Populars: Dades actualitzades el 21 de novembre del 2026

Categories

Internet1.421
Comunicació329
Àudio i vídeo938
Gràfics i disseny566
Ofimàtica i productivitat1.184
Eines per a desenvolupadors3.288
Eines d’IA610
Seguretat i privadesa408
Eines del sistema3.194
Jocs545
Educació i ciència242
Xarxa i TI294

oh-my-pi: baixar versions anteriors

⌥ AI Coding agent for the terminal — hash-anchored edits, optimized tool harness, LSP, Python, browser, subagents, and more

Can Bölük Assistents d’IA i LLM #10 Assistents d’IA i LLM

Aquí hi ha 290 versions d’oh-my-pi, des de 22 de maig del 2026 fins a l’actual, 19.0.1. Cadascuna enllaça amb els instal·ladors del mateix desenvolupador.

2026: 290 290 26
VersióDataMidaFitxers
19.0.1Darrera versió 19 de novembre del 2026 — 64 bits Portable
19.0.0 19 de novembre del 2026 — 64 bits Portable
18.22.2 18 de novembre del 2026 — 64 bits Portable
18.22.0 16 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed slow streaming of Gemma tool calls with large arguments.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Moved binary analysis from the standalone ida and ghidra tools into the backend-neutral dsm namespace (xd://eval/dsm).
- Replaced the ida.enabled, ida.maxOpen, and ida.idleCloseSec settings with disassembler.backend, disassembler.maxOpen, and disassembler.idleCloseSec; shared disassembler database APIs are now available under @oh-my-pi/pi-coding-agent/disasm.
Added
- Added Tern plan review editing and comments: double-click plans to edit them, select text to add Google Docs-style comments, and send comments to Refine.
- Added Ghidra support with automatic setup of Ghidra, JDK, and PyGhidra when analysis is explicitly requested; Auto mode uses an installed IDA Pro when available and otherwise falls back to Ghidra.
- Added support for sharing live analysis databases with child agents through db.handle() and dsm.open(handle).
- Added a Force Remove option to worktree.onExit that removes session-created worktrees without prompting, including worktrees with uncommitted changes; new commits remain on the worktree branch.
Changed
- Updated Tern plan review with a single Approve control, comment-aware Refine option, model selection, and a Save action that writes the plan while keeping review open and plan mode active.
- Plain executable reads now display Mach-O, ELF, and PE metadata without opening an analysis database or downloading a backend; request an overview, function details, or dsm.open() to begin analysis.
- Improved the model switcher (Alt+P / /switch) with provider filtering and per-switch thinking-level selection. In Tern, it now opens above the composer with integrated search, provider and Task/Main filters, and a thinking-level slider.
Fixed
- Deleting or moving large directory trees inside a project no longer stalls language-server file watching (#19490 by @H4vC).
- Reduced CPU and memory usage for long-running subagents as their sessions grow (#19486 by @H4vC).
- Fixed commit.draft() incorrectly classifying files deleted or renamed through bash or eval as “Not yours”; they now appear under Review.
Removed
- Removed the legacy standalone disassembler tools and their associated prompt instructions.
- Removed the Copy and Edit-in-editor buttons from the Tern plan review sheet; use selection copying and double-click editing instead.
@oh-my-pi/collab-web
Fixed
- Improved streaming performance in the collab viewer, preventing long replies from becoming progressively slower as they are received.
@oh-my-pi/pi-natives
Breaking Changes
- TtyWriter now requires a drain threshold and callback and only supports independently reopenable TTYs. Its pending UTF-8 byte bounds are reported conservatively.
Changed
- Large terminal repaints are handled more efficiently, reducing JavaScript-thread work and releasing oversized buffers after output drains.
Fixed
- Fixed stalled terminal writers continuing to emit output or leaking file descriptors after shutdown.
- Fixed terminal output preserving Unicode across chunk boundaries and safely replacing malformed surrogate sequences.
@oh-my-pi/pi-tui
Breaking Changes
- Terminals that support output backpressure must now implement onOutputDrain(callback), which returns an unsubscribe function; terminals without backpressure can omit both pendingOutputBytes and onOutputDrain.
- Updated PlanReviewOverlay options to use PlanReviewOverlayChoice objects and report choice IDs to onPick and selectOption. Refine and Save actions now reflect plan feedback and saving behavior; removed disabledIndices and externalEditorName, and updated describe to receive DescribeContext.
Added
- Added configurable model search placement, provider scopes, and an optional thinking-level slider to ModelPickerComponent; selected thinking levels are reported through onPick metadata. The native picker sheet now supports anchoring, scope styling, level display, and icon actions.
- Added a native plan
18.21.0 16 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added automatic detection of thinking capabilities for previously unknown models, with persistent results and support for alternate effort and token-budget request formats.
@oh-my-pi/pi-catalog
Added
- Added Modal as a hosted inference provider.
Changed
- Changed the local speech model from Kokoro to Phonon 31M.
- Improved reasoning capability detection by preserving unverified thinking-level support and using endpoint-specific probe results when available.
Fixed
- Fixed xAI Grok OAuth model grok-composer-2.5-fast being incorrectly identified as non-reasoning; it now supports thinking levels from minimal through xhigh.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Updated the AgentSession run lifecycle API: agent_end is emitted once when a run can no longer resume, and agent_pause reports runs that stop while awaiting work. Replaced hasPendingAsyncWork(), settleAsyncWork(), and suppressAutoResumeUntilPrompt() with isRunActive, wakeSources, and waitForRunEnd(). Calling abort() now also ends runs paused on background work; RPC, JSON print, collaboration, and extension streams retain their existing per-loop agent_end behavior.
Added
- Added configurable automatic subagent prewalk through task.agentPrewalkAuto, allowing per-agent on/off settings or model patterns while preserving unconditional subagent prewalk.
- Unknown custom and discovered models now automatically detect reasoning support and remember supported reasoning effort levels for future sessions.
- Added inline image and web-page replies: supported terminals can display local, data, and HTTP(S) images and embedded pages, while other terminals receive linked fallbacks. Added tui.renderEmbeds to disable this rendering.
- Added interactive embedded pages in Tern replies, including click-to-visit and Escape-to-return navigation; common host, path, and file: URL forms are accepted.
- Bash redirects can now write to all writable internal URLs, including job cancellation endpoints and handler-backed agent://, cfg://, ssh://, and xd:// URLs.
- Added image pasting to ask form notes and ui.formDialog options via acceptImages; pasted images are returned with [Image #N] markers and described for text-only models. Unsupported prompts now clearly reject image pastes.
Changed
- Changed the default local text-to-speech model to Phonon 31M for faster synthesis, smaller downloads, and additional voices. Existing Kokoro voice settings migrate automatically; speaking rates now range from 0.8× to 1.3×.
- Tern /btw responses now appear as folded, interactive notes at the point where they were requested, with follow-up, copy, branch, inject, and stop actions; /btw without a prompt still opens the history view.
- Overlapping Bash calls now share shell state, including variables, functions, aliases, options, and traps, and background jobs started by those calls continue running independently.
- Extensions enabled or reconfigured through extensions or disabledExtensions now load into the active session without requiring a restart.
- Rich-text pastes made during approval steering or in dialogs now preserve Markdown formatting and images when inserted into the chat editor.
- Moving a task or resuming a session in another project now switches its agents, extensions, custom tools, and MCP servers to match the destination project without a restart.
Fixed
- Fixed status indicators incorrectly showing the agent as idle while subagents or background work could still resume the run.
- Fixed subagents and security scans completing while child agents were still running or queued peer messages remained.
- Fixed conflict-region resolution when a selected side is empty or shares edge lines with surrounding code.
- MCP stdio server failures before a response now report the process exit code or terminating signal instead of only reporting closed stdout.
- Fixed prewalk handoffs so model and reasoning-effort changes are preserved, and automatic handoffs return to the planning model when a
18.20.3 16 de novembre del 2026 — 64 bits Portable
18.20.2 14 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added the Antigravity CLI ACP provider with managed installation, Google sign-in, refreshable distributed credentials, host tools, and retained conversations across tool steps and history changes.
@oh-my-pi/pi-catalog
Added
- Added Antigravity CLI models for Gemini, Claude and GPT-OSS with selectable native thinking levels.
@oh-my-pi/pi-coding-agent
Added
- Added the antigravity-cli provider for Google Antigravity over ACP, with managed binary installation, /login antigravity-cli Google sign-in, host tool execution, and retained conversations across tool steps and history changes.
@oh-my-pi/pi-utils
Added
- Added a profile-aware data directory for managed Antigravity CLI installations and isolated account state.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.20.1...v18.20.2
18.20.1 14 de novembre del 2026 — 64 bits Portable
18.20.0 12 de novembre del 2026 — 64 bits Portable
18.19.0 12 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added the liquid-systemone judgment API for Liquid AI decision models, including validation that rejects image sets exceeding Liquid’s 10,000 32×32-patch request limit before submission.
Fixed
- Fixed Gemini CLI and Antigravity responses losing or misordering thought signatures and reply text, including responses that follow tool calls.
- Fixed Gemini CLI and Antigravity merging reply text around a thought signature, dropping signatures at the start of a reply or after a tool call, and placing a thought ahead of reply text that streamed before it (#18942 by @JustasMasiulis)
- Fixed Google Antigravity judging Gemini models by quota buckets they do not consume: Gemini 3.x requests ignore the Gemini 4 Argon weekly bucket, and Argon requests ignore gemini-weekly, so an exhausted or low bucket no longer blocks, reserves, or caps an account for the other models (#16104).
@oh-my-pi/pi-catalog
Added
- Added Liquid AI provider support with the d1.1 and d1-ultra judge models and Liquid authentication via /login liquid using LIQUID_API_KEY.
Fixed
- Fixed Google Antigravity Gemini 4 Argon models sharing the general Gemini quota scope instead of their dedicated weekly bucket (#16104).
@oh-my-pi/pi-coding-agent
Breaking Changes
- Changed the browser open API: ignore_https_errors is now insecure, allow_file_access has been removed, and file: pages can access other local files.
- Changed the browser open and targets app option from an object to a string or CDP endpoint/executable path. Use argv instead of app.args and the top-level target instead of app.target.
- Removed task.file and the task tool's taskFile field; subagent instructions must now be provided inline.
- Removed blocked todo tasks and the blocked status/blocker metadata. Existing blocked tasks are restored as pending.
Added
- Added systemPromptTemplates, which maps model-selector patterns to Handlebars template files for customizing system prompts. Templates re-render when switching models, with warnings and fallback to the bundled prompt for missing or invalid templates.
- Added ordered multi-step tasks through task.steps. Subagents receive steps sequentially, and the final step determines the task result; interrupted tasks resume with their remaining steps.
- Added Liquid AI judge models liquid/d1.1 and liquid/d1-ultra, including image resizing for Liquid's request limits.
- Added per-step progress, state, and duration reporting for GitHub Actions watches, with expandable details and native Tern Surface Protocol views.
- Added task.autoParallel (off, suggest, or on) to identify subtasks that can run concurrently and optionally delegate them in a single parallel task call.
- Added portable awk, tar, lsof, shasum, md5, and strings support to the bash tool, including Windows support and internal-URI paths.
Changed
- Browser sessions now prefer the bundled ABKit WebKit engine when available, with automatic installation and SHA-256 verification. WebKit provides headed, user-visible tabs and broader browser compatibility; Chromium remains the fallback and can be selected explicitly with app: "chromium", while app: "webkit" requires WebKit. Configure this behavior with browser.webkit and browser.webkitPath or the corresponding environment variables.
- Browser visibility controls now show or hide WebKit tab windows immediately. Headed WebKit tabs can be interacted with directly, while Chromium tabs retain their existing mode-switching behavior.
- In Tern environments with native DOM access, browser observation and interaction use trusted native input across frames and closed shadow roots, with caret-aware typing and case-insensitive text selectors.
- edit.modelVariants now uses the longest matching model pattern, so specific patterns take precedence over generic ones regardless of configuration order.
- The ask tool now submits the form's current values when its timeout expires; updated settings and documentation reflect this behavior.
- The read tool loads det
18.18.9 11 de novembre del 2026 — 64 bits Portable
18.18.8 11 de novembre del 2026 — 64 bits Portable
18.18.7 10 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- LiteLLM usage adapter (@oh-my-pi/pi-ai/usage/litellm) reporting a gateway's key and user USD budgets (#17873 by @maykonmichel)
- UsageApi.replaceConfiguredProviders registers configuration-sourced usage providers between runtime overrides and the built-in resolver, and UsageProvider.usesRequestKey makes aggregate usage probe only the key inference would send (#17873 by @maykonmichel)
Changed
- Providers with a login-warning auth rule show their account risk before a new login and start it only after the user answers yes; any other answer or a cancel stores nothing, and their login-list entry is marked "(at your own risk)" (#18499)
Fixed
- Claude Fable launch resets now report the Fable weekly limit they clear instead of an unknown window (#18661 by @TheAifam5)
@oh-my-pi/pi-catalog
Added
- Added Muse Ember (muse-ember-1.0) on Meta Model API and Muse Code with the minimal–max reasoning effort ladder and its >200K long-context pricing
- Added the login-warning "<text>" auth rule, which makes a new login accept the stated account risk first (#18499)
Changed
- Resumed Google Antigravity behind a login warning instead of pausing it (#18499)
Fixed
- Morph (api.morphllm.com) requests now send x-session-id, keeping a conversation on the worker that holds its prompt cache (#18704 by @wolfiesch)
@oh-my-pi/pi-coding-agent
Added
- /usage and omp usage show LiteLLM key and user budgets for the built-in litellm provider and every models.yml provider with discovery.type: litellm, unless listed in disabledProviders (#17873 by @maykonmichel)
- Extensions can ask the operator before an interactive quit, or cancel it, with session_before_shutdown; a further /exit or Ctrl+D skips the pending prompts and closes (#18424 by @aw1cks)
Changed
- Saved-reset auto-redeem is set per window: codexResets.autoRedeem5h/autoRedeemWeekly and claudeResets.autoRedeem5h/autoRedeemWeekly/autoRedeemFable replace each provider's autoRedeem, whose value migrates into the 5h and weekly keys; a Claude Fable launch reset is governed by autoRedeemFable alone and spent only for a block that includes the Fable weekly cap, or before expiry when Fable usage is above zero (#18661 by @TheAifam5)
- Sessions without a prompt UI (such as omp -p) no longer spend saved resets while auto-redeem is unset, even ones about to expire (#18661)
- Google Antigravity is back in /login and omp login, marked "(at your own risk)": a new login first explains that Google may suspend accounts that sign in from third-party tools and continues only after you type yes; saved Antigravity logins work again (#18499)
- On Windows, the bash tool's description tells the model it runs in an emulated bash environment, not cmd or PowerShell, so it writes bash instead of dir, %VAR% and other cmd syntax
Fixed
- Extension prompts shown from session_before_* handlers no longer time out after 30 seconds while the operator is still answering (#18424 by @aw1cks)
- A second exit request now skips pending worktree.onExit prompts and keeps those worktrees instead of waiting for an answer (#18424 by @aw1cks)
- Advice that makes the agent keep working after it finished no longer comes back a second time, often as a blocker, while the agent is fixing it: advisor reviews of the earlier answer are cancelled and the advisor reviews the fix instead (#18735 by @H4vC)
- On Windows, the bash tool's shell resolves MSYS drive paths (/c/Users/..., /mnt/c/...) everywhere, so running /c/.../tool.exe, cd /c/..., [ -f /c/... ] and < /c/... reach drive C: like they already did for ls and other built-in utilities
- On Windows, cmd-style command lines such as dir | find /c /v "", sort /r, timeout /t 5 and whoami /all in the bash tool run the Windows programs they were written for, instead of pi's GNU-style find/sort/timeout/whoami (which read /c as drive C: and could walk the whole drive)
- Changing or unsetting PATH in the bash tool's shell now forgets remembered command locations, as bash does, so a command no lon
18.18.6 9 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Fixed
- LiteLLM lists only the models the configured key can call (its authenticated /v1/models), keeping the proxy's metadata for them; a proxy whose /v1/models fails now lists no models instead of every model it serves (#18601)
@oh-my-pi/pi-coding-agent
Added
- In Tern, a turn's long-press menu on a phone can edit and resend a prompt, retry the turn, open the rewind page on it, or fork the session from it into a new session
- A rewind from the transcript says how many turns it dropped and offers Undo on its toast in Tern, which puts the turns and your draft back
Changed
- Browser role/ selectors in Chrome tabs now match the role and name Chrome's accessibility tree reports, so they reach open shadow roots and skip hidden elements; in Tern every name observe() reports selects its element again (#18632 by @will-bogusz)
Fixed
- A late usage-limit error from a previous provider no longer blocks or cools down the selected provider; when the old provider has spent its retry budget, a replay-safe turn still tries the selected model without the old quota delay (#17098, #17571 by @Bensonluo)
- Multi-question ask forms in Tern show Submit on the last question tab, not only on the Review tab
- MCP headers and stdio env values that use !command now run again on a 401/403 (the request is retried once), on /mcp reconnect and other reconnects, and on /mcp reload, so servers whose command mints short-lived tokens keep working past expiry instead of staying 401 until restart (#18631)
- Pausing an unapproved plan no longer tells the agent that the draft was approved or to continue executing it, and pausing with /plan closes an open plan review so it can no longer save and execute the plan (#18610).
- Turning off lsp.enabled now also stops write and edit from starting language servers for diagnostics or formatting; previously only the lsp tool was disabled (#18205, #18490 by @yuzu-octopus)
- Fixed background tabs omp borrows (relay app.target, app.cdp_url) ignoring omp's clicks and keys or timing out on hover and drag; the page reads as visible while omp drives it and hidden again once released (#17831 by @will-bogusz)
- Fixed the transcript jumping at the screen's top edge while scrolled back during a streamed reply, and the screen dropping two rows when a reply finished (#12307, #16117 by @galtamam-cmyk)
- Focusing a subagent or returning to the main session no longer loses or freezes the next assistant reply when a fast tool turn finishes before the previous reply's links finish resolving, and messages and tool runs that arrive during the switch show exactly once, in order (#18564 by @eggpeat)
- omp gc --strip now keeps every response of the latest replay turn (including earlier tool-call steps joined by their tool results), reference-only images, and the transcript's file permissions (#18636 by @eggpeat)
- Session files whose entries were written without ids are re-indexed when loaded, and migrating old sessions no longer gives two entries the same id (#18636 by @eggpeat)
- wait and write proc://<id>/kill now say how a settled job's result reached the caller: delivered by the call itself (it finished while the call ran), handed over by this snapshot (it superseded the job's queued completion push, so no duplicate follows), or genuinely late — the last keeps its "not auto-delivered; recovered by this snapshot" warning and adds how long before the call it had finished (#16052, #18635 by @yuzu-octopus)
- Fixed ask notes added with n right after picking an answer landing on the previous question instead of the one shown; a Tern click then Note still notes the answer just clicked, and the Note button names it (#18317 by @DonMikone)
@oh-my-pi/pi-natives
Fixed
- Fixed GTK4 apps such as Files (Nautilus) on X11: element clicks and computer.elementAt() now find their elements where they are on screen, and ax() no longer marks their enabled controls (disabled) (#18643 by @will-bogusz)
- grep no longer searches the file its own ou
18.18.5 9 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Added
- Added Claude Fable 5.2 and Claude Mythos 5.2 ($8/$40 per MTok, $0.20 cache reads, 1M context)
Fixed
- Fixed newly discovered Anthropic Fable/Mythos models showing as text-only, non-reasoning, and mispriced before models.dev catalogs them
- Fixed Claude Mythos 5.1 not using thinking prefix-binding controls, which it needs on every account
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.18.4...v18.18.5
18.18.4 9 de novembre del 2026 — 64 bits Portable
18.18.3 9 de novembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Changed
- Added a balanced delegation bias between eager and restrained; models without a delegation-bias rule now resolve to it instead of eager
@oh-my-pi/pi-coding-agent
Breaking Changes
- The ask tool takes one HTML form ({ html }) instead of questions[]: the model writes a <form> of radios, checkboxes, selects, text inputs and textareas with its recommendation preselected, and receives the submitted form data as JSON. AskToolDetails is now { answers: { name, question, answer }[]; timedOut? }; QuestionResult, the "Other", "Chat about this", answer-note and answer-image paths of the tool are removed, and /tree re-answer reopens the original form (NavigateTreeResult.reopenAsk carries html).
Added
- In Tern, right-clicking an ask answer offers Note, which annotates that answer without picking it (#18551 by @H4vC)
- task.eager (preferred/always) now shapes only the main agent's system prompt; subagents fall back to their model's own delegation policy instead of being pushed to spawn further subagents
- ask forms can show pictures: <img src> from an http(s) URL, a data: URI or a local file (relative to the session's working directory), and inline <svg> diagrams, drawn in the terminal form and natively in Tern
- ask forms give every radio and checkbox group a free-text "Other" answer: its text is the radio's value or is appended to the checkbox group's array (field-by-field fallback, collab guests and ACP elicitation included)
- Extension UI gains optional ui.formDialog(html, options), which renders an HTML <form> and resolves to { data, timedOut? } (undefined when cancelled). The interactive TUI renders it as a form dialog, ACP clients with form elicitation get one elicitation with a property per field, and hosts without it (RPC) answer the ask tool field by field through select/input/editor.
- Eval cells can run shell commands with sh: sh("git status") in Python, await sh`ls ${dir}` or await sh("make", { cwd }) in JavaScript. Commands run on the embedded bash engine to completion (never auto-backgrounded), return their output, and raise ShellError with the exit code on failure.
- omp zsh on Windows runs omp's builtin utilities (ls, cat, head, jq, …) for commands PATH does not have, unless a command_not_found_handler handles them (#18532 by @H4vC).
Changed
- Clarified tool and internal-URL descriptions for history:// concise transcripts, proc:// job retention and parked-agent lookups, and read JSON query text output (#18553 by @xiangnan0811)
- Models without a delegation rule (Claude, Gemini, …) no longer get pushed to delegate: the "map unknown code via task" rule and the "Task + scout" nudges in search tool descriptions are replaced by a light suggestion to explore unfamiliar subsystems with a scout
- In Tern, clicking an ask answer and then Note now annotates the answer highlighted on the question shown instead of the one just clicked (#18551 by @H4vC)
Fixed
- Fixed ask notes added with n right after picking an answer landing on the previous question instead of the one shown; a Tern click then Note still notes the answer just clicked, and the Note button names it (#18317 by @DonMikone)
- Fixed /export failing on Windows when saving in the current directory, including relative paths such as sub/../session.html; the export message now shows the full path of the written file (#18369, #18525 by @minhngoc25a).
- Todo initialization now names empty phases Tasks, so subsequent updates can address them (#18552 by @xiangnan0811)
- omp zsh and the bash tool on Windows: background commands without job control ignore Ctrl-C (INT, and QUIT in zsh) as zsh and bash do, so the console's Ctrl-C leaves them running (#18532 by @H4vC).
- omp zsh finds the system zsh's function library (compinit loads) on Debian, Ubuntu and Arch, and on Windows from an MSYS2, Cygwin or Git for Windows zsh (#18532 by @H4vC).
- omp zsh on Windows: typed commands no longer end in ^M, programs are found through Path, piped scripts run every line,
18.18.2 8 de novembre del 2026 — 64 bits Portable
18.18.1 8 de novembre del 2026 — 64 bits Portable
18.18.0 7 de novembre del 2026 — 64 bits Portable
18.17.0 6 de novembre del 2026 — 64 bits Portable
18.16.0 2 de novembre del 2026 — 64 bits Portable
18.15.0 1 de novembre del 2026 — 64 bits Portable
18.14.2 31 d’octubre del 2026 — 64 bits Portable
18.14.1 31 d’octubre del 2026 — 64 bits Portable
18.14.0 29 d’octubre del 2026 — 64 bits Portable
18.13.3 28 d’octubre del 2026 — 64 bits Portable
18.13.2 27 d’octubre del 2026 — 64 bits Portable
18.13.1 27 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Hosts can distinguish safely paused agents from turns still draining with Agent.isPaused (#17073 by @anatoli-tsinovoy).
@oh-my-pi/pi-ai
Fixed
- Codex accounts whose spend cap is used up are no longer picked through credit overage or from usage cached before the upgrade (#16973 by @DusKing1)
- Fixed Codex capacity 429s (turn_rate_limit_exceeded, "Codex is at capacity") reporting "Try again in ~N min" from the plan's usage-window reset, which made retries wait hours instead of the response's Retry-After (#17087 by @H4vC)
@oh-my-pi/pi-catalog
Added
- Added GPT-6.1 Terra: openai/gpt-6.1-terra ships bundled ($1/$5, $0.05 cached, >272K long-context tier, 1.05M window), and Codex-discovered gpt-6.1-terra/-wm get API-equivalent pricing, the 922K extended-context maximum, and routing to Plus-or-higher accounts.
- Added Kimi K3.1 Pro and K3.1 Pro Highspeed support on Kimi Code (k3.1-pro, k3.1-pro-highspeed) and the Moonshot API (kimi-k3.1-pro, kimi-k3.1-pro-highspeed): official pricing ($2/$10 and $4/$20 per MTok), a 131K output cap, and K3's reasoning and wire contract (#16901 by @eggpeat)
@oh-my-pi/pi-coding-agent
Added
- Added tools.restrictedFolders: every action on a listed folder (the folder and everything below it) asks for an explicit Approve/Deny in every approval mode, yolo included, and no remembered grant, session approval, or ACP "Always allow" covers it; recursive searches rooted above a restricted folder ask too, commands are matched by the directory they run in, and a deny policy still denies (#16049 by @Peribunt)
- /pause shows draining turns and running jobs across the main agent's subagent tree, without waiting on finished turns, safely parked agents, or suspended advisor reviews (#17073 by @anatoli-tsinovoy).
Changed
- Reduced redraws during background subagent progress while keeping elapsed previews and native checklists current (#17131 by @iliaal, based on #16647 by @Komzpa).
- Restoring a session with many subagents refreshes the running-subagent badge and lists once per burst instead of once per agent (#16642 by @Komzpa)
- ask answers now name the option a note is on (m: [X, Y] (note on Y: …), User added note on Y: …) (#17126 by @H4vC)
Fixed
- Fixed advisors dropping their review and showing a rate-limit error when the provider returns a timed 429 such as Codex's turn_rate_limit_exceeded; the advisor now waits out the Retry-After (within retry.maxDelayMs) and retries like the main agent (#17087 by @H4vC)
- Fixed a hashline PUT being refused with "would hold one more copy of line N … than the range did" when its body restates the line next to the range, fills the range, and adds a new line equal to that neighbour (a second @property or #[test], or a break; above a new case); it applies again, dropping only the restated line (#17028)
- Fixed task subagents receiving AGENTS.md files, including the user-level ~/.omp/agent/AGENTS.md, when the session has additional workspace roots (--add-dir, /add-dir, workspace.additionalDirectories)
- Fixed an ask note added right after picking an answer landing on the next question's answer, in Tern and the TUI; Note now goes to the answer just picked until the next question gets other input (#17126 by @H4vC)
@oh-my-pi/pi-tui
Added
- ANSI and TSP /pause screens show unsettled turns and jobs, hide settled counters, and preserve counts and resume controls on short terminals (#17073 by @anatoli-tsinovoy).
- Added noteOption to ExtensionAskDialogResultItem and QuestionResult: the label of the option an ask note is on (unset for a note on the custom answer) (#17126 by @H4vC)
Changed
- StatusLineComponent.setRunningSubagents() and setSubagentTreeCost() now report whether their state changed, allowing callers to skip redundant redraws (#17131 by @iliaal).
Fixed
- Fixed the ask dialog's Note (button or n) right after a pick moved to the next question annotating that question's highlighted row instead of the answer just picked (#17126 by @H4vC)
- Fixed the as
18.13.0 26 d’octubre del 2026 — 64 bits Portable
18.12.6 21 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Added
- Added the prompt-cache-persistent catalog axis (isPromptCachePersistent) for providers whose own endpoint never expires prompt caches; logged-in OpenAI Codex declares it
Changed
- Updated the bundled Yolo-Auto catalog to its current models, qwen3.8-27b (now the default), yolo, yolo-small, and yolo-128k, and removed the retired deepseek-flash-v4 and qwen3.8-flash
@oh-my-pi/pi-coding-agent
Added
- Added the _session/steering ACP extension (advertised as _meta.steering.supported), so clients such as agent-shell can steer a running turn; a steer sent while a permission dialog is open cancels that dialog, and _meta.steering.queue: "followUp" queues a follow-up instead (#15885 by @marcintustin)
- /mcp reauth <name> now works over ACP, RPC, and print mode (omp -p '/mcp reauth <name>'): it opens the browser, prints the authorization URL, reconnects the server in the running session, and print mode exits 1 when reauthorization fails (#16046 by @marcintustin)
Changed
- ACP session/cancel now also cancels prompts that are queued or still dispatching, and stops autonomous runs, instead of being ignored when no client turn is streaming (#15885 by @marcintustin)
- Idle compaction no longer runs on logged-in OpenAI Codex accounts, whose prompt cache never expires; compacting while idle only threw away a warm cache
- ACP sessions no longer fail to start when a client-declared MCP server cannot connect; the failure is reported in the session, and an HTTP/SSE server can be repaired with /mcp reauth (#16046 by @marcintustin)
- /mcp reauth, /mcp test, and /mcp unauth now pick the project entry when a server is defined in both project and user config, and skip project config when mcp.enableProjectConfig is off (#16046 by @marcintustin)
- /mcp reauth removes the server's stale Authorization header from its config when it writes the new OAuth binding (#16046 by @marcintustin)
- omp -p '/compact', /handoff, and /retry given as CLI arguments now run the command instead of sending the text to the model, and exit 1 when the command fails; slash text arriving on stdin or from @file is still sent to the model (#16387 by @marcintustin)
- Interactive startup no longer waits for extensions that register only tools, commands, flags, renderers, or event hooks (for example MCP bridge plugins): they load once the composer accepts input, and your first message waits for them instead. An extension is deferred only after one launch has confirmed it registers no providers, shortcuts, composer shapes, or file fallbacks. Launches that resume a session, pass --tools, or pass flags omp does not recognize still load every extension first (#16684 by @H4vC)
Fixed
- Fixed subagents whose role has an explicitly empty fallback chain falling back through the default chain when the default role uses the same model, including after a retry fallback, prewalk or context promotion briefly switched models or the subagent was revived (#16548 by @eggpeat)
- Fixed /usage and omp usage --json listing identity-less OAuth accounts such as Devin as missing usage beside their working reports (#16549 by @eggpeat)
- Fixed releasing or reaping a browser tab returning while Chromium still listed the page; the close now waits until Chromium reports the tab destroyed (#16680 by @H4vC)
- Fixed the advisor prompt forbidding all repeated advice, including real escalations: advisors may now re-raise a prior nit or concern as a blocker once new evidence makes it one (#13887, #16484 by @pjh890321)
- Fixed Esc no longer interrupting in the TUI after a second MCP OAuth login replaced an unfinished first one (#16046 by @marcintustin)
@oh-my-pi/pi-tui
Changed
- In Tern, centered sheets (/context, reports, keyboard shortcuts, BTW history, session info, the pause screen) fill the pane when it is narrower than 100 columns instead of floating with margins (#16676 by @H4vC)
Fixed
- Fixed the /agents and /models hubs in Tern squeezing the left sidebar as long lines (such as a streami
18.12.5 21 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added OpenAI 12-hour prompt caching for GPT-5.6+ on the Responses API: cacheRetention: "long" now requests prompt_cache_options.ttl: "12h" (instead of the ignored legacy prompt_cache_retention: "24h"), and promptCache.ttl accepts "12h"
Fixed
- Fixed every OpenAI Codex request failing with HTTP 404 {"detail":"Not Found"} on personal ChatGPT accounts whose token is not region-pinned (#16493)
@oh-my-pi/pi-catalog
Fixed
- Fixed Moonshot kimi-k3 pricing cache writes at $0; usage cost now includes cache writes at Moonshot's $3/M 5-minute rate (#16508)
@oh-my-pi/pi-coding-agent
Added
- providers.cacheRetention: long now keeps OpenAI GPT-5.6+ prompt caches for 12 hours (Responses API; cache writes cost 2.5× input instead of 1.25×)
Fixed
- Fixed subagents spawning at a caller-supplied effort (e.g. hi → max) while task.enableEffort is off; the field is now ignored and the agent's configured thinking level (including :auto) applies (#16402)
- Fixed omp update --check failing against JFrog Artifactory registries that add empty dist-tags to a tagged manifest; the update check now uses the registry's curated latest version (#16375)
- Fixed <path>:conflicts on a file past the 10 MiB scan cap reporting no conflicts with no hint: it now names the line the scan stopped at and the ranged read that registers the blocks after it (#16269)
- Fixed the ts-set-map rule stalling tool-call streaming for seconds on a long single-line Map literal, and missing Map entries written ["key", value] as const (#16242)
- Fixed the ts-import-type rule missing inline type imports whose member name is non-ASCII, such as import("./units").温度 (#16242)
@oh-my-pi/pi-natives
Fixed
- Fixed a macOS background action leaving its target in front of your app after you clicked another of your windows, when placing the click took longer than the click itself (#16217)
What's Changed
- fix(update): handled empty Artifactory dist-tags on tagged manifests by @roboomp in https://github.com/can1357/oh-my-pi/pull/16379
- fix(task): ignored caller effort while task.enableEffort is off by @roboomp in https://github.com/can1357/oh-my-pi/pull/16406
- fix(catalog): stopped declaring no_constraint codex residency by @roboomp in https://github.com/can1357/oh-my-pi/pull/16501
- fix(catalog): priced moonshot kimi-k3 cache writes at the 5min rate by @roboomp in https://github.com/can1357/oh-my-pi/pull/16529
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.12.4...v18.12.5
18.12.4 20 d’octubre del 2026 — 64 bits Portable
18.12.3 19 d’octubre del 2026 — 64 bits Portable
18.12.2 18 d’octubre del 2026 — 64 bits Portable
18.12.1 17 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed every request failing behind a TLS-inspecting proxy that refuses compressed uploads: a 400/403/415 whose body is not JSON, or the Anthropic API's "request body is not valid JSON" after a proxy drops Content-Encoding, now resends the body uncompressed and stops compressing for that host (#16108 by @alphastorm)
@oh-my-pi/pi-catalog
Fixed
- Anthropic's request-encoding-rejection now also matches the API's "request body is not valid JSON" 400, which a gzip body gets when a proxy drops its Content-Encoding header (#16108 by @alphastorm)
@oh-my-pi/pi-coding-agent
Breaking Changes
- Removed display() from eval in both Python and JavaScript; print (and console.log in JavaScript) covers it
Added
- Added advisor.continueSeverity (nit, concern, blocker) to choose the minimum advisor note severity that wakes the agent after it finished with a final answer; lower-severity notes stay as a card at the end of the transcript
- Printed DataFrames/Series render as tables, large dicts/lists and objects as JSON trees, and Markdown objects as Markdown for the user, while the model sees the printed text
- Printing a figure or image in eval attaches the image to the result
Changed
- An advisor concern raised after the agent's final answer now continues the run by default instead of waiting at the end of the transcript; set advisor.continueSeverity to blocker for the previous behavior
Fixed
- Fixed context-file @ imports failing to load paths containing backslash-escaped spaces (#16090)
@oh-my-pi/pi-natives
Fixed
- Fixed native memory growing without bound, and eventually crashing the process, when files in a downloaded (WASM) grammar language such as C# were parsed repeatedly (#16067)
@oh-my-pi/pi-tui
Breaking Changes
- numberedCode, readContentCode and nativeDiff take the file's path (path) instead of a language, and describeGroupedOutput drops its lang option: Tern picks the grammar of file contents from their path
Changed
- Advisor notes in Tern read as a callout tinted by their worst severity: its icon and a tally (Concern, 1 blocker · 2 concerns) head the card with the advisor's name at the end, and several notes each lead with their own severity icon
Fixed
- Fixed images rendering blank inside tmux on Kitty-capable terminals when tmux's allow-passthrough is off (the default); omp now falls back to SIXEL or the text placeholder instead of sending Kitty graphics tmux drops
- Fixed Tern drawing tool cards' file contents (write, read, edit, grep, find, ast-grep, ast-edit) plain or in the wrong language when omp did not know the file's extension (.ptx) or mapped it to a neighbor (.cu as C++); omp now sends the file's path and Tern picks the grammar
What's Changed
- fix(ai): resend uncompressed when a proxy rejects a compressed request body by @alphastorm in https://github.com/can1357/oh-my-pi/pull/16108
- fix(tui): answered allow-passthrough in fake tmux client fixture by @roboomp in https://github.com/can1357/oh-my-pi/pull/16112
- fix(ast): retired per-thread wasm parser store after a byte budget by @roboomp in https://github.com/can1357/oh-my-pi/pull/16078
- fix(coding-agent): support escaped spaces in context imports by @roboomp in https://github.com/can1357/oh-my-pi/pull/16091
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.12.0...v18.12.1
18.12.0 17 d’octubre del 2026 — 64 bits Portable
18.11.3 16 d’octubre del 2026 — 64 bits Portable
18.11.2 16 d’octubre del 2026 — 64 bits Portable
18.11.1 15 d’octubre del 2026 — 64 bits Portable
18.11.0 14 d’octubre del 2026 — 64 bits Portable
18.10.4 13 d’octubre del 2026 — 64 bits Portable
18.10.3 12 d’octubre del 2026 — 64 bits Portable
18.10.2 12 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added AgentLoopConfig.onDetachedToolUpdate: a tool's onUpdate frames that arrive after its run's event stream ended (a returned call whose background job keeps reporting) reach it instead of being dropped; Agent forwards them to its subscribers as tool_execution_update events
@oh-my-pi/pi-ai
Added
- Added support for CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS on the anthropic provider: when set, requests omit pre-release fields (context_management, beta tool fields such as strict and defer_loading, task budgets, structured-output format, thinking.block_binding, and Anthropic server compaction) along with their beta headers, so ANTHROPIC_BASE_URL gateways that reject them keep working (#15458 by @H4vC)
@oh-my-pi/pi-catalog
Added
- Added the honors-claude-code-beta-kill-switch axis (compat.honorsClaudeCodeBetaKillSwitch), on for the anthropic provider, so a rule or models.yml override decides which providers honor CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS (#15458 by @H4vC)
@oh-my-pi/pi-coding-agent
Added
- Added CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS support: set it to stop Anthropic requests sending experimental beta fields that a company ANTHROPIC_BASE_URL gateway rejects, such as the Sonnet 5.5 thinking.adaptive.block_binding 400 (#15458 by @H4vC)
- Added the task.structured setting (default off) for structured task results: when on, the task tool accepts outputSchema/schemaMode and subagents return results through yield as before
Changed
- Subagents spawned by the task tool now return their final message as plain text, with no yield tool, output schemas, or "Submit Result" fields; eval agent()/workpool spawns still return structured results
- In Tern, a backgrounded bash command or eval cell keeps filling its own card with live output (In background, with a running clock) and settles it with the job's outcome, Cancelled when the job is killed; wait lists each job on one line (type, id, label, duration) instead of repeating its output
- ACP clients now see a backgrounded call's job output stream into its tool call, which settles completed or failed with the job
Fixed
- Fails unstructured task runs that conclude without a final message
- Background task cards no longer stop updating once the agent ends its turn while the task is still running
- Transparent images the model receives (small icons, oversized screenshots) keep their transparency instead of turning into a black square, and resized edges no longer pick up a dark fringe
@oh-my-pi/pi-natives
Added
- Added in-process env, du, chmod, fold, rmdir, od, df, uptime, id, and dd shell builtins matching GNU coreutils 9.x options, output, and exit codes (chmod and id on Unix only; dd is disabled with the destructive set by PI_DISABLE_UUTILS_DESTRUCTIVE).
- Added an in-process nl shell builtin (ported from uutils coreutils 0.8.0, matching GNU nl), with header/body/footer numbering styles including pBRE regex styles, logical-page section delimiters, -l blank-line grouping, and ln/rn/rz number formats.
@oh-my-pi/pi-tui
Added
- Added opt-in /tree branch navigation across visible branches and nested fork depths; bind app.tree.previousBranch / app.tree.nextBranch (e.g. to Shift+Left / Shift+Right) in keybindings.yml (#12623 by @azain47).
Changed
- The wait tool's native job rows (omp.wait.job) are a single row: badge, id and label, progress and clock, without the output preview or capture notice beneath
- The native bash run foot reads In background with a live clock (omp.run.time without stopped) while a parked job's frames stream (details.async.state === "running", partial), instead of Running
- Native model pickers (/model, the quick switch and the advisor's) mark providers with their logo (mark.icon: "provider:<id>", initials where the terminal has none), and every model row leads with its provider's mark
Removed
- Removed the unused app.tree.foldOrUp / app.tree.unfoldOrDown keybinding ids, which no longer did anything.
- Added parksAsBackground(toolName,
18.10.1 12 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Added a Copy button that leads the turn footer under an answer in Tern (⧉ · 2m13s · 71K tok · $0.85 this turn, in place of the clock), which copies that answer
Changed
- Running agents and background jobs in Tern now share one compact counter over the composer (an icon and a count each, the words on hover) instead of two "N running" pills; each count still opens its own sheet
- In Tern, /session pin and /usage reset open the same picker sheet as the other small selectors (the active account checked, a confirm strip before a reset is spent, unusable accounts dimmed with the reason); the thinking level picker shows each level's depth as a meter, the plugin and queue mode pickers use badges (version, installed, scope, recommended) and plain-language names instead of bracketed text, and the theme picker no longer repeats "(current)" beside its check
- In Tern, the text editor an extension opens (and the ask tool's custom answer and note) is now a sheet in the composer's place: the title over its details, the text written straight on the sheet, and Open in editor, Cancel and Submit buttons whose keys show in their tooltips instead of a keycap hint row
Fixed
- Fixed grep on a single file showing in Tern as separate boxed snippets with no file row or syntax colors; it now looks the same as a grep over a directory
@oh-my-pi/pi-tui
Added
- AssistantMessageComponent's native turn footer leads with a Copy icon (omp.turn.copy) and a dot instead of the clock icon; it copies the answer's text and shows when the host handles transcript actions
Changed
- UserMessageComponent's native hover toolbar (omp.user.tools) leads with Copy and Rewind icon buttons (copy, rewind) instead of text labels, then a dot and the time, like the turn usage footer
- ContextUsageView's native body (on terminals with meter) leads with the tokens used and the room left before auto-compaction, then one stacked bar ticked at the threshold, the breakdown as rows (omp.context.cat, .cat.free, .cat.buffer: swatch, name, tokens, share; the exact count in the row's title) and the snapcompact estimate as labelled rows (omp.context.snap); the blocks meter and legend kv are gone, and contextUsageHead no longer repeats the window size
- AskDialogComponent's native root is omp.editor.ask (collapsed: omp.editor.ask.hidden, the countdown, the question as omp.ask.summary, Skip and Show): a chosen answer is the accent-toned row (a check also carries the check icon) instead of the highlighted one; multi-select Other… is omp.ask.other.check; the question tabs check answered questions and say their keys in the strip's title; the countdown moves from the head to the action bar beside Hide; the preview is omp.ask.preview; the Review tab leads with omp.ask.notice and lists answers (omp.ask.answers) by question name with inline code mono and no (Recommended) suffix; option descriptions run inline code mono, notes ride the row's detail as em; Submit/Next's tooltip names only Enter (and Space on a multi-select)
- MCPAddWizard's native sheet asks one question per step: a phase strip (omp.mcpAdd.progress of omp.mcpAdd.phase, toned done/current/upcoming and neutral when no auth was needed), the question and its lead (omp.mcpAdd.title, omp.mcpAdd.tag for optional fields, omp.mcpAdd.lead), then a mono field with a placeholder and its rule or refusal (omp.mcpAdd.field, tone: "error"; omp.mcpAdd.note), choice rows with icons and mono tags (omp.mcpAdd.choices), status lines (omp.mcpAdd.status, omp.mcpAdd.checks, omp.mcpAdd.error) or the review rows (omp.mcpAdd.summary). Cancel, Back and Continue / Save server buttons run Ctrl+C, Esc and Enter instead of a key-hint strip. The connection test shows its own busy step, the OAuth client secret field is masked, and the review saves on Enter (no Yes/No choice)
- ReportPanel's native body carries the role omp.report.body
- HookInputComponent describes a dock sheet (omp.overlay.hook-input col) like HookEditorComponent's: the 
18.9.2 11 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- Improved context caching for repeat turns, reducing redundant prompt processing and making subsequent turns faster when the system prompt is unchanged.
@oh-my-pi/pi-ai
Breaking Changes
- Renamed TypeSafeJudge to NativeJudge and TypeSafeApiError to JudgmentApiError, including their associated option types. JUDGMENT_ROUTES has been removed; use isJudgmentApi or create a judge for catalog models with createNativeJudge(model, { apiKey }).
Added
- Added native judge support for OpenAI’s Decisions API and Cloudflare Workers AI decision models, including the openai-decisions and workers-ai-systemone APIs. The auth gateway’s System One route supports both providers.
- Added createNativeJudge(model, { apiKey, headers, fetch }) for creating judges from catalog models with their configured request model IDs and question limits.
- Added Cloudflare Workers AI authentication with loginCloudflareWorkersAi, storing the API token and account ID; logging in again with the same token replaces the existing credential.
- Added automatic question batching to NativeJudge for endpoints with question limits, with up to four concurrent batches and merged answers and usage reporting. Usage from successful batches is reported through partial-usage callbacks and both session and auth-gateway usage ledgers, even if another batch fails.
Changed
- Improved Anthropic prompt caching for long conversations, reducing memory usage while selecting cache breakpoints.
Fixed
- Fixed Devin-routed OpenAI models returning invalid_argument on every turn after an existing session was switched from openai-codex.
@oh-my-pi/pi-catalog
Removed
- Removed parseCloudflareAiGatewayCredential, serializeCloudflareAiGatewayCredential, and CloudflareAiGatewayCredential from @oh-my-pi/pi-catalog/wire/cloudflare-ai-gateway. Use the equivalents from @oh-my-pi/pi-catalog/wire/cloudflare instead.
@oh-my-pi/pi-coding-agent
Fixed
- Improved responsiveness when adding assistant text after tool calls in long transcripts.
- Fixed apply_patch handling of context-free hunks when the target text appears multiple times.
- Fixed find filename ranking and sketch scoring with judge endpoints that limit the number of questions per request.
- Fixed omp gc --apply leaving temporary session files after an interrupted save.
- Fixed omp auth-gateway serve handling of OpenRouter routing variants such as :floor and :nitro.
- Improved validation errors for computer helpers with missing or non-string text arguments by identifying the helper and argument involved.
- Fixed /tan forks incorrectly inheriting the parent task's subagent spending while preserving token counts.
- Fixed omp:// documentation links so they open the corresponding files in source checkouts.
- Fixed bare model IDs in models.yml incorrectly resolving to auto-discovered Ollama, llama.cpp, or LM Studio providers.
- Added warnings when configured Ollama, llama.cpp, or LM Studio endpoints cannot be reached, identifying the relevant environment variable.
@oh-my-pi/pi-natives
Fixed
- Fixed stalls when highlighting Markdown tool previews containing many inline spans or backslash escapes.
@oh-my-pi/pi-tui
Added
- Added native TSP rendering support for Tern terminal sessions running inside tmux, with automatic session detection and safe title-setting behavior.
Fixed
- Improved responsiveness when inserting assistant messages after tool calls in long transcripts.
@oh-my-pi/pi-utils
Changed
- Improved streaming JSON parsing performance for tool arguments containing many string values.
Fixed
- Fixed parsing of long strings that contain closing delimiters other than double or single quotes.
What's Changed
- fix(natives): register corrected Markdown grammar at build time by @iliaal in https://github.com/can1357/oh-my-pi/pull/15272
- perf(natives): vendor the Markdown grammar without exponential backtracking by @sjawhar in https://github.com/can1357/oh-my-pi/pull/13838
- perf(utils): avoid bulk scan setup for short escaped
18.9.1 11 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added compat.supportsThinkingBindingControls for Bedrock Converse: setting it to false stops prefix-bound Claude requests from sending thinking.block_binding, the thinking-binding-controls-2026-08-01 beta and the /input_transformations response path, for gateways or regions that reject them (#15316 by @harshaygadekar)
Fixed
- Fixed Bedrock Claude Fable 5.1 requests with thinking off and a forced tool choice sending the forced choice next to adaptive thinking; the choice now falls back to auto (#15316 by @harshaygadekar)
@oh-my-pi/pi-catalog
Added
- Added the supports-thinking-binding-controls axis to Bedrock Converse models (on by default) so a rule or models.yml override can turn off Claude thinking-binding controls on hosts that reject them (#15316 by @harshaygadekar)
@oh-my-pi/pi-coding-agent
Added
- Added supportsThinkingBindingControls to Bedrock compat in models.yml, so a Bedrock gateway or region that rejects Claude thinking-binding controls can turn them off (#15316 by @harshaygadekar)
@oh-my-pi/pi-natives
Fixed
- Fixed executable scripts without a #! line failing with "Exec format error" in the embedded shell (#15271 by @iliaal).
@oh-my-pi/pi-tui
Fixed
- Fixed inline charts and images leaving overlapping copies over text when moved, resized, replaced, or hidden, while preserving images in scrollback.
What's Changed
- fix(bedrock): add supportsThinkingBindingControls catalog opt-out by @harshaygadekar in https://github.com/can1357/oh-my-pi/pull/15316
- fix(shell): preserve executable text fallback with posix_spawn by @iliaal in https://github.com/can1357/oh-my-pi/pull/15271
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.9.0...v18.9.1
18.9.0 11 d’octubre del 2026 — 64 bits Portable
18.8.9 10 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Updated state field rendering to trim trailing whitespace and normalize empty collection formatting for consistent prompt output
@oh-my-pi/pi-coding-agent
Added
- Added capability-driven extension terminal launches for tmux, Zellij, Herdr, and CMUX, with consolidated multiplexer detection, provider/shell capability feedback, and required POSIX-shell confirmation for shell-input launches; CMUX shell input preserves non-ASCII arguments and pane working directories (#13620 by @anatoli-tsinovoy).
- Added /fork pane|window|tab to open a fork of the session in a new multiplexer pane or window while this session keeps running; bare /fork still forks in place (#13620 by @anatoli-tsinovoy).
- Added Orca CLI-based launches for /fork pane|window|tab and ctx.ui.openTerminal(), with POSIX-shell confirmation; when Orca can only start a tab as a background terminal, the launch reports a warning instead of a visible tab (#13620 by @anatoli-tsinovoy).
Changed
- CLI --fork now pairs tool calls the source session left unresolved with an unknown-outcome result, and /tan clones report such calls as unknown-outcome instead of aborted; forking a source whose process already exited keeps the process-exit recovery and pending-tool warning (#13620 by @anatoli-tsinovoy).
- Table charts (tui.autoGraph) now cover before/after tables whose rows are different metrics (time, memory, counts): each row's change is drawn as a factor of its "before" value (48× less, +27%), and rows written in prose are skipped and named under the chart. A single column of scores like 12/12 is drawn as bars filling toward 100%
- In terminals that speak the Tern Surface Protocol (Tern), the composer's bottom bar shows the git branch and its status beside the model chip when the status line has the git segment; it outlasts the other facts as the bar narrows, and clicking it opens /git (#15220 by @H4vC)
- Table charts plot scores like 154/160 as the share of their total instead of the first number, and tables comparing two or three columns across four or more metrics now get a chart: a grid of one small panel per metric
Fixed
- Fixed table charts failing on a table cell such as 3 constructor or 2 valueOf calls
@oh-my-pi/pi-tui
Breaking Changes
- Removed the isInsideTmux(), isInsideZellij(), and isInsideHerdr() exports; use hasTerminalMultiplexerSession("tmux" | "zellij" | "herdr", env) instead (#13620 by @anatoli-tsinovoy)
Added
- classifyTerminalMultiplexer() reports "orca" inside Orca terminals while isInsideTerminalMultiplexer() stays false there, so Orca keeps the direct-terminal render path; TERMINAL_MULTIPLEXER_ENV_KEYS lists every environment variable multiplexer detection reads (#13620 by @anatoli-tsinovoy)
- Added terminalMultiplexerSessions(), terminalMultiplexerForTerm(), classifyTerminalMultiplexerModule(), and routeTerminalMultiplexerNotification(), driven by one multiplexer registry that now owns notification routing and per-multiplexer capability overrides (#13620 by @anatoli-tsinovoy)
- Added change and progress table-chart kinds with ChartSpec.baseline and TableColumn.scores: a before/after table whose rows have their own units charts each metric as a factor of the baseline column (rows without two numbers are named in the caption), and analyzeTable reads whole-number scores (12/12, 154/160) as the percent of their total, keeping a / b pairs and lists as written
Changed
- StatusLineComponent.describeComposerFacts() includes the configured git segment among the composer's facts (branch and status, the status.git click action, pinned so it outlasts the other facts) instead of leaving the branch to Tern's pane header (#15220 by @H4vC)
- Small-multiple table charts with more than three panels and at most six categories draw as bands of per-metric panels with a color legend, and worthCharting counts their panels like categories
- Table charts take the look of Apple's charts and Tern: a UI sans, hairline gridlines, rounded bars and cells, smooth 
18.8.8 10 d’octubre del 2026 — 64 bits Portable
18.8.7 9 d’octubre del 2026 — 64 bits Portable
18.8.6 8 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added support for warm-cache-aware conversation pruning, keeping pruned history within the model’s prompt-cache lookback window while preserving Anthropic prompt-cache efficiency.
- Added AgentLoopConfig.hasQueuedAsides (also available on Agent) to allow interruptible waits to detect queued asides without consuming them or signaling other tools.
@oh-my-pi/pi-ai
Fixed
- Fixed Google Gemini and Cloud Code Assist (Antigravity) requests failing when tool schemas contain unsupported JSON Schema keywords or fields that allow multiple types.
- Fixed auth broker account selection and usage-limit enforcement to consistently use the selected account’s quota when multiple accounts are present.
- Fixed Anthropic-family model streaming so encoded marker tokens are decoded correctly in text, tool-call updates, partial messages, and completed tool calls.
@oh-my-pi/pi-catalog
Added
- Added prompt-cache lookback support for Claude models across all hosts, including the public prompt-cache-lookback catalog axis and resolvePromptCacheLookback API.
Fixed
- Fixed model discovery when providers publish models before they are recognized by the catalog; unsupported models are now skipped with a warning so other available models remain discoverable.
@oh-my-pi/pi-coding-agent
Added
- Added per-session Git worktree support with worktree.onStart and worktree.onExit settings to create an isolated worktree for each session and clean it up when the session ends.
- Expanded xAI web search with X post search, including X-only and author-specific queries, author exclusions, date and recency filters, and automatic xAI routing when credentials are available.
- Added xAI-powered reading of X posts, threads and replies, profiles, searches, and hashtags when logged in, replacing the unavailable Nitter mirrors.
Changed
- Web search now prefers an authenticated xai-oauth login over an xai API key when both are available, unless modelProviderOrder specifies a different order.
Fixed
- Fixed judge-gated features continuing to use a stale model chain after switching judge roles.
- Improved Anthropic prompt-cache reuse when pruning tool results from long conversations.
- Fixed resumed Claude sessions losing earlier thinking context and prompt-cache reuse when extensions or MCP tools were registered before the first message.
- Fixed subagent advisors configured with @advisor using the built-in slow model instead of the configured advisor role.
- Fixed the /switch command and alternate model picker crashing when stored model speed statistics contained an unnamed model.
- Fixed lsp and generate_image attempting to read FIFO, terminal, or unbounded device paths, which could hang or exhaust memory.
- Fixed aside messages from extensions being blocked behind a running wait operation.
- Fixed extensions importing @oh-my-pi/pi-tui/native/* failing to load in compiled omp binaries.
- Fixed raw token markers appearing instead of Nerd Font icons in Anthropic idle recaps, /btw and /omfg replies, and streaming previews.
- Fixed sessions moved with /wt disappearing from resume lists; sessions in Git worktrees now remain discoverable and can be resumed or relocated if their worktree was removed.
- Fixed live config reload ignoring edits made during startup or right after a config symlink was retargeted, until the next unrelated edit.
@oh-my-pi/pi-tui
Changed
- Improved rendering performance for streaming long Markdown lists while preserving nested items, numbering, and reference links.
Fixed
- Added package exports for the native Tern/TSP modules, allowing extensions in compiled omp binaries to import them.
- Fixed the Tern /model picker's Roles tab so typing and Backspace consistently edit the search field without triggering role commands.
- Fixed the session picker so sessions retain and display their directory when switching between the current folder and all-project views.
What's Changed
- fix(agent): stop tool-result pruning from re-writing the whole
18.8.5 8 d’octubre del 2026 — 64 bits Portable
18.8.4 8 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Breaking Changes
- AuthBrokerClient.notifyUsageStale and UsageLedgerStore.invalidateUsageCache (including RemoteAuthCredentialStore.invalidateUsageCache) now take an optional leading provider argument: the signature is (provider?: string, signal?: AbortSignal) instead of (signal?: AbortSignal) (#14761 by @will-bogusz)
Added
- Auth gateway route option excludeProviders leaves those providers' accounts out of /v1/usage and /v1/credentials/check (#14755 by @will-bogusz)
Fixed
- Fixed a single-provider usage refresh from an auth broker client (omp usage invalidate --provider, Codex auto-redeem, a saved-reset redeem) wiping the broker's cached usage for every provider, so other accounts showed no usage on every client until their next successful probe (#14761 by @will-bogusz)
- Fixed Anthropic sessions missing the prompt cache a second time after the API dropped thinking blocks from a changed conversation (#14748 by @will-bogusz)
- Fixed Cursor logins never renewing: every refresh was rejected with "Invalid User API Key", so logins lapsed about 60 days after sign-in; a session Cursor has ended is now disabled with a re-login hint (#14753 by @will-bogusz)
- Fixed the auth broker logging caller-supplied X-Forwarded-For / X-Real-IP values and unknown request paths; peers come from the socket unless trustProxyHeaders is set (#14762 by @will-bogusz)
- Fixed false thinking-loop detections on Gemini, DeepSeek, and Grok when the model drafts or answers with repetitive code or markup (VRML, SVG, JSON); such output is no longer discarded and retried.
- Fixed Cursor provider leaking conversation checkpoints, blob stores and abort listeners after requests end (#14669 by @H4vC)
- Fixed session-affinity pins growing without bound in long-lived gateways (now capped at 256 sessions per provider) (#14672 by @H4vC)
- Fixed completed Cursor turns failing with "Cursor stream ended before turnEnded" when the connection closed after the answer had fully arrived (#14851 by @kyle-elliott-asymptote).
- Fixed Cursor provider errors that Cursor marks as not retryable being retried until the retry budget ran out (#14851 by @kyle-elliott-asymptote).
@oh-my-pi/pi-catalog
Fixed
- Fixed Claude Haiku 5.5 opening at its full 1M window on Amazon Bedrock, Google Vertex, and other non-Anthropic hosts: every priced host now carries the 5x pricing band above 100K input tokens, so the window stays at 100K unless extended context is on (#14903 by @H4vC).
@oh-my-pi/pi-coding-agent
Changed
- Agent Hub keeps existing agents in place while open; new agents appear first in the flat roster or within their tree sibling group (#13066 by @kmccleary3301)
Fixed
- Fixed omp auth-broker serve logging every client as unknown (or as whatever a caller put in X-Forwarded-For); it now logs the socket address, with --trust-proxy-headers for brokers behind a reverse proxy (#14762 by @will-bogusz)
- Fixed relative file links in Tern assistant replies opening against the folder omp was started in after /wt or /move; they now open the file in the session's working directory (#14879 by @H4vC)
- Fixed the BTW history, git shortcuts, and autoresearch sheets in Tern having no Close button, and the plan review sheet having no Cancel button (#14894 by @H4vC)
- Fixed the agent transcript viewer and /annotate review in Tern having no clickable way out; both now show a clickable esc at the top right (#14894 by @H4vC)
- Fixed the /move folder picker in Tern having only key hints; Accept (Tab), Cancel (Esc) and Confirm (Enter) are now clickable buttons (#14894 by @H4vC)
- Fixed omp usage listing a second "7 days … not reported" row for a Codex account that reports only a weekly window, and misaligning its rows against accounts that report a 5-hour window too (#14760 by @will-bogusz)
- Fixed omp auth-gateway serve showing the usage and credential health of providers listed in disabledProviders, account emails included, on /v1/usage and /v1/credentials/check (#14755 by @will-bogusz)
- Fixed 
18.8.3 7 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added OAuthRefreshUnavailableError, a retryable error that keys.getWithCredential and oauth.access reject with when every usable OAuth credential failed to refresh transiently; keys.get resolves undefined instead so availability probes move on to their next candidate (#14843 by @H4vC)
Fixed
- Fixed a single transient OAuth token-refresh failure (network blip, timeout, 5xx) ending a running session, including subagents restricted to an account pool, with a non-retryable "No API key for provider" error while the stored credential was still valid; the refresh error now surfaces and the request is retried (#14843 by @H4vC)
@oh-my-pi/pi-catalog
Added
- Added Claude Haiku 5.5 with adaptive thinking (low through max effort), image input, a 1M-token context window, 128K output, and its tiered pricing above 100K input tokens.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed repeat opens of the model picker, model hub, or agents views stacking duplicates; the open one is brought forward instead (#14846 by @H4vC)
What's Changed
- fix(coding-agent): focus the open model picker and hubs instead of stacking duplicates by @H4vC in https://github.com/can1357/oh-my-pi/pull/14846
- fix(ai): retry transient OAuth refresh failures instead of reporting a missing key by @H4vC in https://github.com/can1357/oh-my-pi/pull/14843
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.8.2...v18.8.3
18.8.1 7 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added the public validateAgentToolArguments() API for consistent, lenient-aware validation of agent tool arguments across agent and coding-agent workflows.
@oh-my-pi/pi-ai
Added
- Added session restrictions for OAuth account pools via AuthStorage.sessions.restrict, limiting selection, fallback, rotation, and authentication to specified accounts until the returned lease is released with sessions.unrestrict. API keys and other accounts are not used when a session is restricted.
- Exported resolveCredentialIdentityKey for determining the identity key used to match credentials with broker account pools and session restrictions.
Fixed
- Fixed Codex Fast (priority) and Ultrafast usage being recorded, billed, and reported as Standard when the backend echoed a default service tier; the requested tier is now preserved in usage and performance records.
@oh-my-pi/pi-catalog
Fixed
- Fixed Codex Fast (priority) pricing to use OpenAI’s 2.5× included-usage rate for supported models, excluding GPT-5.5 and GPT-6 Astra.
- Fixed GitHub Copilot models with tier-specific prompt limits incorrectly defaulting to the long-context window.
@oh-my-pi/pi-coding-agent
Added
- Added account pools for task agents, allowing an agent and its related work—including advisors, title generation, subagents, and resumed sessions—to use only specified OAuth accounts and fail rather than fall back to another account or an API key.
- Added omp usage accounts to list OAuth account provider and identity keys without exposing tokens, making account-pool configuration easier.
Changed
- Generated session titles now include their card icon and short code, so the /resume picker and session listings show this context; title.icons applies to newly generated titles.
- Symlinked routing configurations now reload when an intermediate file or profile-directory link is replaced, without requiring a session restart.
Fixed
- Improved Tern ask questions so the question no longer obscures the transcript and can be answered while reviewing and scrolling through the relevant context.
- Added warnings for unrecognized compat keys in models.yml while continuing to load the configuration; warnings identify the file and key path.
- Fixed /new so new sessions reset plan and goal modes, todo state, and related planning configuration instead of carrying stale state forward; fresh prewalk behavior is also restored after handoffs, cancellations, or automatic recovery.
- Fixed OAuth alias login and logout flows so they refresh the underlying provider's models, report authentication state correctly, and list and remove the accounts associated with that provider.
- Fixed model speed statistics so throughput from non-default OpenAI or Codex service tiers is kept separate and displayed for the tier used by the session.
- Fixed stale model catalogs remaining in memory after provider or catalog refreshes.
- Fixed write and edit from hanging or exhausting memory when given FIFOs, terminals, device files, or other non-regular targets; unsupported targets are now rejected safely.
- Fixed long conversations from losing user and tool-result images when the conversation also contained assistant-generated images.
- Fixed local:// session paths from escaping their intended storage directory when given . or .. session identifiers.
- Fixed omp gc --archive --apply from leaving orphaned session-title records behind.
- Fixed /retry after an interrupted reopened ask prompt when an extension had registered a context handler.
- Fixed browser relay sessions from missing cross-origin iframe content that loaded before the tab was opened.
- Fixed browser clicks on invisible native radio and checkbox inputs, including controls commonly used by GOV.UK forms.
- Fixed /model so selecting the model already assigned to a project's default role still switches to that model correctly.
- Fixed wait and background task state when a completion has already been consumed elsewhere, preventing premature 
18.8.0 7 d’octubre del 2026 — 64 bits Portable
18.7.0 6 d’octubre del 2026 — 64 bits Portable
18.6.3 6 d’octubre del 2026 — 64 bits Portable
18.6.1 4 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed native OpenAI context compaction for sessions containing many screenshots, preventing image-size estimates from incorrectly causing compaction requests to be rejected.
@oh-my-pi/pi-ai
Fixed
- Fixed compatibility with Command Code DeepSeek and other DeepSeek-family models by preserving the reasoning context required for warm OpenAI Responses sessions and correctly handling incomplete DSML tool-call wrappers in visible output.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed waiting for subagent follow-up messages: responses now appear as background jobs that can be waited on or canceled, and are delivered only once.
- Improved /switch autocomplete so model and role suggestions use the same relevance ordering as the model picker, including support for @role aliases.
- Fixed concurrent skill:// searches blocking other filesystem operations and delaying subagent artifact publication.
- Fixed compatibility checks for browser relays from other OMP versions and added guidance for resolving stale connection-refused errors.
- Fixed follow-up hashline edits being incorrectly rejected after earlier edits shifted anchored lines, while continuing to reject genuinely stale line references.
- Fixed pi.exec() reporting exit code 0 when a process was terminated by a timeout or signal; terminated processes now report code -1.
- Fixed /collab guests being unable to respond to setting-change approval and tool-issue report consent prompts.
- Fixed follow-up hashline edits being rejected as anchored on lines "never displayed" after an earlier edit shifted those lines, when the replacement carries the anchored content; genuinely stale line numbers are still rejected (#14254 by @abilliontokens).
- Fixed long /btw answers in Tern being clipped with no way to scroll: /btw now answers in the scrollable BTW history sheet (#14331 by @H4vC)
- Fixed /btw answers longer than 4 KiB being cut off with […truncated] once they finished (#14331 by @H4vC)
- In Tern, Esc puts the BTW history sheet away while an answer keeps streaming (/btw reopens it); x cancels the answer (#14331 by @H4vC)
@oh-my-pi/pi-natives
Fixed
- Fixed concurrent searches through host-provided filesystem callbacks so they no longer starve other asynchronous filesystem operations, and ensured canceled searches release promptly.
@oh-my-pi/pi-tui
Fixed
- Fixed the Space key in the Git diff pane so it stages or unstages the focused hunk instead of scrolling or acting on the wrong change.
- Fixed the BTW history sheet in Tern: the history list no longer collapses beside a long answer, the panes lose their foldable ## headings, and the arrow/page keys scroll the answer (#14331 by @H4vC)
- Fixed a single saved side question in the BTW history panel opening on its one-row list: its answer now has focus, and Enter or f to follow up jumps to the bottom of the conversation (#14331 by @H4vC)
What's Changed
- fix(pi-edit): accept shift-consistent hashline anchors carrying current content by @abilliontokens in https://github.com/can1357/oh-my-pi/pull/14254
- fix(agent): estimate image tokens from dimensions for native compaction fit by @H4vC in https://github.com/can1357/oh-my-pi/pull/14260
- btw ux uplifts, tern specific. by @H4vC in https://github.com/can1357/oh-my-pi/pull/14331
- fix(collab): mirrored cfg:// approval prompts to collab guests by @roboomp in https://github.com/can1357/oh-my-pi/pull/14282
- fix(ai): filled reasoning_text on warm summary-only reasoning replay by @roboomp in https://github.com/can1357/oh-my-pi/pull/14294
- fix(natives): moved provider searches off libuv workers by @roboomp in https://github.com/can1357/oh-my-pi/pull/14298
- fix(browser): diagnosed stale relay after omp upgrades by @roboomp in https://github.com/can1357/oh-my-pi/pull/14299
- fix(tui): restored Space row staging in git diff by @roboomp in https://github.com/can1357/oh-my-pi/pull/14329
- fix(ai): preserved prose after incomplete dsml wrappers by @roboomp in https://github.com/can1357
18.6.0 3 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- When a DeepSeek turn ends with raw <|DSML|…> tool-call text that could not be parsed into a real call, and no tool call was made, the broken markup is now removed from the message before it is saved to history. The agent then tells the model the tool call failed, shows it the correct format, and asks it again, at most twice in a row. Previously the markup stayed in history, and the agent stopped as if the model had finished (#14202 by @H4vC).
@oh-my-pi/pi-ai
Fixed
- Fixed Antigravity chat and image requests sending an outdated client version when the model list came from cache, which could make newer models such as Claude Opus 5.5 unavailable.
- When a DeepSeek model writes a broken DSML tool call (for example with the opening <|DSML|tool_calls> and <|DSML|invoke> tags missing), its closing tags are now kept in the streamed text instead of being dropped. This lets the agent remove exactly the broken call while keeping any text after it (#14202 by @H4vC).
@oh-my-pi/pi-catalog
Fixed
- Fixed DeepSeek V4 model IDs and the V4.1 Flash alias lacking version information in model identity and dashboards (#14194).
- Fixed Antigravity models such as Claude Opus 5.5 and Sonnet 5.5 disappearing after omp models refresh. When the update check failed, omp reported an outdated Antigravity client version (2.8.0), so the server left the newer models out of the list. The fallback version is now 2.19.1.
- DSML tool calls from DeepSeek models are now parsed on every host. This includes local servers (llama.cpp, LM Studio, vLLM), custom providers from models.yml, and gateways that weren't on the old list of supported hosts. Before, a complete <|DSML|tool_calls> envelope from these hosts showed up as plain text and the tool never ran (#14202 by @H4vC).
@oh-my-pi/pi-coding-agent
Added
- The /models Roles view shows which saved model preset is in effect, and Ctrl+←/→ (or p/P on the role rows, for macOS where Ctrl+←/→ switches Spaces) switches to the next or previous one, in Tern and text mode (#14210 by @H4vC)
Changed
- Picking a model for a role in /models now puts the cursor on the model list, so ↑/↓ choose a model and Enter assigns it right away instead of moving through the sidebar and dropping the role selection; ← still reaches the providers (#14210 by @H4vC)
- Sped up secret redaction on long conversations: it no longer slows down as history grows (#14213 by @H4vC)
- Startup is faster with plugins that bundle large dependency trees: the extension loader no longer re-reads and re-checks the same files while loading them (e.g. ~280 ms → ~185 ms with the IDA MCP plugin) (#14219 by @H4vC)
Fixed
- Fixed an EPIPE: broken pipe unhandled rejection crashing the session when a debug adapter, eval kernel, IDA worker, or RPC server exits mid-write (seen on Windows) (#14196 by @andrebrait)
@oh-my-pi/omp-stats
Fixed
- Fixed the Frustration page splitting DeepSeek V4 provider variants and the V4.1 Flash alias into separate model-version rows (#14194).
@oh-my-pi/pi-tui
Fixed
- Fixed Alt+Up (restore queued steering), arrows and other escape-sequence keys on Windows acting as Esc (interrupting the turn) and typing [1;3A into the editor when the console host relays them one byte at a time (#14216 by @H4vC).
What's Changed
- fix(catalog): agy anthropic models display by @H4vC in https://github.com/can1357/oh-my-pi/pull/14197
- fix(stats): merged DeepSeek V4 frustration rows by revision by @roboomp in https://github.com/can1357/oh-my-pi/pull/14199
- fix(dsml): parse DeepSeek DSML on every host and keep unparsed markup out of history by @H4vC in https://github.com/can1357/oh-my-pi/pull/14202
- feat(tui): switch model presets with ctrl+arrows in the /models roles view by @H4vC in https://github.com/can1357/oh-my-pi/pull/14210
- perf(secrets): speed up secret redaction by @H4vC in https://github.com/can1357/oh-my-pi/pull/14213
- fix(tui): reassemble escape keys relayed as win32-input-mode text records by @H4vC in https://github.com/can1
18.5.1 3 d’octubre del 2026 — 64 bits Portable
18.5.0 3 d’octubre del 2026 — 64 bits Portable
18.4.12 2 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added createAuthGatewayRouter, the auth-gateway's routes without the HTTP listener, and serveAuthGatewayStdio, which serves them as JSON lines ({"id", "path", "body"} in, {"id", "status", "body"} out) for a parent process.
@oh-my-pi/pi-coding-agent
Added
- Added omp auth-gateway stdio: a long-lived inference server for other programs speaking JSON lines on stdin/stdout ({"id", "path": "/v1/chat/completions", "body"} in, {"id", "status", "body"} out) with your own sign-ins; a request's model takes any --model selector (@smol, sonnet, @commit,@smol) and falls back along retry.fallbackChains when an attempt fails.
Changed
- Reduced CPU use while streaming with several agents active: extension message_update handlers are delivered through a lighter queue, and RPC no longer processes subagent events unless a client subscribed to them (#13244 by @iliaal).
Fixed
- Fixed web search stopping at Perplexity's anonymous signup wall instead of falling back to the next configured provider (#12756).
- Fixed imported Claude Code sessions on Windows reporting the encoded C--… directory name instead of the registered project path when the transcript records no cwd (#13363 by @jchanghong023).
- Fixed JavaScript eval wait()/handle.wait() ignoring a positional timeout; h.wait(30) now waits up to 30 seconds like { timeout: 30 }, and mixing an options object with positional arguments throws a TypeError (#12720 by @F0Rextasy).
- Fixed Herdr and other lifecycle-tracking extensions showing a collab guest (omp join) as idle while the host is working; extension-initiated turns (pi.sendMessage with triggerTurn, pi.sendUserMessage) are refused as host-only while joined instead of running on the guest's local model (#13156 by @Fruitseller).
- Fixed omp update failing with "Malformed npm registry response … missing version" on registries such as Sonatype Nexus that answer the /<pkg>/latest lookup with the full package document or a non-manifest body; the updater now reads the version from the package's dist-tags (#14115).
- Fixed wait and proc:// failing with "Daemon list request timed out" when the project daemon broker hangs; background jobs and agents are still reported, and proc:// shows that services are unavailable (#14120 by @H4vC).
- Fixed agent://<id> showing an agent's previous published output as current while that agent runs a follow-up or wake turn; the read now says the output is from the previous run (#14120 by @H4vC).
- Fixed proc:// listing agents as running when they have no turn in flight; their rows now say the run is stale or already finished, as the jobs tool does (#14120 by @H4vC).
- Changed subagents to skip their own builds, tests, and smoke runs and leave verification to the main agent, avoiding CPU spikes from many subagents verifying at once.
@oh-my-pi/pi-tui
Changed
- Changed Ctrl+Z in Tern's native composer to undo the last edit (TSP undo event) instead of suspending
@oh-my-pi/pi-utils
Fixed
- Fixed multi-second temp directory removal stalls on Windows by forcing a major GC before the first deletion retry (#13044 by @jchanghong023).
What's Changed
- fix(eval): accept positional timeout in JS wait() via optionsArg by @F0Rextasy in https://github.com/can1357/oh-my-pi/pull/12720
- fix(web-search): classify Perplexity anonymous signup wall as failure by @roboomp in https://github.com/can1357/oh-my-pi/pull/12838
- fix(omp-rpc): verify teardown and retain a survivor handle for reaping by @roboomp in https://github.com/can1357/oh-my-pi/pull/12843
- fix(utils): force a GC before retrying temp dir removal on Windows by @jchanghong023 in https://github.com/can1357/oh-my-pi/pull/13044
- fix(collab): mirror guest lifecycle events to the extension runner by @Fruitseller in https://github.com/can1357/oh-my-pi/pull/13156
- perf(coding-agent): reduce streaming event queue overhead by @iliaal in https://github.com/can1357/oh-my-pi/pull/13244
- fix(coding-agent): encode Windows drive colons in Claude project dirs by @jchan
18.4.10 2 d’octubre del 2026 — 64 bits Portable
18.4.9 1 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed rejected HTTP 400 requests from consuming unbounded disk space by automatically cleaning up old request logs and enforcing a size limit.
- Fixed unnecessary credential and session updates that could trigger needless authentication reloads in other running processes.
- Fixed context-overflow recovery for Strata requests that exceed the model context limit but return no usage information.
@oh-my-pi/pi-catalog
Fixed
- Fixed model catalog caching so unchanged catalogs refresh without unnecessary rewrites, and offline snapshots for endpoint-less models (such as Azure models) are now handled correctly across startups.
@oh-my-pi/pi-coding-agent
Added
- Added opt-in stale-session garbage collection with omp gc --stale or gc.stale, removing orphaned session markers and terminal breadcrumbs and expiring old debug reports and collaboration replicas according to configurable retention limits.
- Added RPC controls to cancel or steer individual foreground and background subagents without aborting the session.
- Added RPC word-completion commands so web and IDE hosts can provide the same ghost-text completion available in the terminal editor.
- Added an opt-in RPC ask-dialog mode that lets hosts render all questions together with checkbox or radio-button controls and submit their answers in one response.
- Added SDK notifications when a session moves to a new persistence file, including the previous and new paths.
- Added SDK APIs to inspect and cancel background jobs, including their command, working directory, process IDs, exit status, and captured output.
Changed
- Improved the omp predict comparison view with an interactive native interface, table rows, action controls, and clearer status information.
- Improved MCP authorization prompts with clickable links, native copy/open context menus, and a clearer URL layout.
- RPC hosts are now notified when omp cancels an expired select, confirm, input, or ask dialog, allowing stale UI prompts to be closed.
- Limited saved bash, Python, and JavaScript evaluation output artifacts to 16 MB by default while preserving both the beginning and latest output; configure the limit with tools.artifactMaxBytes, or set it to 0 for unlimited output.
Fixed
- Fixed /wt on filesystems without copy-on-write cloning, including NTFS, so unchanged files are not incorrectly marked modified and staged edits, additions, and deletions retain the correct contents.
- Fixed Windows Ctrl+V taking about a second to paste by avoiding unnecessary PowerShell clipboard checks.
- Fixed local:// paths being misinterpreted as local filesystem paths by the read, write, and search tools.
- Fixed read handling of semicolon-separated URLs, local paths, and line selectors so each entry is processed independently.
- Fixed session persistence conflicts between multiple omp processes, preventing lost or interleaved turns and continuing in a new session file when necessary.
- Fixed session image handling to avoid unnecessary rewrites, preserve images after interrupted writes, and prevent garbage collection from removing images that are referenced again.
- Reduced unnecessary disk writes and improved persistence efficiency across sessions, model data, configuration, and background jobs.
- Fixed the native composer showing the main session's effort level instead of the selected subagent's level.
- Fixed the omp predict comparison view and MCP authorization prompt rendering with their full native interfaces, including clickable link actions.
@oh-my-pi/pi-natives
Added
- Added readTextFromClipboard() for reading plain text from the system clipboard without starting a subprocess.
- Added Shell.pids() to retrieve the IDs of still-running processes spawned by an in-flight shell command.
@oh-my-pi/pi-tui
Added
- Exported wordCompletionQuery() so hosts outside the editor can apply the same prose eligibility rules used by ghost-text word completion.
- Added a full-featured Background jobs view with selectable jobs, live
18.4.8 1 d’octubre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-natives
Fixed
- Fixed omp 18.4.3 and later crashing with a segmentation fault at startup on Apple silicon Macs running macOS older than 27; Apple Foundation Models support now loads only on macOS 27 and later
@oh-my-pi/pi-tui
Added
- Added an optional terminal section to theme JSON (background, foreground, chrome, widget, 16 ansi colors) naming the terminal a theme was made for, for hosts that paint the terminal themselves; the built-in themes ported from known schemes (GitHub, Nord, Dracula, Catppuccin, Solarized, Gruvbox, Tokyo Night, One, Monokai, Rosé Pine, Poimandres, Celestial) carry their scheme's.
Fixed
- Fixed native-terminal (TSP) frames held back by unacknowledged credits waiting for an unrelated render after the 5-second stall fallback expired; a credit-blocked change now renders as soon as the oldest frame counts as stalled.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.4.7...v18.4.8
18.4.5 30 de setembre del 2026 — 64 bits Portable
18.4.4 29 de setembre del 2026 — 64 bits Portable
18.4.3 29 de setembre del 2026 — 64 bits Portable
18.4.2 28 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added tool_execution_end events that fire as each tool call settles for live UI updates
Changed
- Emitted tool result messages in the order of tool calls, preserving call order regardless of completion order
- Reduced repeated token-counting work with a bounded, model-scoped cache of exact text and short-message fragment counts.
Fixed
- Fixed an issue where streaming tool call arguments could be incorrectly modified in-place
@oh-my-pi/pi-ai
Fixed
- Fixed successful Cursor agent turns being treated as context overflows, which ran overflow compaction and showed "Compaction freed too little context to make progress" while /context read well under the window; overflow detection now uses the reported context size instead of input totals summed across a turn's model calls (#13608 by @H4vC)
- Fixed Anthropic requests with thinking enabled failing on models whose output ceiling cannot fit the minimum thinking budget; thinking is now disabled for those requests instead (#13359 by @jchanghong023)
- Fixed Cursor native Grep/Glob results showing no matches or raw output, Write failing to create files, StrReplace missing edits beyond the read limit, and Read/Shell/Delete results misreporting content or metadata (#13600).
@oh-my-pi/pi-catalog
Changed
- Improved model cache invalidation efficiency by implementing deep equality checks on cached rows when database version signals change
- Optimized read-row cache by enabling granular cache-hit logic across concurrent connection handles
@oh-my-pi/pi-coding-agent
Added
- The shell's cp builtin accepts macOS's -c (clone where possible, else copy; same as --reflink=auto).
Changed
- Removed unused timestamp metadata from model usage tracking queries to optimize database overhead
- Improved session storage reliability by using atomic inode identity verification for lock management
- Optimized internal role chain resolution and caching to reduce event-loop contention during judge initialization
- On macOS, the shell's cp builtin now also clones (copy-on-write) when overwriting an existing file, instead of rewriting its data; the destination keeps its permissions, and hard-linked destinations are still written in place.
- The find tool now fails with a timeout error after 20 seconds instead of blocking the turn when the judge model stalls.
- Reduced main-thread stalls with many agents running: read code summaries now parse off the main thread, streamed tool-call snapshots no longer deep-copy growing argument text on every delta, and model-cache reads skip re-parsing unchanged rows.
- Reduced CPU while an agent streams: the working-row token rate and status line no longer re-tokenize or re-resolve settings every frame, and ^ model mentions no longer rebuild the model scope per keystroke.
- task and /vibe subagents now get their own Python kernel and JS/Ruby/Julia eval state instead of sharing the parent's, so agents can no longer overwrite each other's variables or reset each other's kernels.
- On Windows, the shell's cp builtin now clones files (copy-on-write) on ReFS and Dev Drive volumes by default, falling back to a regular copy elsewhere; --reflink and -c no longer fail there.
Fixed
- Fixed mouse handling ignoring per-mode TUI mouse setting
- Fixed the shell's cp --reflink=always emptying an existing destination when the filesystem cannot clone; the destination is now left untouched.
- Fixed block-clone task isolation on Windows ReFS and Dev Drive volumes failing on files whose size is not a whole number of clusters, larger than 4 GiB, or sparse.
- Fixed Google Antigravity requests failing with 429 RESOURCE_EXHAUSTED on every turn: the system prompts' RFC 2119 conventions line is reworded so the endpoint no longer rejects it (#13379 by @heshuoshuo0512)
- Fixed CRLF SKILL.md files injecting raw YAML frontmatter into user-invoked and autoload skill messages (#13590).
- Fixed Cursor native Grep ignoring requested context, Read negative offsets starting at the top, and 
18.4.0 28 de setembre del 2026 — 64 bits Portable
18.3.5 27 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Breaking Changes
- Removed the stream-level Anthropic prompt-cache keep-alive: StreamOptions.anthropicCacheRefresh, StreamOptions.anthropicCacheRefreshRequest, and the zero-output refresh request path. Prompt-cache warming now lives in the coding agent's session-level cache warmer (#12699 by @KamijoToma).
@oh-my-pi/pi-catalog
Added
- Added the openai web-search grounding for OpenAI API models that support Responses web search (gpt-5.5, gpt-5.6-luna, gpt-6-astra, gpt-6-luna) (#13467 by @anatoli-tsinovoy).
- Added Model.promptCache, per-retention-tier prompt-cache entry lifetimes in seconds (short / long), declared per provider through the prompt-cache KDL rule (bundled: direct Anthropic, 5 min / 1 h). Custom models and modelOverrides opt in with the models.yml promptCache key (#12699 by @KamijoToma).
@oh-my-pi/pi-coding-agent
Added
- Added API-key-billed OpenAI Responses web search (openai/gpt-6-luna, then openai/gpt-5.6-luna), tried after every Codex entry in the default search fallback chain so ChatGPT-subscription search is exhausted before any API usage is billed (#13467 by @anatoli-tsinovoy).
- Added prompt-cache warming, ported from earendil-works/pi: shortly before a prompt-cache entry expires, the main agent loop replays its last request and cuts the replay off at the first generated token, so idle gaps no longer force a full-prefix cache re-write. A refresh fires only when the expected avoided-miss cost clears its cost by $0.05, and warming stops as soon as a refresh misses the cache. Controlled by providers.cacheWarming (off / streaming / idle, default idle); idle warming covers 5-minute entries only, and models without a declared promptCache lifetime are never warmed. Extensions can override each decision through the cache_warming_decision event (#12699 by @KamijoToma).
@oh-my-pi/pi-natives
Changed
- Improved syntax highlighting to use about 5x less memory and run 3-5x faster by compiling grammars with Oniguruma instead of fancy-regex; highlighted output is unchanged.
Fixed
- Fixed SmolLM word completion (spelling.autocomplete: smollm) being about 25x slower on Windows x64 and Intel Macs; suggestions are unchanged (#13488 by @H4vC)
@oh-my-pi/pi-tui
Added
- Added an "OpenAI API" option to the setup wizard's web-search step; the existing ChatGPT-OAuth option is now labeled "OpenAI Codex" (#13467 by @anatoli-tsinovoy).
What's Changed
- feat(coding-agent): port upstream pi's cache-warming strategy by @KamijoToma in https://github.com/can1357/oh-my-pi/pull/12699
- perf(pi-predict): stopped SmolLM CPU kernels lowering mul_add to libm fmaf on x86-64-v2 by @H4vC in https://github.com/can1357/oh-my-pi/pull/13488
- feat(web-search): add OpenAI API-billed search by @anatoli-tsinovoy in https://github.com/can1357/oh-my-pi/pull/13467
- chore(pi-builtins): removed unused same-file dependency by @H4vC in https://github.com/can1357/oh-my-pi/pull/13491
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.3.4...v18.3.5
18.3.4 27 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Anthropic OAuth requests capping output at 64k tokens; they now request the model's full ceiling (128k on Opus 5.5), matching Claude Code and API-key requests
@oh-my-pi/pi-coding-agent
Breaking Changes
- Replaced the task tool's complexity field with solutionSpace, a description of how open-ended the subtask is; auto thinking for spawned subagents now picks effort from it alone
Changed
- auto thinking now picks effort for every turn by how open-ended the problem is, so large volumes of mechanical work no longer raise it
Fixed
- Fixed agents looping for hours when every turn spends the whole output limit on reasoning: length-stop retries now tell the model its reasoning was discarded and to act in smaller steps, and a subagent whose length-stop recovery gives up now fails with that error instead of being re-prompted into the same loop
- Fixed tagging a model with ^ mid-session dropping the provider prompt cache for every following turn: new m<N> pseudonyms now arrive as a hidden session notice instead of rewriting the task description, which only absorbs them at a base-prompt rebuild
@oh-my-pi/pi-natives
Fixed
- Fixed the native addon keeping every bun test --isolate/--parallel test file's global object and module graph alive, which grew each test worker by ~15 MB per file until the run was OOM-killed.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.3.3...v18.3.4
18.3.3 27 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added live steering support, allowing models to receive and act on user steering messages during an active stream.
@oh-my-pi/pi-coding-agent
Added
- Added a unified predictive text engine with N-gram, SmolLM2, and macOS native providers, including cross-engine blending, background model downloads, and a cross-process prediction daemon.
- Added the omp predict command for evaluating completion performance and support for ingesting existing Claude Code and Codex prompt histories to bootstrap predictions on new installations.
- Added omp skill list [dir] [--json] to report skills resolved for a session directory, including discovery warnings in JSON output.
- Added a centralized progress display for background tool and model downloads, including support for downloading the SmolLM2-135M word-completion model.
- Added dynamic evaluation guidance through hidden session notices.
- Added a required complexity rationale to the task tool to improve automatic thinking-depth selection.
- Agents using task or bash now receive the wait tool for background-process coordination, and subagents can receive it when explicitly requested.
- Added context-aware suggestions to empty composers based on agent activity and effort.
- Added optional global or per-project memory scopes to the retain and learn tools when Mnemopi scoping is enabled.
- Added /btw to focused subagent views for asking questions about that agent's transcript with separate side-conversation history.
Changed
- Completion behavior now uses the N-gram engine for standard auto completion across platforms, with blended N-gram and SmolLM confidence scoring where applicable; the SmolLM2 model uses a 145 MB GGUF (Q8_0) download and is prefetched only when explicitly activated.
- Updated spelling.autocomplete to use an enum-based engine configuration.
- Completion ghost text is now preserved through manual keystrokes.
- Window input actions now default to background execution; set takeover: true to opt into foreground activation, with clarified cross-platform coordinate and activation behavior.
- omp tiny-models download can now download the word-completion model.
- Updated /play help, read-tool summaries, platform-aware shortcut labels, and other UI hints for clearer interaction guidance.
- Updated the empty-submit behavior to account for live-steered messages and surface pending live-steering status in the UI.
- ps --all now includes exited global services, while the default view shows live global services.
- Orchestrator task documentation now follows a Target/Change/Acceptance format.
- Slash-command and hint usage tracking is now persistent and namespaced.
Fixed
- Preserved MCP structuredContent in live tool-result details so evaluation callers can consume server data without reparsing model-facing JSON; spilled results continue to retain an artifact reference without duplicating the payload in session history.
- Fixed Collab hosts becoming unable to reclaim a room after a brief network interruption; hosts now retry room recovery without losing guests or queued updates.
- Fixed one-shot commands that stopped before completing, such as omp config set on a fresh Windows profile, incorrectly exiting successfully without output; they now report failure with diagnostic guidance.
@oh-my-pi/pi-mnemopi
Fixed
- Fixed recall search matching query terms inside unrelated words, improving result relevance and ensuring exact matches are prioritized.
- Fixed sleep summaries replacing shortened terms within larger words, preventing unintended word corruption.
@oh-my-pi/pi-natives
Added
- Added a TextPredictor N-API binding for managing the high-performance ghost-text completion engine.
Changed
- Updated desktop input-control capabilities to use a unified takeover setting, including takeover: true for forced foreground pointer interaction.
@oh-my-pi/pi-tui
Added
- Added responsive ghost-text word completion with pluggable backend providers, context-aware prose filteri
18.3.2 26 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed tool calls that put their payload in the intent field i (for example a file body in write) silently running with the leftover arguments; they now fail with an error telling the model to retry (#13140, #13141 by @radkawar)
- Fixed the Anthropic compaction failure log omitting why no compaction block came back; it now names the stop reason (#13300 by @alphastorm)
@oh-my-pi/pi-ai
Fixed
- Fixed capped Anthropic and Bedrock Claude requests with thinking enabled, including on-demand compaction, ending at max_tokens with no answer; every capped request now gets its effort's thinking budget on top of the requested output (#13300 by @alphastorm)
@oh-my-pi/pi-coding-agent
Added
- Added ctx.agent to the extension context, reporting whether the session is the top-level agent or a subagent, plus its registry id, agent definition name, task depth and parent id, so handlers rebound to subagent sessions can tell which agent they serve (#13314 by @andrebrait)
- Added tracking of Anthropic's usage-limit wrap-up allowance for Claude subscription accounts: after the 5-hour or weekly limit is reached, the status line and /slow status show limit reached · wrapping up · resets HH:MM, and the agent is told to wrap up when neither low priority nor extra usage will continue the work (#13340 by @H4vC)
Changed
- providers.anthropic.slowMode now controls only the low-priority lane; the usage-limit wrap-up allowance is tracked for every first-party Claude subscription account (#13340 by @H4vC)
- Enter on the /model hub sidebar now moves focus to the model list (like →) instead of acting on the highlighted row (#13347 by @H4vC)
Fixed
- Fixed the Windows bash tool exporting TEMP, TMP, and TMPDIR with 8.3 short names such as ADMINI~1, so they now match the long-form pwd/$PWD after cd "$TEMP" (#13265 by @CoderTCY)
- edit and write no longer refuse handwritten files named generated.go, generated.ts, generated.js, or generated.py; these are treated as auto-generated only when their header carries a generated-code marker (#13138, #13139 by @radkawar)
- Fixed the edit tool warning that valid Go 1.26 new(expr) calls (e.g. new(f(x))) introduced a syntax error, and ast_grep/ast_edit reporting parse errors on them (#13148, #13149 by @radkawar)
- Fixed hashline PUT N* / CUT N* on the first statement of a block (for example a Go or Python function that opens with an if) also replacing or deleting every statement after it (#13153, #13154 by @radkawar)
- Fixed the Full output: artifact:// link on large background bash and eval results pointing at a truncated copy with […elided…] gaps instead of the complete output (#13142, #13143 by @radkawar)
- Fixed grep paths like dir/*.go also matching files in subdirectories of dir (#13146, #13150 by @radkawar)
- Fixed auto-compaction re-sending a failed native (server-side) compaction on every turn, re-reading the full context each time; after a failure a retry would repeat, the next configured method runs instead until a compaction succeeds (#13310 by @alphastorm)
- Fixed a /slow off session resending requests indefinitely when another session had activated the shared Anthropic low-priority lane (#13340 by @H4vC)
What's Changed
- fix(pi-edit): allow editing handwritten generated.go files by @radkawar in https://github.com/can1357/oh-my-pi/pull/13139
- fix(agent): reject tool calls whose intent field carries the payload by @radkawar in https://github.com/can1357/oh-my-pi/pull/13141
- fix(ast): parse Go 1.26 new(expr) without syntax errors by @radkawar in https://github.com/can1357/oh-my-pi/pull/13149
- fix(natives): fail the build when an installed host addon cannot load by @alphastorm in https://github.com/can1357/oh-my-pi/pull/12682
- fix(coding-agent): link background job Full output to the raw capture by @radkawar in https://github.com/can1357/oh-my-pi/pull/13143
- fix(pi-ast): PUT/CUT N* on a block's first statement stops at that statement by @radkawar in https://github.com/can1357/oh-my-pi
18.3.1 25 de setembre del 2026 — 64 bits Portable
18.3.0 24 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added support for documenting agent tools on demand through the new AgentTool.docTopics method.
- Added TOOL_INTERRUPT_ABORT_REASON so interruptible tools can distinguish queued steering, peer messages, or background completions from a full run abort.
Changed
- Improved interrupt handling so tools respect wait mode and can be interrupted when appropriate.
- Updated Anthropic compaction compatibility with signature verification.
@oh-my-pi/pi-ai
Added
- Added support for Anthropic User Profiles, including schema-validated API responses.
- Added support for Apple Foundation Models running on-device, including tool calling and vision capabilities.
- Added multi-account authentication and authorization for Codex cyber access programs, including automatic request replay after access-program rejections.
- Added credential-aware authentication routing with per-account OAuth policies, deterministic account selection, protected quota reserves, persistent rate-limit tracking, automatic recovery, and sticky session-to-credential affinity.
- Added deprecated getApiKey and reload methods for backward compatibility.
@oh-my-pi/pi-catalog
Added
- Added compaction support for the claude-mythos-preview model.
- Added provider and authentication rule definitions for on-device Apple Foundation Models.
- Added per-account discovery of Codex cyber access entitlements.
- Added org-scoped-identity and oauth-token-env authentication rules for providers.
@oh-my-pi/pi-coding-agent
Breaking Changes
- The hub tool is deprecated; use wait, write, and the proc:// protocols instead.
- The irc.timeoutMs configuration setting has been removed.
- The edit mode syntax now uses *** Edit File:, *** Find, and *** Replace headers instead of SM: headers.
- Cancelling a process through write now requires an explicit proc://<id>/kill target; other write targets validate content normally.
Added
- Added omp:// documentation scopes for find and omp find. Search all embedded harness documentation with omp:// or a specific document with omp://<file>.md; results are returned as canonical URLs that read can open, including range selectors.
- Added extension support for ephemeral, /btw-style side turns through ctx.runEphemeralTurn(), with optional tool suppression and output/context limits without adding the turn to session history.
- Added background job and service management through the wait tool and proc:// URLs, including supervised services in bash and direct agent messaging through agent:// write targets.
- Added *** Insert Before and *** Insert After edit operations for adding lines without replacing existing code.
- Added the toks command for offline token counting, including support for Jev (TypeSafe Jev 1.13) encodings.
- Added automatic discovery of Apple Foundation Models on supported Apple silicon devices.
- Added /changelog last [N] for viewing the latest release or a selected number of recent releases.
- Added terminal-based OAuth authentication with omp login, including browser-assisted login, account and organization details, and automatic model discovery refresh. Added provider support for org-scoped-identity, oauth-token-env, and per-account OAuth priority/reserve policies through auth.accountPolicies, with policy state shown by omp usage.
- Added the daybreak badge to omp usage for enabled accounts.
- Added /export and /usage to focused subagent views for exporting a focused transcript and viewing account usage without returning to the main session.
- Pasted clipboard images are now saved in the session artifact directory, allowing agents to read, copy, or upload them by file path.
- Added /annotate for attaching notes to diffs, replies, session messages, files, or quoted text and inserting or sending those notes in prompts and reviews.
- Added configurable MCP startup behavior through MCP_STARTUP_TIMEOUT_MS/mcp.startupTimeoutMs and OMP_MCP_REQUIRE_READY=1, allowing headless runs to require MCP servers to become ready before 
18.2.11 23 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed background job completions interrupting foreground Bash and eval calls, which could cause those calls to be repeatedly moved into the background.
@oh-my-pi/pi-ai
Fixed
- Fixed Claude Opus 5.5 not applying a mid-session switch to high-effort reasoning when the session started without an explicit effort setting.
- Fixed Alibaba Token Plan monthly quotas not appearing in usage reports or the status line.
@oh-my-pi/pi-catalog
Added
- Added built-in StepFun provider support for the step-5-preview model, including live model discovery.
Fixed
- Fixed cost estimation for OpenAI Codex GPT-6 Sol and Luna models across standard and worker routes.
- Fixed GitHub Copilot Grok 4.x models failing to make requests.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed nested eval Todo updates not being reflected by the Todo tracker, including cases where a cell fails after committing an update.
- Fixed strict-mode structured-output validation for JSON Schemas without a root type, preserving their items and required keywords.
- Improved streamed TTSR whole-buffer matching to avoid repeated scans from the beginning of the buffer.
- Fixed plural browser queries when compiled binaries provide shallow stack traces.
- Fixed browser tab.fill timing out on pages whose animation frames stall.
- Fixed the first LSP diagnostics request returning no results while a newly started language server is still analyzing.
- /shake thinking now reports the number of tokens freed.
@oh-my-pi/pi-natives
Changed
- Improved warmHighlighter() so it prepares TypeScript, TSX, JavaScript, Bash, Python, Rust, and Markdown highlighting in the background, reducing delays on the first highlight for these languages.
@oh-my-pi/pi-tui
Added
- Added a Skills section to the settings.
- Enabled OSC 8 hyperlinks in rio.
Fixed
- Fixed scrolling startup release notes without requiring an initial Escape keypress.
- Improved usage quota display so names remain distinguishable and readable in narrow and multi-column dashboards.
@oh-my-pi/pi-wire
Added
- Added public API contract and authentication schemas for the Skillshare registry.
What's Changed
- fix(coding-agent/tools): keep native JSON Schema roots out of JTD conversion by @roboomp in https://github.com/can1357/oh-my-pi/pull/12896
- fix(browser): drive tab.fill and frame actions off puppeteer Locator by @roboomp in https://github.com/can1357/oh-my-pi/pull/12898
- Add Pi-compatible RPC primitives get_entries(since), get_tree, get_available_thinking_levels to RPC protocol by @44madfire in https://github.com/can1357/oh-my-pi/pull/12900
- fix(browser): tolerate missing Puppeteer callsites by @Dante-dan in https://github.com/can1357/oh-my-pi/pull/12902
- fix(tui): preserve distinguishing usage quota labels by @lockwo in https://github.com/can1357/oh-my-pi/pull/12903
- fix(coding-agent): explain model fallback causes by @lockwo in https://github.com/can1357/oh-my-pi/pull/12904
- fix(catalog): route Copilot Grok 4.x through the Responses API by @roboomp in https://github.com/can1357/oh-my-pi/pull/12905
- fix(tui): make startup changelog scrollable after welcome animation by @michaelr in https://github.com/can1357/oh-my-pi/pull/12906
- feat(coding-agent): expose before_subagent_spawn hook for extensions by @aloyzzz in https://github.com/can1357/oh-my-pi/pull/12907
- fix(coding-agent): report tokens freed by /shake thinking by @Gablinas in https://github.com/can1357/oh-my-pi/pull/12916
- feat(catalog): added StepFun provider with live model discovery by @ArpitMishra17 in https://github.com/can1357/oh-my-pi/pull/12918
- fix(todo): refresh tracker after nested eval updates by @tommymorgan in https://github.com/can1357/oh-my-pi/pull/12921
- feat(tui): enable OSC 8 hyperlinks on rio by @abcfy2 in https://github.com/can1357/oh-my-pi/pull/12922
- feat(coding-agent): configure extended context for custom providers by @LingLambda in https://github.com/can1357/oh-my-pi/pull/12923
- fix(agent): stopped background completions f
18.2.10 22 de setembre del 2026 — 64 bits Portable
18.2.9 22 de setembre del 2026 — 64 bits Portable
18.2.8 21 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added support for text embeddings, document reranking, video generation, image generation across multiple providers, audio speech synthesis, and audio transcription services.
- Added support for the System One judgment API, including configurable request headers for proxy routing and custom authentication.
Changed
- Updated API response cost reporting to use aggregate usage totals.
- Model list responses now optionally include a model kind.
Fixed
- Fixed detection of Claude usage-limit errors.
@oh-my-pi/pi-catalog
Added
- Expanded OpenRouter provider support with embedding, reranking, video generation, text-to-speech, and speech-to-text capabilities, including five new speech-to-text models.
- Added speech-to-text support to the OpenAI provider.
@oh-my-pi/pi-coding-agent
Added
- Added comprehensive browser automation tools for accessibility auditing, React inspection, console and network monitoring, performance tracing, semantic DOM queries, tab management, screen recording with cursor overlays, downloads, custom initialization scripts, persistent storage, and WebMCP cross-frame tool discovery.
- Added support for buffered cloud transcription with OpenAI-compatible models.
- Added visual change detection for video processing, including FFMPEG analysis and SVG overlays.
- Added support for declaring native judges through custom providers using the typesafe and openrouter-decisions API values, with configurable base URLs, API keys, and headers.
Changed
- Expanded browser security and resilience controls with configurable HTTPS error handling, domain allow-listing, and automatic tab recycling when security-sensitive state changes.
- Updated background job notifications to deliver output as follow-up messages and discourage unnecessary polling.
- Expanded the bash tool's documented auxiliary utilities and removed its truncation footer notice.
Fixed
- Improved responsiveness in long sessions by significantly reducing the time required to scan provider context for credential patterns.
- Fixed native judges failing to honor configured request headers, enabling authenticated and header-routed judge providers to work as configured.
- Fixed LSP requests hanging when aborted while waiting for an earlier write to complete.
@oh-my-pi/pi-tui
Changed
- Improved Bash tool background-task notices by providing completed output as a follow-up and discouraging unnecessary polling.
What's Changed
- fix(judgment): support custom headers and models.yml config for native judges by @Thisen in https://github.com/can1357/oh-my-pi/pull/12678
- perf(secrets): gate built-in credential regexes behind literal prefix probes by @DarkPhilosophy in https://github.com/can1357/oh-my-pi/pull/12666
New Contributors
- @Thisen made their first contribution in https://github.com/can1357/oh-my-pi/pull/12678
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.2.7...v18.2.8
18.2.7 21 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Breaking Changes
- Anthropic streaming and provider request helpers must now be imported from @oh-my-pi/pi-ai/providers/anthropic instead of the package root.
- Moved the public NO_AUTH_SENTINEL export from providers/openai-shared to auth-retry.
Fixed
- Anthropic organization-level OAuth permission errors now reliably rotate to sibling credentials and persist blocks across usage reports.
- Fixed error handling for provider responses that do not include token usage information.
@oh-my-pi/pi-catalog
Added
- Added model-kind and grounded-search capability metadata, along with catalogs for local inference and search-engine models.
- Added OpenRouter image-model discovery and live TypeSafe judge-model discovery.
- Added the buildDiscoveredModel helper for defining custom providers.
- Added glob-based patterns for identity overrides.
Changed
- Updated input cost for TypeSafe models to 0.042
- Improved model routing and thinking-policy handling for llama.cpp Qwen models, Bonsai lineage aliases, and custom provider names.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Image-generation overrides now use model selectors, and web-search CLI overrides use --model instead of --provider.
- Removed the bash tool's env parameter.
- Eval judge(state, questions) is now awaited and returns answers directly; JudgmentHandle and judgment support in wait() have been removed.
Added
- Added find tool for semantic workspace searching, allowing agents to locate behaviors and symbols using natural language
- Added find CLI command for performing semantic workspace searches
- Added batch evaluation with judge_batch(states, questions) / judgeBatch(...), including bounded background execution, incremental result and status access, per-item failure reporting, and the ability to wait for or reattach to jobs across turns or after a reset.
- Added the jevify magic keyword to have the agent establish an evaluation rubric before classifying bulk items and inspect only items flagged by the judge.
- Added omp web-search as an alias for omp search.
- Added tui.titleSpinner configuration to select the terminal-title working-state spinner (braille, dots, or line).
- Added Handlebars-based system prompt templates through SYSTEM_TEMPLATE.md, --system-prompt-template, and the SDK, with access to live settings and tool data.
- Added configurable image, web, speech, dictation, judge, and memory model roles with ordered fallbacks, legacy backend-setting migration, and omp models --kind filtering.
- Added native OpenRouter image generation, model-selected web-plugin search, and live discovery of TypeSafe judge models.
Changed
- Updated agent system prompts to prioritize the find tool over grep and glob for behavioral lookups
- Refined system prompt instructions for XML tag handling and agent persona
- Updated sloppy edit tool syntax to use plain text headers instead of XML tags
- Improved startup performance by validating provider-qualified model selectors against only the relevant provider catalog.
- Reduced launch time for npm and compiled builds by embedding the model catalog more efficiently.
Fixed
- Fixed system prompt configuration validation so systemPromptTemplate and customSystemPrompt cannot conflict with a full systemPrompt replacement, including when values are empty.
- Added browser-relay support for listing eligible pages without attaching to or claiming them.
- Fixed Codex compatibility with the sloppy edit tool.
- Capped concurrent eval judge and completion requests to prevent large fan-outs from overwhelming judge and fallback models.
- Temporarily avoids retrying judgment requests with credentials that recently failed due to authorization or billing errors.
- Fixed image and speech fallback models disappearing after discovery and eliminated incorrect incompatibility warnings for providers without credentials.
- Fixed resume and continue flows to hide empty sessions.
- Fixed edit operations that could loop after empty insertions or fail on Unic
18.2.6 18 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Anthropic prompt-cache head re-baselining on every memory recall refresh: the system breakpoint now anchors on the last stable segment instead of the volatile recall suffix, and the stable-system fingerprint ignores recall blocks, so a recall refresh re-bills only the suffix instead of the whole tools+system head.
- Fixed auth-broker client config resolution failing silently on Windows when reading the token file or config.yml; reads now use node:fs instead of Bun.file.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed clipboard paste stalling on an empty clipboard; image and text clipboard reads now run concurrently so the empty-clipboard status surfaces after the slower read instead of the sum of both.
- Fixed memory recall blocks carrying a minute-resolution Current time stamp that dirtied the cached system prompt on every refresh; recall rows already carry dates, so the stamp is removed.
- Fixed omp auth-broker token and omp auth-gateway token exiting silently without creating a token on Windows when no token file exists yet; token and config reads now use node:fs instead of Bun.file.
What's Changed
- fix(auth-broker): use node:fs for token and config reads on Windows by @H4vC in https://github.com/can1357/oh-my-pi/pull/12390
- docs(changelog): restore contributor attribution for #12177 and #12236 (@shivamklr) by @shivamklr in https://github.com/can1357/oh-my-pi/pull/12396
- fix(ai): keep prompt-cache head stable across memory recall refreshes by @H4vC in https://github.com/can1357/oh-my-pi/pull/12431
- fix(coding-agent): run clipboard image and text reads concurrently on paste by @H4vC in https://github.com/can1357/oh-my-pi/pull/12447
- fix(ai): remove dead breakpoint locals in head-caching recall test by @H4vC in https://github.com/can1357/oh-my-pi/pull/12453
- fix(build): pin find-msvc-tools to 0.1.12 for Windows cc compat by @H4vC in https://github.com/can1357/oh-my-pi/pull/12455
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.2.5...v18.2.6
18.2.5 17 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Improved agent performance by reducing redundant tool-schema processing during repeated model calls and optimizing streamed tool-call argument parsing.
@oh-my-pi/pi-ai
Added
- Added support for templating and custom base and authentication URLs in OAuth flows.
Fixed
- Fixed Anthropic prompt-cache breakpoints stalling when conversations include mid-conversation tool changes, preventing growing message tails from being unnecessarily re-billed as uncached input.
@oh-my-pi/pi-catalog
Added
- Added the stencil authentication provider for omp stream, supporting OAuth code + PKCE sign-in with auth.stencil.so, configurable via STENCIL_API_KEY, STENCIL_AUTH_URL, and STENCIL_BASE_URL. This is an authentication-only provider, not a model provider; OAuth-code login configuration also supports base-url and auth-url nodes with {base} and {auth} URL placeholders.
Fixed
- Corrected Yolo-Auto metadata for Qwen Flash: qwen3.8-flash and the paid yolo route now report the documented 256K context window and use the Qwen chat-template reasoning dialect, with qwen3.8-flash as the provider default.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Moved terminal UI modules—including themes, tool renderers, chat, overlay, status-line, composer, setup wizard, and Git/PS/debug apps—to @oh-my-pi/pi-tui. The corresponding @oh-my-pi/pi-coding-agent subpaths no longer exist; names re-exported from the package root remain unchanged.
Added
- Added omp stream for livestreaming terminal sessions at live.omp.sh/<your Stencil username>, with viewer chat, pane-per-session display for sessions in the same directory, screen redaction, and configurable stream.serverUrl and stream.redactPatterns settings. Use --server to override the stream server, --title to set a title, and --no-tui to retain the line-based log interface.
- Added Stencil account support to /login. omp stream uses a signed-in Stencil account or STENCIL_API_KEY for channel ownership and authentication. Sensitive environment, dotenv, secrets.yml, credential-shaped, and configured pattern-matching values are redacted before screen data is transmitted.
- Added faster keyless web search fallback by prioritizing the default keyless Parallel provider ahead of Perplexity.
Changed
- Improved parent IRC message prompts to make interruption handling more reliable.
- Improved subagent task labels and plan filenames to use concise, action-oriented descriptions.
- Updated CLI byte sizes to use decimal KB units and made duration displays coarser and easier to read.
Fixed
- Fixed edit auto-repair waiting up to 60 seconds when the smol model does not respond; it now times out after 20 seconds and reports repair start and timeout details.
- Fixed subagents leaving queued parent messages behind after tool interruptions.
- Fixed a subagent burning its whole run on yield calls that never finish it: an incremental-only yield turn no longer bypasses the request budget, and the forced final yield ends the run (#12351 by @pedropaulovc).
- Fixed browser.open({ app: { relay: true } }) waiting for the full tool timeout when no relay extension is installed or reachable; it now fails promptly with an actionable error while preserving the wait for a connected extension to recover.
- Fixed edit handling of ellipsis markers, inline closing tags, copy-ready corrections, and retries, including cases that could insert literal markers, misreport matches, omit the file target, or panic.
- Enabled edit.enforceSeenLines by default to reject hashline edits anchored to content that was not displayed, and prevented stale-tag recovery from applying edits to a structurally different duplicate construct (#12369 by @pedropaulovc).
- Fixed startup failures when the plugins directory or its manifest cannot be read; inaccessible plugin roots are now skipped with a warning.
- Fixed generation token-rate displays for subagents and restored the main session's reading after switching focus.
- Fixed subagent HUD labels and plan
18.2.4 17 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added the judgment module for typed questions over JSON state, including choice, yes/no, and score judgments through the Judge interface.
- Added TypeSafeJudge support with TypeSafe System One authentication, credential rotation on unauthorized responses, and retry-aware backoff.
- Added TextJudge and chatTextBackend for model-based judgments, with structured state rendering and safeguards that prevent embedded requests from being executed.
- Added automatic format-correction retries to TextJudge when models return malformed output.
- Added the guardState option to TextBackend to control whether safety guidance is included in prompts.
@oh-my-pi/pi-catalog
Added
- Added typesafe authentication for TypeSafe System One judgments via the TYPESAFE_API_KEY configuration and API-key validation against the TypeSafe models endpoint.
@oh-my-pi/pi-coding-agent
Added
- Added an optional live generation speed readout via composer.tokenRate, showing smoothed tokens-per-second output in the working row and keeping the rate visible between turns.
- Added TypeSafe provider support through /login typesafe or TYPESAFE_API_KEY. TypeSafe can power thinking-level detection, unexpected-stop detection, and AI-assisted git staging with calibrated judgment probabilities; configure providers.judgmentProvider as auto, typesafe, or llm to select the judgment backend.
- Added the judge(state, questions) evaluation helper for Python and JavaScript cell code, supporting typed choice, boolean, and score judgments. It returns a handle whose .wait() method provides answers and probabilities, using TypeSafe when configured and available or a fallback chat model otherwise.
Changed
- Unified thinking-level detection, unexpected-stop detection, and AI-assisted staging around a shared judgment system with automatic fallback across configured models when TypeSafe is unavailable or cannot complete a request. AI-assisted staging now evaluates files as a single batched judgment while preserving one yes/no decision per file.
@oh-my-pi/pi-tui
Fixed
- Fixed inline images disappearing or temporarily blanking when resizing the terminal in kitty and Ghostty.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.2.3...v18.2.4
18.2.3 17 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- stream() and streamSimple() support asynchronous model header resolution for each request attempt, including authentication retries and cancellation.
- Provider login prompts can request masked entry with secret: true.
@oh-my-pi/pi-catalog
Added
- Models can carry deferred request-header resolvers, and model managers can reconstruct omitted cached headers from authoritative local configuration without persisting credentials.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Config-backed headers now resolve asynchronously through ModelRegistry.getProviderHeaders() or resolveModelHeaders(); removed the synchronous config/model-config-values module.
- Removed the unused ConfigFile.getMtimeMsAsync(), tryLoadAsync(), loadAsync(), and loadOrDefaultAsync() methods.
- Custom SQL session clients must support transactions for atomic renames.
Added
- Type ^ to tag a model for delegation, with atomic display-name chips and session-persisted m1, m2, … agents available to task and eval.
- Provider login and setup support masked secret prompts; RPC rejects secret prompts rather than requesting ordinary input.
Changed
- Shell-backed API keys and headers resolve asynchronously without freezing terminal input or running during catalog construction.
Fixed
- macOS process discovery now retains the complete PID list when locating executables and descendants. (#12290 by @iliaal)
- Reduced snapshot-recording stalls when a session retains large file histories. (#12279 by @iliaal)
- Cancelled background jobs remain tracked until execution finishes, so cleanup cannot report completion prematurely after retention expires. (#12278 by @iliaal)
- Fixed localized edits rewriting unrelated bytes in files with invalid UTF-8; these edits now fail without modifying the file. (#12277 by @iliaal)
- Fixed sloppy edits crashing with a char-boundary panic instead of reporting a match error when the file contains multibyte (e.g. CJK) text.
- Fixed retry timing reliability in agent sessions by ensuring sleep durations are monotonic
- Fixed data stability issues when processing streamed lines
- Resolved same-path move failures in indexed session storage
- Restricted and revived subagents retain parent-loaded extension hooks without enabling extension-contributed tools.
- Revived subagents honor the owning session's extension-discovery restrictions.
- Secret login answers stay hidden in later prompts and cannot be recovered through undo or yank.
- SQL session renames preserve data on same-path moves, missing sources, and failed overwrites.
- MySQL session writes no longer use deprecated upsert value references.
- MCP SSE requests honor one response deadline and report timeouts correctly without replaying accepted tool calls.
- Legacy extension package-import patterns follow native prefix precedence.
- Bundled extensions observe theme initialization and changes through the existing live theme export.
- Configured discovery models retain request-time credentials after offline cache reloads and failed refreshes.
- Runtime API-key overrides retain precedence over configured credentials.
- Element handles returned by tab.waitForSelector, tab.$, and related selector helpers can now be passed as arguments to tab.evaluate inside tab.run instead of failing with "JSHandles can be evaluated only in the context they were created".
@oh-my-pi/pi-tui
Added
- Editors support whitespace-delimited ^ mention autocomplete and expose registered atoms for host-defined chip rendering.
Changed
- fuzzyRank accepts readonly candidate arrays without copying them.
Fixed
- Masked inputs no longer expose their text in diagnostic previews.
Removed
- Removed the internal fastTailSplices and resetFastTailSplices Markdown instrumentation exports.
@oh-my-pi/pi-utils
Fixed
- Optimized model configuration command execution by deduplicating requests and adding failure backoff
- Prevented unnecessary credential command execution when runtime API keys are configured
- Retained readLines() 
18.2.2 16 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added configurable baseUrl support for bedrock-converse-stream requests, enabling Amazon Bedrock providers and compatible custom providers to use VPC or PrivateLink endpoints, FIPS hosts, and internal gateways, including endpoints mounted under a path or authenticated with query parameters.
Fixed
- Corrupt credential databases are now backed up privately and recreated instead of preventing startup; signing in again restores credentials.
- Fixed malformed Anthropic thinking signatures that could freeze sessions at 100% CPU.
- Fixed Anthropic-compatible gateway tool-call handling so client-declared tools are reported with stop_reason: "tool_use" and already-executed native provider tools are not exposed for clients to run again.
- Fixed gateway session isolation when clients omit a session key, preventing one conversation's retained provider state from affecting another.
- Fixed retained provider state across credential switches so account-specific capabilities are re-evaluated while reusable endpoint capabilities remain available.
- Fixed session retention limits closing provider state while a request is still streaming.
- ChatGPT accounts that have exhausted a plan's usage window but still have available Codex credit can now continue to be selected for Codex requests.
- Cursor requests now honor explicit max-mode markers on wire-backed models instead of inferring the mode from the model suffix.
- OpenAI-compatible chat responses containing only structured tool calls now report time to first token correctly.
@oh-my-pi/pi-catalog
Added
- OpenAI-compatible model discovery now fills in reasoning-effort tiers for unrecognized models using the shared catalog’s published reasoning options, while preserving explicit discovery metadata and reviewed model rules.
Fixed
- Fixed recovery of corrupted model caches so private backups are preserved and concurrent recovery cannot overwrite a cache that has already been restored.
- Fixed pricing for Devin (SWE-2, SWE-1.7, and GLM-5.2 High) and Kimi Code models when upstream discovery omits cost information. SWE-2 now reflects its promotional pricing through December 31, 2026, then switches to list pricing on January 1, 2027.
- Fixed pricing and chat routing for Devin Fusion models so composite models use their own headline rates and supported Fusion lanes connect directly instead of failing through an incompatible routing path.
@oh-my-pi/pi-coding-agent
Added
- Expanded built-in secret obfuscation to detect credentials in connection URLs regardless of environment-variable name, including PostgreSQL, MongoDB, MySQL, Redis, AMQP, and other supported schemes.
- Expanded built-in secret obfuscation to cover AWS access keys, Google API keys, Slack, npm, Stripe secret/restricted keys and webhook secrets, Hugging Face and SendGrid tokens, JWTs, Bearer tokens, and PEM private keys.
- Added the tui.titleSpinner setting to choose the terminal-title working-state animation (braille, dots, line, or pulse), alongside the existing tui.titleState toggle.
Changed
- Session-stop hooks that block with a reason now keep the session running until they allow it or the user interrupts; explicit aborts are no longer restarted by a stop hook.
- Corrupt agent and prompt-history databases are now backed up before fresh stores are created, allowing startup to continue; credentials may need to be entered again.
- Agent and history database startup errors now identify the affected database file.
- Terminal-title spinner animations now work on native Windows; WSL retains the static separator to avoid unnecessary CPU usage.
- Explicit model refreshes now re-evaluate command-backed API keys and headers, allowing rotated credentials to take effect without restarting.
- Background job entries are removed shortly after their results are consumed or recovered, while unconsumed jobs remain available for inspection.
Fixed
- Fixed the transcript collapsing into a compact no-spacing layout whenever the prompt,
18.2.1 15 de setembre del 2026 — 64 bits Portable
18.2.0 15 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Assistant turns from Anthropic-compatible hosts (direct, or via OpenRouter's reasoning_details) carry upstreamModel, the serving model id recovered from the signed thinking block, so callers can detect a router substituting a different model than requested.
Fixed
- Fixed OpenCode Go window-limit 429s (5-hour/Weekly/Monthly usage limit reached. Resets in …) not pinning the exhausted credential to the server-stated reset; the window phrasing is now covered by a regression test over the rotation classifier. (#12091 by @H4vC)
@oh-my-pi/pi-catalog
Breaking Changes
- Removed getCatalogProviderEntry and the CATALOG_PROVIDERS constant in favor of providerEntry and providerEntries
Added
- Added support for DeepInfra, LiteLLM, Qwen Portal, SiliconFlow, and SiliconFlow-CN providers
- Added support for Gemini 3.8 Flash models
Changed
- Cached provider catalogs restore policy-versioned materialized models instead of rebuilding each row on launch.
- Model policy resolution reuses indexed rule matches and cached target results.
- Model-aware delegation prompts reuse policy decisions until the model's identity or capabilities change.
- Provider catalog entries (default model, env keys, discovery wiring) and the bundled fallback rows for providers that cannot be discovered at generation time (Anthropic, OpenAI Daybreak, xAI OAuth, Meta, Muse Code, Bedrock Mantle, Devin, Z.AI, Sakana, ai&, Abliteration, Yolo-Auto, GMI Cloud, Fire Pass, QwenCloud Token Plan, Cloudflare AI Gateway, GitLab Duo Workflow) now live in src/compat/rules/providers/<id>.kdl and compile into rules.json; KnownProvider is generated from them, and the generator bundles seed rows by each entry's declared bundle policy instead of per-provider code.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Settings.getGroup() now returns shallow-frozen snapshots, reused until effective settings change.
- Removed parseSSE, MCPToolsResponse, and MCPCallResponse; callMCP() now returns the shared JsonRpcResponse with an unknown result instead of an unchecked generic payload.
Added
- Added ollama web search provider using Ollama's hosted web search API (POST https://ollama.com/api/web_search), authenticated via OLLAMA_CLOUD_API_KEY (#3791).
- Added readUrl support for Ollama model pages (ollama.com/<model> and ollama.com/library/<model>), extracting descriptions, tags, and architecture metadata.
- @upstream routing selectors accept tiered OpenRouter slugs (openrouter/google/gemini-3.8-flash@google-ai-studio/priority), and omp bench labels each routed model with its upstream.
- /skill:<name> in the composer becomes an atomic skill chip (icon + name, linked to its SKILL.md) once you finish typing it or accept it from autocomplete — it deletes as one unit and survives draft restores, like image chips.
- The transcript now flags a gateway serving a different Claude model than requested: a ⚠ served claude-haiku-4-5-20251001 · requested claude-opus-5 · via openrouter/Amazon Bedrock divider under the first affected turn, shown once per substitution per session.
Changed
- Compiled binaries ship precompiled bytecode: omp boots in ~30 ms instead of ~250 ms and the interactive prompt accepts input ~300 ms sooner, at the cost of a larger binary.
- Welcome recents refresh after first paint, and attachment bands reuse cached chip state until the draft changes.
- Interactive startup paints its speculative frame before loading the session runtime; model/auth dialogs and browser/computer preludes load on first use.
- Status-line redraws reuse unchanged segment output, settings groups, and tool token estimates while preserving live invalidation.
- Skill invocations render as a normal user turn: a mid-prompt skill shows as an inline chip in the user bubble; a leading skill shows as a railed callout with the chip and prompt size, with the rest of your message rendered as full multi-line Markdown instead of a single collapsed header.
Fixed
- hub wait on a process now reports what it was actua
18.1.22 14 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- 400-request debug dumps now redact provider-specific auth headers (x-goog-api-key, x-amz-security-token, and any header whose name carries a key/token/secret), not just a fixed allow-list, so a shared dump can no longer leak a live API key (#12007).
@oh-my-pi/pi-catalog
Added
- Enabled assistant prefill support for Ollama models
Fixed
- OpenCode Zen GPT-6 Astra requests now use the Responses endpoint instead of failing through chat completions with HTTP 500 (#12030).
@oh-my-pi/pi-coding-agent
Breaking Changes
- Hub message/job waits now always use the adaptive window (5s, lengthening to 5m across back-to-back waits); removed the timeoutMs argument and async.pollWaitDuration setting.
Added
- Added a privacy warning to memory reports reminding users to review data for secrets before sharing
- omp git / /git: delete discards the selected file's changes (press twice to confirm) — in the sidebar on a file or whole directory, in the diff pane on the shown file; untracked files are removed, staged files reset to HEAD
Changed
- Pressing c on a /btw answer now shows a green "✓ Copied to clipboard" confirmation in the panel and history detail, and BTW history accepts Ctrl+/ to switch panes (#12052 by @H4vC).
Fixed
- Automatic session titles no longer draw from canned prompt examples.
- Sessions titled by a local Ollama model (e.g. LFM2.5) no longer stay unnamed when the model's chat template spends the whole output budget on reasoning.
- /debug memory reports now include numeric memory statistics instead of raw heap snapshots that could expose provider and MCP credentials.
- Multi-step logins (e.g. Perplexity email → code) now move the input field under the latest prompt instead of leaving it stuck beneath the first one.
- Todo updates made through Eval's tool.todo(...) now persist to the session, so they survive resume/rewind/fork and no longer trigger false incomplete-todo reminders.
- Native background security scans now accept provider-owned AWS authentication for Amazon Bedrock and Bedrock Mantle without requiring a stored OAuth account (#12013).
@oh-my-pi/pi-utils
Fixed
- Fixed extractRetryHint sleeping hours past the provider's stated wait when a timezone-naive reset at timestamp overshoots the relative retry hint: the skewed stamp is now ignored instead of winning longest-wins (#12070 by @H4vC).
What's Changed
- fix(catalog): route Zen Astra through Responses by @roboomp in https://github.com/can1357/oh-my-pi/pull/12032
- fix(ai): redact provider-specific auth headers in 400 dumps by @roboomp in https://github.com/can1357/oh-my-pi/pull/12010
- fix(security): support provider-native Bedrock auth by @roboomp in https://github.com/can1357/oh-my-pi/pull/12017
- feat(coding-agent): confirm /btw copies with visual feedback by @H4vC in https://github.com/can1357/oh-my-pi/pull/12052
- fix(utils): ignore zone-skewed naive reset-at past relative retry hint by @H4vC in https://github.com/can1357/oh-my-pi/pull/12070
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.21...v18.1.22
18.1.21 14 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Flatpak Chromium launcher executables (including com.google.Chrome, org.chromium.Chromium, and io.github.ungoogled_software.ungoogled_chromium) so app.path is treated as a browser and gets managed Chromium profile handling
- Fixed Chromium --user-data-dir handling by normalizing --user-data-dir <dir> and relative profile paths to absolute --user-data-dir=... values before launch
- Browser automation now works alongside an already-running Chrome using an isolated profile, keeps requested profiles separate, and never kills reused browser processes.
- First-use Chromium installation and browser operations no longer consume Eval's runtime timeout or reset its kernel while waiting.
- Browser startup reuses a successful system-Chrome fallback instead of retrying an unavailable download during the same open.
- Browser clicks and other interactions no longer stall when OMP-owned tabs are in the background, including after worker timeout recovery.
@oh-my-pi/pi-utils
Added
- Added getBrowserProfilesDir() (~/.omp/browser-profiles; XDG: $XDG_STATE_HOME/omp/browser-profiles) for profiles of Chromium browsers spawned by the browser tool.
Fixed
- Timed out stalled Chrome-for-Testing metadata requests after 30 seconds when looking up download metadata
- Concurrent browser installations share one download without replacing a running browser, and stalled downloads time out with partial files cleaned up for retry.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.20...v18.1.21
18.1.20 13 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Windows OAuth sign-in failing on every attempt after an upgrade when a previous run left a stale native callback registration behind; handlers registered by older binaries are now recognized as owned and rolled back instead of blocking recovery (#11967 by @H4vC).
@oh-my-pi/pi-catalog
Fixed
- Fixed DeepSeek V4.1 Flash dropping image attachments on hosts without their own carve-out, so the natively multimodal model is no longer treated as text-only (#11992 by @infernix).
- LiteLLM model groups keep the image input their deployment declares, instead of having attachments replaced with the "model does not support vision" placeholder (#11982, #11985 by @lz37).
- Fixed OpenCode Zen/Go Muse Spark models failing every tool-call turn with a 400 "reasoning encrypted_content was not issued to this caller" error: the gateways proxy the Responses lane to Meta but can't round-trip encrypted reasoning, so those SKUs no longer request or replay it (#11928).
@oh-my-pi/pi-coding-agent
Added
- Added collab.autoStart (off | view | control): every local interactive session hosts itself as it starts and rotates its room on /new, /resume, fork, or branch, so a phone or dashboard can reach any running session without running /collab first (#11908 by @alphastorm and @sorphwer).
- Added omp collab list [--json] and /collab list to enumerate every live local Collab host (instance, generation, session, cwd, model, participants, relay/attention state, access) without exposing links, plus omp collab link <instanceId|pid> [--view] to fetch one generation-bound browser URL from a private per-room Unix socket/named pipe registry; room keys, write tokens, and URLs never touch disk (#6099; #11908 by @alphastorm and @sorphwer).
Changed
- Documented that native JS/TS hook factories must live in .omp/hooks/pre/ or .omp/hooks/post/ (not directly in .omp/hooks/), and cross-linked the hooks and extension-loading docs (#11942).
Fixed
- The hidden notice announcing a mid-session tool-availability change now states that it lists only what changed, so an additions-only notice no longer reads as the complete tool set and the model keeps using tools that are still callable (#11824 by @camjac251).
- TTSR stream buffers now reset at every assistant message boundary, not only at turn start, so a scope: text or tool-argument rule can no longer fire on a later message because of text streamed by an earlier response in the same turn (#11957 by @srobroek).
- Eval completion() calls now use configured retry fallback chains when their role model fails (#11989).
- Eval completion() fallback chains now also apply to unqualified role models, walk into a failed fallback's own model chain, stop at retry.maxRetries, and resolve session-sticky credentials with the session id (#11989).
- Eval completion() fallbacks now keep depth-first chain order, inherit the failed candidate's effort for bare nested entries, and skip keyless candidates without spending retry.maxRetries budget (#11989).
- Eval completion() fallbacks reached at different efforts now each walk their shared descendants instead of truncating the later effort's path (#11989).
- Fixed ranged grep rejecting existing files with glob characters in their names (#11977).
- Notified Collab guests when admitted prompts are discarded, including room retirement during a session change (#11908 by @alphastorm).
- Preserved pending Collab dialog answers across session-switch rollback without accepting them after commit, stop, or writer departure (#11908 by @alphastorm).
- Fixed prompts awaiting setup crossing a fork, branch, or tree-navigation commit, multi-question extension dialogs moving later questions to a replacement Collab room, and stale rooms blocking /collab or /join after a failed session change (#11908 by @alphastorm).
- Fixed background task cards missing their final completion or failure after an early result or live-session focus replay.
- Ranged reads on Windows no longer intermittently open th
18.1.19 12 de setembre del 2026 — 64 bits Portable
18.1.18 11 de setembre del 2026 — 64 bits Portable
18.1.17 10 de setembre del 2026 — 64 bits Portable
18.1.16 9 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Codex SSE streams that end without a terminal completion event now retry when replay-safe and remain transient errors when partial output prevents replay (#11349).
@oh-my-pi/pi-coding-agent
Added
- /rename without a title now generates a session name from recent conversation using the configured tiny model.
- Added opt-in experimental notes-backed context windows with persistent branch-local notes, searchable original session history, retained latest user requests, and a model-callable rollover tool, including in Code Mode.
- /loop accepts --until '<cmd>' / --while '<cmd>' to gate each iteration on a shell command's exit status, so a loop can stop on real project state instead of only a count or duration. (#10858 by @andyhite)
Fixed
- Read error and preview rendering now sanitizes tabs and Windows-style CRLF (e.g. ssh host-key failures, tab-indented fetched content) so raw output can no longer tear the result frame.
- Unset tiny model roles now honor the configured @smol fallback in direct execution and the /models Roles view (#11311).
- Extension Control Center (/extensions) search now accepts j and k, so extensions like jira/json are searchable; bare j/k no longer move the list selection (use arrow keys or the configured tui.select.up/down) (#11350).
- Codex turns interrupted before terminal completion now auto-continue after resolved tool calls instead of stopping (#11349).
- Fixed the status line's pi brand/working segment double-padding the first separator, so every gap around a separator is a single space (#11103).
- /handoff no longer leaves the TUI in a running state when completion races with delayed session events (#11263).
- Fixed legacy Pi extensions failing to load when calling ctx.isProjectTrusted() in an event handler; the extension context now exposes it (always true, since OMP applies no project-trust gating) (#7955).
- Fixed Ctrl+Z jobs exiting successfully after fg instead of restarting the TUI because terminal teardown left Bun without a referenced event-loop handle while waiting for SIGCONT (#8585).
- Extensions loaded by the npm CLI now apply settings overrides to the active session, so generated agents and model choices remain isolated between sessions (#11047 by @mgpai22).
- Live task dispatch now reloads added, changed, removed, and deleted project task and retry settings before resolving subagents (#11191).
- Reset /loop iterations combined with --while / --until no longer keep submitting without resetting when vibe mode is enabled while the condition command is still running; the loop now disables itself instead (#10858).
@oh-my-pi/pi-utils
Fixed
- Fixed $which capturing Bun.which at import on Linux and Windows, so Bun.which stubs installed later (e.g. per-test spies) are honoured and PATH-only language servers no longer leak into test results.
What's Changed
- feat(coding-agent): add experimental notes-backed context windows by @wolfiesch in https://github.com/can1357/oh-my-pi/pull/10893
- fix(coding-agent): sanitize tabs and CR in read error rendering by @H4vC in https://github.com/can1357/oh-my-pi/pull/11305
- TUI: single delta marker for turn elapsed time in usage rows by @JamieJ5926 in https://github.com/can1357/oh-my-pi/pull/11332
- fix(agent): retry truncated Codex streams by @roboomp in https://github.com/can1357/oh-my-pi/pull/11356
- fix(tui): let /extensions search accept j and k by @roboomp in https://github.com/can1357/oh-my-pi/pull/11352
- fix(coding-agent): honor configured smol fallback for tiny by @roboomp in https://github.com/can1357/oh-my-pi/pull/11312
- fix(tui): clear stale loader after handoff completion by @roboomp in https://github.com/can1357/oh-my-pi/pull/11264
- fix(auth): refresh desktop database for native zcode:// oauth by @roboomp in https://github.com/can1357/oh-my-pi/pull/11287
- fix(update): validate foreign symlink targets by @roboomp in https://github.com/can1357/oh-my-pi/pull/11155
- fix(compaction): rearm idle timer after setting changes b
18.1.15 8 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- GitHub Copilot sign-in keeps the GitHub-owned Copilot CLI OAuth app on Enterprise domains: private instances run their own OAuth registry and reject the github.com-registered OpenCode client, which is now used only for public github.com sign-ins (#11285 by @H4vC)
- GitHub Copilot sign-in uses the minimal-grant OpenCode OAuth app again (read:user only): GitHub renders each app's existing per-user grant on the consent page, so Enterprise organizations that block the Copilot CLI app's broad historic grant can log in as on 18.1.4. API request identity still mimics the Copilot CLI, and tokens minted by either app keep working (#11280 by @H4vC).
- GitHub Copilot plan/model-policy 403s no longer count as credential failures for credential-lifetime decisions: the token is valid, so stored credentials are preserved instead of wiped (#11280 by @H4vC).
- Fixed custom google-generative-ai providers failing mid-turn model fallback when Gemini 3 tool calls are replayed without their original thought signature (#11270).
@oh-my-pi/pi-coding-agent
Added
- Added advisor.maxNotesPerUpdate setting and WATCHDOG.yml configuration (default 4): allows reasoning verifiers to batch findings in a single review update without being rate-limited.
- Headless browser tabs now freeze when a turn settles so idle animated/WebGL pages stop burning CPU/GPU, resuming automatically on next use; tabs idle past browser.idleCloseSec (default 30 minutes) are closed. persist: true on browser.open opts a tab out of both (#8246 by @H4vC).
Changed
- When enabled (task.showResolvedModelBadge), subagent model badges show the thinking-level icon, model name, and attached-advisor eye before the agent name in task, eval, job, and HUD rows.
Fixed
- Task descriptions containing tabs no longer misalign or overflow task rows; tabs are expanded before measuring and rendering.
- GitHub Copilot model-policy 403s (plan, model policy, org restriction) no longer delete stored credentials, so the provider stays listed in /model after a per-model access denial instead of disappearing until the next /login (#11280 by @H4vC).
- Bash results no longer replace a failing command's output with the shell minimizer's lossy summary when the original capture cannot be persisted as an artifact; the raw diagnostics are kept so a failure stays actionable (#11081).
- Fixed worker subprocesses failing to declare themselves as worker hosts before dispatching selectors, which prevented nested thread worker spawns during /usage stats sync on multi-core systems.
- Fixed /usage displaying a misleading generic database read failure when activity loading fails; the error detail is now sanitized, collapsed to a single line with shortened paths, and surfaced in the dashboard.
- Advisor notes now report rate limiting accurately, blockers always interrupt even after a lower-severity note in the same update, and deferred notes flush when the primary run completes, including after advisor quota exhaustion (#11062).
- Fixed the built-in clangd registration omitting CUDA source and header files (.cu and .cuh) (#10782 by @alphastorm).
- Fixed ast_grep skipping CUDA headers and ignoring an explicit lang override for ambiguous file extensions (#10782 by @alphastorm).
- Python cells are no longer replayed automatically after a kernel crash, preventing duplicate side effects; the next call starts a fresh kernel.
- Session rewrites preserve open-reader snapshots and replacement identity when a rename needs an EPERM fallback.
- Fixed WorkPool children retaining a stale Gemini-formatted yield declaration when pooled items were installed or cleared.
@oh-my-pi/pi-natives
Fixed
- Fixed C++ language inference excluding CUDA header (.cuh) files (#10782 by @alphastorm).
@oh-my-pi/pi-tui
Fixed
- Stopped long flicker when moving or resizing an omp pane in Warp. Resize repaints in place there after the drag settles (override with PI_TUI_RESIZE_IN_PLACE=0), with no alternate-screen borrow, no scrollback replay, 
18.1.14 7 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed reasoning-off requests (e.g. GitHub Copilot gpt-6-astra) surfacing 400 Unsupported value: 'none' … Supported values are: … instead of retrying at the lowest allowed effort: the reasoning-effort fallback now recognizes Supported values phrasing (#11128 by @H4vC).
- Fixed Cursor GPT off-tier requests sending raw -none sibling ids (e.g. gpt-5.6-sol-none-fast), which the Run endpoint rejects; they now normalize to the base model id with no reasoning parameter, matching every other effort tier (#11128 by @H4vC).
@oh-my-pi/pi-catalog
Fixed
- Bills Astra API requests above 272K input at the documented 2x input / 1.5x output long-context tier; the Codex subscription route stays exempt with free cache writes (#11157 by @H4vC).
- Fixed Astra's extended window over-advertising input by 128K; it now uses the documented 922K input cap inside the 1.05M total context (#11157 by @H4vC).
- Fixed explicit Codex context-window overrides widening past the server-honored maximum; they now clamp to the documented ceiling like upstream Codex (#11157 by @H4vC).
- Fixed Codex Astra using its larger window without opt-in; its default is 272K and Extended Context enables at least the documented 1.05M window (#11126 by @H4vC).
- Fixed GitHub Copilot enterprise-only model ids inheriting another provider's wire routing (e.g. gpt-5.6-sol-fast pinning every request to the -none sibling id regardless of thinking level) (#11128 by @H4vC).
@oh-my-pi/pi-coding-agent
Fixed
- The startup update notice counts every change in a release: bullets written above a ### heading now count under Other, and +/* markers and lightly indented bullets count like -.
- Fixed Codex Astra retaining its larger window after disabling Extended Context, including cached models; explicit model overrides still take precedence.
- Fixed explicit Codex context-window overrides widening past the server-honored maximum; they now clamp to the documented ceiling like upstream Codex (#11157 by @H4vC).
- Fixed Astra's extended window over-advertising input by 128K; it now uses the documented 922K input cap inside the 1.05M total context (#11157 by @H4vC).
- Bills Astra API requests above 272K input at the documented 2x input / 1.5x output long-context tier; the Codex subscription route stays exempt with free cache writes (#11157 by @H4vC).
- Fixed Extended Context silently enabling without a settings source (SDK embedding, early boot); it now matches the off default until opted in (#11157 by @H4vC).
- Fixed /copy link captions showing Markdown delimiters for formatted labels and splitting across two rows for multiline labels (#11086 by @mustafaabidali).
- Fixed Ask custom answers requiring another submission after paste or remaining on the same multi-select question; pending clipboard text is preserved before submission, and single-question multi-select answers still go through review (#11099 by @camjac251).
- The startup update notice no longer counts standalone * * * and - - - separator lines as changes.
- Fixed /loop replacing the repeating prompt with a mid-turn interjection; steering while the agent runs is now one-off, and only an idle submission becomes the new loop body (#11159 by @H4vC).
@oh-my-pi/pi-tui
Fixed
- extractMarkdownLinks() now returns one-row visible labels for formatted and multiline links (#11086 by @mustafaabidali).
What's Changed
- fix(lsp): evaluate idle timeout per workspace client (#8389) by @harshaygadekar in https://github.com/can1357/oh-my-pi/pull/10237
- test(mcp): event-gate the reauth manual-input claim instead of a wall-clock budget by @rigel-mintaka in https://github.com/can1357/oh-my-pi/pull/11117
- fix(coding-agent): preserve and advance Ask custom answers by @camjac251 in https://github.com/can1357/oh-my-pi/pull/11099
- fix(tui,copy): flatten Markdown link labels to one visible row for /copy captions by @mustafaabidali in https://github.com/can1357/oh-my-pi/pull/11086
- fix(catalog,coding-agent): keep Astra 1.05M window behind ex
18.1.13 7 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-utils
Fixed
- Fixed filterChildShellEnv applying the omp process's own launch-environment provenance (the pre-dotenv NODE_ENV and launcher-owned names read from /proc/self/environ) to caller-supplied environment objects; launch provenance now only applies when filtering the live process.env/Bun.env, and an explicit env resolves its dotenv mode from its own NODE_ENV.
What's Changed
- fix(tui): deliver notifications inside Herdr panes by @lemonleks in https://github.com/can1357/oh-my-pi/pull/11054
- fix(utils): apply dotenv launch provenance only to the live process env by @pedropaulovc in https://github.com/can1357/oh-my-pi/pull/11074
- fix(catalog,coding-agent): default Astra to documented 1.05M window by @H4vC in https://github.com/can1357/oh-my-pi/pull/11089
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.12...v18.1.13
18.1.12 6 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added Muse Code subscription sign-in, credential refresh, inference, and quota reporting in /usage, with durable rate-limit backoff so quota refresh recovers instead of repeatedly retrying.
@oh-my-pi/pi-catalog
Added
- Added Muse Code as a provider with Muse Spark models and live account-scoped discovery.
- Muse Code subscriptions now resolve a compact edit-prompt variant, cutting recurring per-request tool bytes without touching other providers.
- Added Meta's new max reasoning effort tier to Muse Spark 1.3 (standard) on the Meta Model API and Muse Code.
Fixed
- Fixed OpenCode Go/Zen live model discovery (GET /v1/models) missing x-opencode-session and omp's User-Agent: discovery requests now attribute with the stable install id so the requests OpenCode flags as Bun fetch carry the required session header.
- Fixed GPT-6 Astra requests through GitHub Copilot failing with an unsupported endpoint error (#10874 by @xpcmdshell).
- Fixed GPT-6 Astra showing as free with a 272K-token window in the OpenAI Codex catalog by applying its documented pricing; /extended-context enables the wire-advertised 872K-token maximum (#10980 by @H4vC).
- Made extended-context catalog rebuilds faster by resolving each model's maximum window once per process (#11039 by @H4vC).
@oh-my-pi/pi-coding-agent
Changed
- Ranged reads of text without bracket characters skip unnecessary lexical context scanning.
- Muse Code sessions send a compact hashline edit description (~3 KB less per request); all other models keep the full prompt.
Fixed
- Fixed GPT-6 Astra extended-context support and preserved maximum context windows reported by OpenAI Codex discovery ([#10980](https://github.com/can1357/oh-my-pi/pull/10980) by [@H4vC](https://github.com/H4vC)).
- Subagent yield no longer rejects a valid data payload because a non-strict OpenAI-compatible backend filled the optional error field with ""; previously the worker retried the identical call until the invalid-yield cap and the parent received nothing.
- Fixed fullscreen /copy outlining only a lazily created grouped Read card, so Enter copies the assistant yield instead of tool output.
- memory:// now resolves against the session that issued it: a caller's own memory backend answers memory://<id>, so co-located sessions no longer read each other's memory rows, and a caller whose session is no longer live fails closed instead of being answered by a peer. Prompt completion binds to the same caller, so memory://<memory-id> stays on offer while a subagent shares the working directory. Advisors retain their owning session's memory access even without a session file.
- Fullscreen /copy now opens on the recent tail of the branch instead of replaying the whole session, so it appears immediately and steps without lag on long sessions (a loads the earlier turns). Both it and the esc-esc rewind selector also cache each transcript row set instead of re-stripping it every frame.
- Fixed the fullscreen /copy and esc-esc rewind selectors repainting the whole frame for a wheel notch that cannot move the viewport; because both open scrolled to the newest turn, wheeling down there made the frame twitch.
@oh-my-pi/pi-tui
Fixed
- Avoid inserting a trailing space when auto-completing directory paths with @, and keep autocomplete open when accepting a directory with Tab or Enter.
- Horizontal wheel reports (the sideways drift of a two-finger trackpad scroll) no longer decode as a vertical wheel direction, so fullscreen selectors such as /copy and the rewind picker stop jumping up and back down at the end of a scroll gesture.
What's Changed
- feat(ai): add Muse Code subscription provider by @eggpeat in https://github.com/can1357/oh-my-pi/pull/10677
- fix(catalog): route Copilot GPT-6 models through Responses by @xpcmdshell in https://github.com/can1357/oh-my-pi/pull/10874
- fix(eval): validate bridged tool arguments before execution by @DarkPhilosophy in https://github.com/can1357/oh-my-pi/pull/10953
- fix(catalo
18.1.11 5 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- GitHub Copilot sign-in now requests only basic profile access, restoring login for Enterprise organizations that reject repository, gist, and Codespaces permissions (#10656).
@oh-my-pi/pi-coding-agent
Added
- Added the retry.waitForUsageReset setting: when a provider reports usage-limit exhaustion with a reset time (5-hour or weekly quota windows on any provider), the session sleeps until the reset instead of failing fast past retry.maxDelayMs.
- Added opt-in bash.allowCompoundCommands approval for conservative literal && chains, with ordered per-segment rules and normal bash policy fallback for unmatched segments. The opt-in requires a positively classified POSIX-quoting shell; incompatible and unknown shells retain legacy approval. Whole-chain denies take precedence over earlier prompts.
Fixed
- Report oversized selected lines that cannot fit after read context, with a working raw recovery selector instead of a looping continuation hint (#10775).
- Fixed WorkPool child sessions crashing during startup while constructing their incremental yield tool schema.
- Commit summaries written in Vietnamese, Korean, and other accented scripts are no longer rejected for exceeding the length limit, and keep their accents as typed.
@oh-my-pi/pi-utils
Fixed
- Fixed extractRetryHint dropping the longer timing signal when an error body carries both an account reset and an appended retry hint: competing signals now merge by longest window instead of first match, so retries honor the provider's full backoff.
What's Changed
- fix(coding-agent): report omitted oversized read lines by @original4422 in https://github.com/can1357/oh-my-pi/pull/10798
- fix: retry transient gateway stream failures by @aktanazat in https://github.com/can1357/oh-my-pi/pull/10686
- fix(coding-agent): avoid WorkPool child startup crash by @eggpeat in https://github.com/can1357/oh-my-pi/pull/10755
- fix(coding-agent): update artifact byte-budget test for oversized-line notice by @H4vC in https://github.com/can1357/oh-my-pi/pull/10831
- fix(commit): recompose commit text after compatibility folding by @ntdatt812 in https://github.com/can1357/oh-my-pi/pull/10794
- fix(ai): reduce GitHub Copilot OAuth scopes by @roboomp in https://github.com/can1357/oh-my-pi/pull/10657
- docs(ai): relocate GitHub Copilot changelog entry to Unreleased by @H4vC in https://github.com/can1357/oh-my-pi/pull/10845
- feat(coding-agent): add retry.waitForUsageReset to sleep until usage limit reset by @H4vC in https://github.com/can1357/oh-my-pi/pull/10838
- feat(coding-agent): opt into literal compound bash approvals by @shawnkoh in https://github.com/can1357/oh-my-pi/pull/10919
- fix(pi-walker): resolve windows clippy errors by @H4vC in https://github.com/can1357/oh-my-pi/pull/10946
- fix(rust): resolve Windows clippy errors and maintain cross-platform signatures. by @H4vC in https://github.com/can1357/oh-my-pi/pull/10950
New Contributors
- @original4422 made their first contribution in https://github.com/can1357/oh-my-pi/pull/10798
- @aktanazat made their first contribution in https://github.com/can1357/oh-my-pi/pull/10686
- @ntdatt812 made their first contribution in https://github.com/can1357/oh-my-pi/pull/10794
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.10...v18.1.11
18.1.10 4 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed Codex V2 remote compaction ignoring explicit thinking-off.
- Fixed Codex V2 remote compaction rebuilding the request prefix differently from normal turns, restoring prompt-cache reuse (#10786).
@oh-my-pi/pi-coding-agent
Changed
- Subagent yield now takes data/error directly instead of nesting them under a result wrapper.
Fixed
- Fixed Codex V2 remote compaction rebuilding the request prefix differently from normal turns, restoring prompt-cache reuse (#10786).
- Restored mouse clicks, hover, and wheel scrolling in Plan Review.
What's Changed
- fix(compaction): preserve codex v2 cache prefix by @roboomp in https://github.com/can1357/oh-my-pi/pull/10789
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.9...v18.1.10
18.1.8 3 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added GPT-6 Astra support for preserving prompt caching when changing the thinking level during a conversation across the OpenAI and OpenAI Codex providers.
Changed
- Updated OpenAI Codex requests to improve routing by communicating the selected model and service tier across Responses, WebSocket, and remote-compaction requests.
@oh-my-pi/pi-catalog
Added
- Added GPT-6 Astra to the OpenAI Codex model catalog, including support for configuration updates and requests using the freeform apply_patch tool.
Fixed
- Fixed omp models refresh so revoked ChatGPT account tokens no longer prevent the remaining OpenAI Codex models from being discovered.
@oh-my-pi/pi-coding-agent
Fixed
- Improved background task results with structured output schemas: parsed results are now available through the agent://<id> resource, while large or invalid inline JSON is replaced with a reliable pointer to the complete result.
- Background task artifacts are retained long enough for follow-up turns to read them, including failed tasks that lack valid structured output, and are cleaned up without blocking shutdown or leaking resources.
- Fixed context compaction incorrectly accepting archived history that was larger because of opaque reasoning data, allowing the next compaction strategy to run instead.
- Fixed the Model Hub sidebar jumping to the top when provider refreshes rebuild the list; the focused model, or its nearest remaining entry, is now preserved.
- Fixed the inspect_image status hint showing the wrong model after switching between image-capable model roles.
- Fixed multi-minute TUI freezes during subagent activity and batch execution.
What's Changed
- fix(coding-agent): deliver structured output for background task spawns by @andyhite in https://github.com/can1357/oh-my-pi/pull/10625
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.7...v18.1.8
18.1.7 3 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed DeepSeek-family Responses replay (e.g. opencode-go) rejecting a resumed thinking-mode turn with 400 The reasoning_text in the thinking mode must be passed back to the API when compaction dropped the turn's reasoning; a non-empty placeholder is now synthesized instead of an empty reasoning_text (#10690).
@oh-my-pi/pi-coding-agent
Breaking Changes
- Removed the Ruby and Julia eval backends and related interpreter configuration; eval now supports Python and JavaScript only.
- Removed the eval parallel() and pipeline() helpers. agent() and completion() now return handles immediately, and wait(handles) provides synchronization.
- Python eval tool calls are now asynchronous coroutines, matching JavaScript; use await tool.read() and similar calls.
Added
- Added asynchronous eval agent and completion handles with status, cancellation, messaging, waiting, and automatic result delivery for unwaited background work.
- Added eval workpools for queueing items onto the least context-loaded keep-alive subagent with configurable concurrency; the pool name is its async-job ID for hub wait, .peek() gives a non-consuming snapshot, per-item {key, data|error} yields finish batches incrementally, and eval.workpool.freshAgents opts into a new agent per item.
- Added support for defining eval tools in Python with @tool or JavaScript with tool(fn, schema), and exposing them to subagents through task, agent, and workpool calls. Configure availability with eval.tools.enabled.
- Added native Windows ARM64 binaries with architecture-aware installation and updates.
- Added an MLX backend for running local tiny models on Apple silicon. Configure providers.tinyModelDevice=mlx, or use PI_TINY_DEVICE=mlx or metal, to run title generation, memory tasks, and automatic thinking classification with MLX models, with an ONNX CPU fallback when Python is unavailable.
- Added Qwen3 1.7B as a local memory and thinking-classification model for the MLX backend.
Changed
- Local tiny models for titles, memory, and automatic thinking classification now share on-demand workers across omp processes, reducing redundant resource usage; workers stop automatically after inactivity.
- PI_TINY_DEVICE=metal now selects the MLX backend on macOS.
- Updated agent reactions to trigger on the opening emoji instead of requiring a newline, consuming any following whitespace.
Fixed
- Fixed transient provider retries incorrectly failing with an “Agent is already processing” error.
- Fixed user-scope marketplace plugins installed through omp losing their skills when the Claude plugin source was not separately enabled.
- Fixed hashline edits failing when targets included apply_patch markers, while rejecting ambiguous bracketed targets instead of editing the wrong path.
- Fixed bracketed hashline edit targets being reported as undefined to extension path allowlists.
- Fixed MCP tools discovered during startup disappearing after plan-mode approval or when leaving default-on plan mode.
- Fixed ACP clients receiving invalid file locations or updates for released terminals, preventing invalid worktree scans and terminal errors on Windows.
@oh-my-pi/pi-natives
Added
- Added Windows ARM64 native addon support, including platform-specific npm packages.
@oh-my-pi/pi-utils
Added
- Added the public getTinyWorkerRuntimeDir() utility, which returns the standard ~/.omp/run/tiny directory for tiny-worker runtime data.
What's Changed
- fix(discovery): restore omp marketplace plugin skills by @roboomp in https://github.com/can1357/oh-my-pi/pull/10666
- fix(acp): filter invalid read locations by @roboomp in https://github.com/can1357/oh-my-pi/pull/10680
- fix(coding-agent): preserve MCP tools after plan mode by @roboomp in https://github.com/can1357/oh-my-pi/pull/10685
- fix(coding-agent): parse bracketed hashline event paths by @roboomp in https://github.com/can1357/oh-my-pi/pull/10689
- fix(ai): synthesize non-empty reasoning_text on DeepSeek Responses replay by @roboomp in https:
18.1.6 3 de setembre del 2026 — 64 bits Portable
18.1.5 3 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added /login abliteration with API key validation against /v1/models, supporting the ABLITERATION_API_KEY and ABLIT_KEY environment variables.
Changed
- Modernized provider authentication and token refresh across the catalog, with shared support for API-key, authorization-code, and device-code sign-in flows and clearer sign-in progress messages for OpenRouter, Kimi, and xAI.
Fixed
- GitHub Copilot now uses the official Copilot CLI identity and OAuth application for requests and new sign-ins, restoring access to client-gated models while preserving existing credentials.
- GitHub Copilot now reports model_not_supported responses immediately instead of repeatedly retrying unsupported models.
- Improved account recovery after Google rate limits are lifted earlier than the reported reset time.
- Fixed unmetered autocomplete models being reported as exhausted when quota is limited.
- Fixed Gemini 3 cross-model sessions in Cloud Code Assist when replaying tool calls without a thought signature.
- Fixed Cursor models behind an authentication gateway incorrectly retrying valid client-declared tool calls.
- Fixed reasoning from models that prefill <think> (including DeepSeek-R1 and hosted Qwen3-Thinking) being shown in the response instead of as a separate thinking block.
@oh-my-pi/pi-catalog
Added
- Added the Abliteration (abliteration.ai) provider, including its documented abliterated-model catalog and live model discovery.
- Added the GLM 5.3 Promo 50 model.
- Added computer-use capability metadata to model configurations.
- Added declarative provider authentication policies covering login, refresh, environment-key, and credential behavior, with generated compatibility data and typed accessors.
Changed
- Gemini 3.8 Flash now supports reasoning modes and image inputs.
- Updated the GitHub Copilot API version to 2026-08-01.
- Reduced input pricing for the minimax/minimax-m2 model.
- Renamed Meta Model API contributor SKUs to use the Muse Spark 1.x (C) naming.
Fixed
- GitHub Copilot model discovery now respects the Copilot CLI identity, ensuring eligible enterprise and experimental models are available.
- Fixed startup failures when discovering Bedrock-style Mistral Mixtral models.
- Fixed Muse Spark 1.3 contributor models on OpenCode gateways so they use the correct Responses API route.
- Updated Meta and OpenCode Muse Spark 1.3 model metadata and capabilities, including context windows, reasoning levels, image input, pricing, and model naming; media-only Muse SKUs are no longer presented as chat models.
@oh-my-pi/pi-coding-agent
Added
- Added Abliteration provider support to /login, including ABLITERATION_API_KEY configuration and help text.
- Added clone-first Git worktree support that carries over ignored build artifacts when creating worktrees, with a configurable worktree.clone setting and fallback to a standard checkout. This is supported by github pr_checkout, omp worktree add, and git worktree add commands entered through the Bash tool.
- Added the omp worktree add command with Git-compatible branch, detach, path, and commit options.
- Added /wt (alias /worktree) to create a linked worktree with uncommitted changes and move the current session into it while leaving the original checkout untouched.
Changed
- Foreign user-level configuration sources (~/.cursor, ~/.codex, ~/.claude, ~/.gemini, ~/.config/opencode, ~/.codeium/windsurf) are now opt-in via enabledProviders, while project-level configurations in CWD and .agents continue to load by default.
- Split subagent isolation configuration into task.isolation.enabled and isolation.backend; existing task.isolation.mode settings are migrated automatically.
- Updated the built-in smol and slow model priority chains to favor newer recommended models and remove older model generations.
- Improved unsupported-model error messages by removing retry guidance that does not apply.
Fixed
- Fixed automatic title generation so --no-title also prevents todo-initializat
18.1.4 2 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Changed
- Enabled Cursor tool schema projection for supported models
Fixed
- Antigravity and Gemini CLI now collapse every Gemini Flash generation from 3.6 on (gemini-3.8-flash-low/-medium/-high and the -tiered alias, and future revisions) into one routed gemini-<rev>-flash entry via a revision-templated variant-family, instead of surfacing raw per-level ids until a per-revision rule lands.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.3...v18.1.4
18.1.3 2 de setembre del 2026 — 64 bits Portable
18.1.2 1 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed preservation of Claude thinking/reasoning context when conversations are compacted or tool results are pruned.
@oh-my-pi/pi-ai
Added
- Added thinking controls for Amazon Bedrock models.
- Added dynamic mid-conversation updates for Anthropic system prompts, tools, and reasoning effort.
- Added deferred tool loading and prompt caching for Anthropic models.
- Added configurable handling for invalid Anthropic thinking blocks through anthropicPrefixMismatchBehavior.
Fixed
- Fixed compatibility issues with Anthropic thinking and prompt-cache breakpoints across deployments, preserving valid reasoning context while preventing invalid-signature errors.
- Fixed incorrect operating-system information reported in request headers on non-Linux systems.
- Fixed Google Antigravity quota handling so requests rotate to another account with available usage instead of unnecessarily switching models.
- Fixed Anthropic authentication for newer models by updating the Claude Code request fingerprint.
@oh-my-pi/pi-catalog
Added
- Added support for turn-scoped system messages, changing tools during a conversation, and setting effort on individual messages.
- Added support for Claude Fable 5.1 models, with improved thinking-prefix handling across supported Claude models.
- Added DeepSeek V4 Flash Vision Exp, Mercury 2.5 Preview, and Xiaomi MiMo V2.5 Pro UltraSpeed models.
Changed
- Updated model pricing and context-window limits.
@oh-my-pi/pi-coding-agent
Added
- Recover stray <SM:EDIT> payloads emitted as plain text into real edit tool calls, with support for disabling this behavior through the edit.recoverInlineEdits setting.
- Advisors now receive context from the active memory backend, including project decisions and recalled instructions; advisors also gain the recall tool when supported by the backend.
Changed
- Replaced the sloppy edit format's symbolic markers with a clearer XML-based format using <SM:EDIT>, <SM:FIND>, and <SM:PUT> tags. Edit errors now include copy-ready XML payloads.
- Increased the default input delay for the trace CLI to 3 seconds.
Fixed
- Improved chat history stability in long-running sessions by avoiding unnecessary updates when date or directory context changes.
- Fixed the trace CLI hanging during proxy connections and added support for forward HTTP proxies.
- Fixed newly started sessions using stale model context-window limits after background model discovery completes; the active model now refreshes automatically so context usage and compaction thresholds match the model catalog.
What's Changed
- fix(coding-agent): rebind active model after background discovery by @roboomp in https://github.com/can1357/oh-my-pi/pull/10491
- fix(mcp): expand Claude plugin stdio env placeholders by @mrexodia in https://github.com/can1357/oh-my-pi/pull/10481
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.1...v18.1.2
18.1.1 1 de setembre del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Fixed a native crash (and multi-gigabyte committed-memory growth held until exit) when git status ran over worktrees with tens of thousands of untracked files: whole-worktree porcelain status now runs through the git CLI with bounded output capture, falling back to the in-process gitoxide walk only when git is not installed, and any panic escaping a native VCS operation now surfaces as a structured VcsError instead of a process-level failure.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.1.0...v18.1.1
18.1.0 1 de setembre del 2026 — 64 bits Portable
18.0.11 29 d’agost del 2026 — 64 bits Portable
18.0.10 28 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added support for continuing interrupted agent runs with pending tool calls, allowing those calls to be retried before requesting the next model response.
@oh-my-pi/pi-coding-agent
Added
- Added the Sharpshooter memory backend for tracking friction-earned project decisions, with /memory queue and /memory sync controls.
- Added /restart to relaunch omp with its original launch flags and resume the current session in place.
- Added the band composer shape, a flush powerline status band above the prompt; it is now the default while existing composer.shape settings remain unchanged.
- Added in-place retry for interrupted or failed tool calls: use F5, Alt+R (app.retry), or /retry to replay an intact failed batch without an additional model round trip.
- Improved the working status display with a timed braille spinner, streamed intent, session accent colors across relevant status elements, and theme-aware session accent generation.
- Updated the unicode and ascii symbol presets to use π/pi for the brand icon, avoiding tofu on fonts without the nerd-font glyph.
Changed
- The /review command's PR-style comparison now uses the merge base against the current branch, excluding commits that exist only on the base branch; selecting the current branch reports no changes.
- Prompt history is now persisted immediately when submitted, and session database state is checkpointed on exit to improve durability and prevent unbounded WAL growth.
Fixed
- Fixed edit-tool parsing of --prefixed MATCH lines so they correctly represent whole-line deletions and can be replaced by a following + run.
- Fixed interrupted and failed Python evaluation cells being reported as successful results instead of errors, improving model handling, telemetry, retries, and background-job failure reporting.
- Fixed native-extension imports such as numpy hanging indefinitely in the Python evaluation tool on Windows.
- Fixed a macOS composer display issue where undercurl could remain attached to stale text after rapid typing.
- Improved xd:// MCP failure messages with actionable transport stages, failure categories, server and tool context, retryability, trace IDs, and redacted JSON-RPC details.
- Fixed ACP read tool-call locations so clients such as Zed Follow receive the resolved filesystem path rather than the OMP line-range selector.
@oh-my-pi/pi-natives
Added
- Added native process replacement support for the CLI’s /restart command.
- Added VcsGitRepo.mergeBase(a, b) to find the best common ancestor of two Git revisions, returning null when the histories are unrelated.
@oh-my-pi/pi-tui
Added
- Press the Right Arrow at the end of a line to accept autocomplete suggestions, inline ghost-text completions, and spelling corrections, just like Tab.
- Added a smooth brand-color transition and an elapsed turn timer to the status line while the agent is working.
- Added a full-width “band” composer style for flush status lines.
@oh-my-pi/pi-utils
Added
- Added postmortem.drainStdout to flush buffered standard output before process exit or exec-replacement.
- Added an exitOnly option to postmortem.register for resources that should remain available during keep-alive cleanup and be released only on actual process exit.
- Added hexToOklch and oklchToHex color conversion utilities with sRGB gamut mapping that reduces chroma when necessary.
- Added checkpointWal to checkpoint committed SQLite WAL frames without blocking concurrent readers.
Fixed
- Fixed repeatable postmortem cleanup behavior so persistent resources and callbacks registered during cleanup remain active until the eventual process exit.
- Fixed asynchronous postmortem cleanup so callbacks registered during a cleanup pass are awaited before cleanup completes, including during signal-driven exits.
What's Changed
- fix(eval): read python runner stdin serially to unblock numpy on windows by @roboomp in https://github.com/can1357/oh-my-pi/pull/7987
- fix(review): pass base and head as separate r
18.0.9 28 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed /shake elide handling of mixed tool results so images are preserved and token savings are reported accurately.
@oh-my-pi/pi-ai
Fixed
- Improved OAuth sign-in flows, including a fallback message when the browser cannot automatically close the OAuth success tab.
- Fixed Cloudflare AI Gateway onboarding and routing so gateway account and endpoint configuration is preserved correctly while gateway credentials are not sent as upstream OpenAI authorization headers.
- Fixed Codex OAuth quota handling so chat and Spark usage remain independent, legacy shared quota limits continue to work, and incomplete usage reports are not incorrectly treated as unlimited.
@oh-my-pi/pi-catalog
Added
- Added Z.AI GLM-5.3-Flash to the bundled GLM Coding Plan catalog, with a 1M-token context window, 131,072-token output limit, native image input, and low/high/max thinking levels. It is now available through model discovery and the model picker.
Fixed
- Fixed Google Antigravity Gemini 3.7 Flash tiered models failing with HTTP 400 errors at minimal or disabled thinking levels; those tiers now use the compatible Gemini 3.7 Flash model routing.
- Fixed OpenCode Go and Zen GLM-5.3 Flash models exposing an unsupported extra-high reasoning level; they now use the supported low/high/max reasoning levels.
- Fixed Cloudflare AI Gateway credential handling so stored credentials and routing metadata are parsed and used correctly.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Removed the git and jj wrapper modules from the SDK surface. VCS operations are now available through @oh-my-pi/pi-natives/vcs, including native handles and typed VcsError support; the package continues to re-export the github (gh CLI) helpers.
Changed
- extendedContext now defaults to off: models with premium long-context pricing tiers (e.g. GPT-5.6 1M) stay capped at their standard-pricing window unless the setting or /extended-context on enables the extended window.
Fixed
- Improved terminal readability on light backgrounds by ensuring TUI surfaces use contrasting foreground colors.
- Coalesced simultaneous autonomous continuation requests to prevent repeated calls while the agent is busy, with clearer continuation diagnostics.
- Fixed Snapcompact so it skips or falls back when compaction would not reduce context size, and now compacts text in mixed tool results while preserving all source images.
- Added Google Antigravity daily quota usage to the status line.
- Fixed status-line background-work counts so queued tasks and evaluation jobs remain visible without double-counting running subagents.
- Fixed nested subagent visibility in RPC subscriptions, the subagent HUD, get_subagents, subagent_* events, and get_subagent_messages.
- Fixed omp token refreshing local MCP OAuth credentials without blocking or losing rotating refresh tokens, and preserved OpenCode MCP OAuth configuration during discovery.
- Prevented process crashes caused by socket-closed errors and unhandled promise rejections during concurrent subagent shutdowns, timeouts, and MCP transport disconnects.
- Fixed automatic startup model selection so ambient AWS credentials do not incorrectly select an unavailable Amazon Bedrock model over a provider the user has authenticated with.
- Kept embedded context usage visible in the status line when long session names or paths consume available space.
- Added a status message when CTRL-O toggles tool-output expansion.
- Fixed omp usage to report Codex Chat and Spark capacity meters separately when they share a usage window.
@oh-my-pi/pi-mnemopi
Fixed
- Fixed working-memory search returning too few or no results when the most relevant matches had been superseded, ensuring valid older entries are still returned.
@oh-my-pi/pi-natives
Breaking Changes
- Replaced the Git-specific watchHead and headWatchTarget API with the backend-neutral watch and VcsRepo.watchTarget APIs.
Added
- Added portable repository discovery and read operations through VcsRepo, r
18.0.8 27 d’agost del 2026 — 64 bits Portable
18.0.7 27 d’agost del 2026 — 64 bits Portable
18.0.6 26 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added the backgroundIdleMs option to customize how long background auth-broker activity remains active before automatically parking.
Fixed
- Fixed auth-broker background activity keeping processes alive unnecessarily, so unused broker-backed auth storage now parks automatically and no longer prevents CLI exit.
@oh-my-pi/pi-coding-agent
Added
- Added fast, cached conventional commit message generation to the git TUI and omp commit --legacy, including automatic handling of whitespace-only changes, clearer commit scopes, and improved grammar and tense in generated summaries.
- The git TUI sidebar now supports collapsing and expanding the Unstaged and Staged sections, with keyboard shortcuts to stage or unstage an entire section.
- Long streaming thinking and reasoning output now continues into terminal scrollback during a turn instead of remaining clipped to the viewport.
Changed
- omp commit --legacy now uses the same conventional commit message generation as the git TUI.
- The git TUI sidebar now groups new files separately from tracked changes in the Unstaged section, while Staged and commit file lists use a unified status-based view.
- Improved resilience when streaming output changes during rendering, preventing incomplete blocks from causing further display updates to fail.
Fixed
- Commit-message generation errors in the git TUI now remain visible in the status bar instead of disappearing and returning to an idle state.
- Fixed omp update leaving standalone Windows binaries on the old version when stale Bun launcher metadata was present, and preserved launchers installed by a newer concurrent update during binary repair (#9806).
- Quitting omp git during commit-message generation now exits cleanly without leaving the process running.
@oh-my-pi/pi-natives
Fixed
- Improved TypeScript and TSX syntax highlighting, including correct handling of type annotations and template literals.
@oh-my-pi/pi-tui
Added
- Added Markdown.getLastRenderStableText() to expose the stable prefix of streamed Markdown text for append-only transcript publication.
@oh-my-pi/pi-utils
Added
- Added conventional commit generation with support for dependency, security, configuration, UX, and infrastructure commit types, plus configurable caching and large-diff analysis behavior.
What's Changed
- fix(cli): repair stale Windows update launchers by @roboomp in https://github.com/can1357/oh-my-pi/pull/9807
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v18.0.5...v18.0.6
18.0.5 25 d’agost del 2026 — 64 bits Portable
18.0.4 24 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- Improved performance in append-only context mode by memoizing message serialization, keeping per-call sync overhead flat as conversations grow.
Fixed
- Fixed an issue where onTurnEnd was skipped for turns ended by a terminal tool result (such as a subagent's final yield).
@oh-my-pi/pi-ai
Fixed
- Fixed Cursor tool calls through OpenAI-compatible authentication gateways losing arguments when complete argument maps are sent without streaming deltas (#9479).
- Fixed Cursor plan entitlement refusals repeatedly selecting ineligible accounts by scoping credential blocks to the requested model during rotation (#9488).
- Improved HTTP 413 error classification to accurately distinguish between payload/media size limits and token context window overflows, preventing inappropriate token compaction attempts and routing to correct recovery/fallback strategies (#9235).
- Fixed Cursor conversation rotation after aborts or mid-turn restarts to properly replay the last user message on a fresh conversation.
@oh-my-pi/pi-catalog
Fixed
- Fixed default reasoning effort for cursor/cursor-grok-4.5 and cursor/cursor-grok-4.6 so requests without an explicit effort setting default to -medium instead of -low, preventing rate limit rejections on Cursor's Start plan (#9478).
- Fixed aliased OpenCode Zen Ox Alpha models exposing incorrect effort levels, ensuring the gateway's native low, high, and max tiers are correctly mapped and reachable (#9349).
- Fixed missing rate card tier for public xAI and SuperGrok models with prompt lengths exceeding 200K tokens (#9512).
@oh-my-pi/pi-coding-agent
Added
- Added the omp git command (and /git slash command): an interactive, fullscreen repository TUI featuring a split/inline/hunk diff viewer with minimap scrollbar, syntax highlighting, a staging sidebar with line-level staging, commit composer with amend support, and author avatars. Supports keyboard navigation, full mouse interaction, and pinning views to specific commits via omp git <revision>.
- Overhauled the /extensions Extension Control Center into a fullscreen alternate-screen dashboard with mouse support, tab navigation, unified inspector views across extension types, live MCP connection management, and expandable details (Ctrl+O).
- Added support for live syntax highlighting in streaming markdown code blocks.
- Added an immediately editable startup composer for interactive launches, preserving drafts typed while session initialization is in progress.
Changed
- Improved streaming markdown and thinking block rendering performance on long sessions by batching token updates and eliminating redundant re-processing.
- Optimized streaming edit verification and session restoration for large files and history-heavy sessions.
Fixed
- Fixed invalid streamed edit patches occasionally reaching the edit tool instead of being stopped early.
- Fixed ! shell commands on zsh/fish by running them inside a real PTY, resolving terminal option errors and preserving ANSI color formatting.
- Fixed transcript layout corruption and viewport compression caused by interrupted streams, empty blocks, or collapsed wrapped diff lines.
- Fixed transcript scrollback loss where output below sticky cards (such as hub-wait or todo) failed to commit to terminal history.
- Improved HTTP 413 error handling: accurately distinguish between true token-context overflows and provider byte/media budget limits, persist terminal errors across sessions, and enable proper fallback-chain model switching.
- Fixed discovery-backed session models failing to restore when resuming sessions with omp --resume or --continue.
- Fixed browser tool initial launch timeouts on slow or cold host environments.
- Fixed eval runtime probes hanging on Windows due to inherited stdin handles.
- Fixed Claude models replaying partial thinking blocks as conversation text when interrupted mid-turn.
- Fixed image request failures with Kimi Code and Moonshot models by ensuring inline base64 image de
18.0.3 23 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed a Fireworks-hosted model aborting mid-generation with an HTTP 400 Floating point NaN (not-a-number) is detected in generation killing the turn instead of retrying; this model-side numerical fault is now classified transient and retried, matching the existing treatment of Copilot fleet-skew 400s (#9458).
- Fixed OpenAI-compatible completions hosts that stream content then terminate with the [DONE] sentinel while omitting (or nulling) finish_reason failing every turn with OpenAI completions stream closed before a finish_reason was received; a [DONE]-terminated stream now finalizes as a clean stop and only a genuine transport EOF (no [DONE], no finish reason) surfaces the incomplete-stream error (#9433).
@oh-my-pi/pi-catalog
Fixed
- Fixed OpenRouter auxiliary requests (e.g. session-title generation) failing with 400 Reasoning is mandatory for this endpoint and cannot be disabled on mandatory-reasoning models such as stealth/ox-alpha. Live discovery now honors the endpoint's reasoning.mandatory flag, clamping thinking-off to the lowest supported effort instead of sending reasoning: { enabled: false } (#9415).
@oh-my-pi/pi-coding-agent
Added
- Added opt-in edit auto-repair (edit.autoRepair.enabled): when an edit breaks a file's AST parse, the smol model repairs the broken region in place — validated by re-parse, revert-rejected, and surfaced as a diff in the tool result — instead of only warning.
- Added update channels: omp update --canary installs canary prereleases from the npm canary dist-tag and omp update --stable switches back; the chosen channel persists and drives the startup update check.
Changed
- Unexpected Stops now offers None, Mechanical (default), and Smart modes; Smart adds small-model classification to recover text-only stops.
Fixed
- Resolved cursor drift and text duplication caused by overlapping or out-of-bounds spelling ranges
- Squeezed transcript tool rows no longer render as a bare unstyled ╭─ Hub frame: a squeezed block keeps its real render whenever it fits the allocated rows, and blocks that genuinely overflow fold to a themed frame that names the tool's activity (e.g. Hub · send → Main).
- Python/Ruby/Julia eval cells that hit their wall-clock timeout during a parallel()/agent()/tool.* fan-out no longer get their kernel force-killed (losing all session state): the timeout now aborts in-flight bridge calls so the runner unwinds as a clean KeyboardInterrupt and the kernel survives.
- Multi-select ask options whose labels end in (Recommended) now show their checked state and avoid duplicate recommendation suffixes (#9452).
- Fixed crash during update output when theme configuration is missing
- Fixed flickering typo undercurls while typing by projecting state during revalidation
- Fixed self-update on Windows leaving the omp command missing or stuck on the previous version when package-manager reinstalls fail on running files
- Ctrl+T now toggles every thinking block in the transcript, including blocks already retired to terminal history (#9440).
- Copilot Grok 4.6 Responses streams that repeatedly close after thinking now stop after one same-model retry instead of consuming the full retry budget (#9427).
- /mcp test now reports cancellation immediately when Esc is pressed during a slow config lookup, instead of staying suspended until the read settles (#9419).
- Fixed remote browser relay endpoints advertising a client-local CDP WebSocket URL: /json/version now reflects a valid request Host and falls back to the relay's loopback address when it is absent or unusable.
- Restored red/green and syntax highlighting in edit-tool result bodies (#9439).
- Fixed goal mode failing to start (No such tool: xd://goal) when goal.enabled was turned on after the session had already started; the goal tool is now registered lazily on goal-mode entry (#9444).
@oh-my-pi/pi-natives
Fixed
- macOSCheckSpelling no longer reports the whole checked string as misspelled: automatic language identificat
18.0.1 23 d’agost del 2026 — 64 bits Portable
18.0.0 22 d’agost del 2026 — 64 bits Portable
17.4.2 21 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Image content blocks accept an optional url mirror: providers whose APIs fetch remote images (Anthropic url sources, OpenAI/xAI Responses and Chat Completions image_url, Google fileData) send the URL instead of the inline base64 payload.
Fixed
- Fixed Cursor thinking-effort selection being cosmetic: collapsed effort-routed families (GPT-5.6 Luna/Sol/Terra, Grok 4.5/4.6) now send the effort-routed wire model id instead of always pinning the -none off tier (#9246).
- Fixed OAuth preflight refresh stranding a peer-rotated credential: when a concurrent process rotated a rotating-refresh-token grant (e.g. Anthropic) during preflight, the resolve pass skipped the freshly reloaded row and failed the request with no credentials for single-account setups (#9194).
- Fixed Cursor reasoning-sibling models (e.g. gpt-5.4-mini-low, gpt-5.6-sol-xhigh) failing with resource_exhausted (errorId 528384): the per-effort GPT slug is now split into its base model id plus a { id: "reasoning", value: <effort> } request parameter, matching the official cursor-agent wire shape, instead of sending the sibling slug as the wire model id with no parameters (#9164).
@oh-my-pi/pi-catalog
Added
- Added identification of provider-side image fetchers (OpenAI, Anthropic, xAI, Google), so a server hosting images by URL can attribute an inbound fetch to the vendor that issued it.
Fixed
- Fixed Cursor model discovery showing separate picker rows for pure effort-suffixed models beyond GPT-5.6 by collapsing each standard and Fast lane into one reasoning-effort model (#9237).
@oh-my-pi/pi-coding-agent
Added
- Added an opt-in image URL broker (images.urls.enabled) that publishes outgoing images through an ordered chain of backends instead of sending inline base64 to URL-fetching providers
- Composer attachment chips (ported from omp2): pasted images and large text pastes stage as rounded preview cards above the prompt — image cards show a live thumbnail (Kitty Unicode placeholders) with pixel dimensions, text cards a snippet with +N lines/N chars — while the editor buffer holds a compact <icon> #N token in the card's identity color.
Changed
- Pasted images now insert only the [Image #N, WxH] marker; the redundant trailing attachment://N URI is no longer added to the composer.
- Added a consolidated CLI reference (docs/cli-reference.md) documenting every top-level subcommand and launch flag, including headless print mode (--print/-p, --print-thoughts) (#9252)
Fixed
- Fixed unreadable colors in macOS Terminal.app by using its supported 256-color mode (#9162).
- Fixed Esc after a fast /mcp test result aborting the active agent turn instead of consuming the advertised cancellation input (#9173).
- Fixed task spawns crashing when legacy boolean per-agent prewalk or advisor overrides are present in config.yml.
- Fixed ACP session/prompt requests hanging forever when a builtin slash command's residual prompt (e.g. /force:<tool> /some-command) resolved locally, which also wedged all subsequent prompts on the session (#9206).
- Fixed eval-spawned subagent output being omitted from per-turn output-token budgets, including failed and isolated runs (#9187).
- Fixed /compact over RPC blocking the serialized command queue for the full summarization round-trip, so a follow-up abort could not interrupt it (#9200).
- Fixed RPC UI select requests dropping option descriptions, allowing hosts to render described choices (#9175).
- Fixed /todo edit failing with "Could not parse Markdown" when checklist items had backslash-escaped brackets (- \[x\]), which editors and markdown renderers commonly emit (#9188).
- Fixed omp setup --check/--json with no component printing usage text to stdout and exiting 0; it now errors on stderr and exits non-zero so scripted JSON health checks fail loudly (#9221).
- Fixed an aggressive task.maxRuntimeMs mislabeling committed subagent outcomes: a budget-killed run is no longer reported as a runtime-limit timeout, and a subagent that yie
17.4.1 21 d’agost del 2026 — 64 bits Portable
17.4.0 20 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Breaking Changes
- Replaced global token counting functions (countTokens, countTokensConservatively, setTokenizerModel, and estimateTokens) with model-scoped, immutable Tokenizer instances (agent.tokenizer). Use tokenizer.countTokens(text, mode?), tokenizer.countMessage(message), or tokenizer.countMessages(messages).
- Updated context management functions (findCutPoint, prepareBranchEntries, collectShakeRegions, pruneToolOutputs, pruneSupersededToolResults, and trimRemoteCompactionInputToContextWindow) to require an explicit Tokenizer instance.
Added
- Added Tokenizer.checkTokenBudget(text, budget) to efficiently verify if text fits within a token limit using fast byte-bound checks before falling back to full tokenization.
- Added provider-anchored transcript token estimation (findTranscriptUsageAnchor, isTranscriptUsageAnchor, estimateTranscriptTokens) to calculate transcript token counts incrementally from the latest reported assistant turn usage.
- Added remotePreserveReusable() to check whether a previous remote compaction payload remains reusable with the active model.
Changed
- Expanded native tokenizer support across catalog models, adding exact embedded token counting for Claude, Qwen 3.5+, DeepSeek V3/V4/R1, Kimi K2/K3, and GLM-5+ models. Tokenizer now constructs from a resolved catalog Model.
- createCompactionSummaryMessage takes an options object after (summary, tokensBefore, timestamp); CompactionSummaryMessage gained optional method and tokensAfter display metadata.
@oh-my-pi/pi-ai
Added
- Added model metadata fields (context_length, max_output_tokens, input_modalities, etc.) to auth gateway model listing responses
Fixed
- Fixed tool-argument repair applying lossy transformations (such as stringifying objects or stripping unrecognized keys) when validating union schemas (anyOf/oneOf), preventing corrupted tool call and subagent payloads
- Fixed 400 errors when communicating with local OpenAI-compatible inference servers that reject chat_template_kwargs.reasoning_effort by improving reasoning effort parameter fallback and compatibility handling
- Fixed DeepSeek-family models on hosts like Fireworks losing reasoning whenever tools were offered: a redundant tool_choice: "auto" is now omitted so the provider keeps thinking enabled; forced and "none" selectors still take priority (#1207)
@oh-my-pi/pi-catalog
Added
- Models now include an optional tokenizer family field across bundled, discovered, and custom models (supporting Claude, Qwen, DeepSeek, Kimi, and GLM families), with support for explicit overrides in model configuration.
- Added long-context cost tiers (cost.longContext) to subscription Codex GPT-5.6 models (Sol, Terra, Luna) matching first-party API pricing above 272K input tokens.
Changed
- Bundled model metadata is prebuilt during generation, reducing catalog startup work.
Fixed
- Fixed tool-call turn failures for opencode-go/muse-spark-1.2 and related variants by ensuring API transport pins apply to live discovery and automatically inferring response routes for gateway-first OpenCode models (#8957).
@oh-my-pi/pi-coding-agent
Added
- /cleanse (and omp cleanse) — run the checker/repair loop in-session, with a live status board of running checkers, repair subagents, and token/cost totals.
- omp ps — interactive monitor for daemon-supervised background processes.
- Composer layouts — composer.shape picks the editor frame (rounded box, Claude Code rules, upstream-pi rules, borderless), with live previews in /settings and the setup wizard.
- Context line — statusLine.contextLine gauge (percentage, annotated, embedded) showing context usage and compaction boundaries.
- Backgroundable Python — eval cells can run async and auto-background like bash, with configurable thresholds.
- Local Claude token counting — Anthropic-family tokens now count via a native local tokenizer, and every counter (session maintenance, advisor, stats, context tools) uses the active model's own tokenizer.
- exte
17.3.8 19 d’agost del 2026 — 64 bits Portable
17.3.7 18 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Send the omp/<version> User-Agent on xAI chat (xai and xai-oauth) unless the request already set its own.
@oh-my-pi/pi-catalog
Changed
- Changed the paid xAI (XAI_API_KEY) and SuperGrok (xai-oauth) default models to grok-4.6.
Fixed
- Raised the GPT-5.6 Sol/Terra/Luna context window on the Codex transport (openai-codex) from 372K to 1M tokens: OpenAI enabled the 1M window for subscription Codex on 2026-08-16, but the Codex model registry still reports the stale 272,000, so discovery now floors these SKUs at 1,000,000 instead of trusting the reported value (openai/codex#38917).
@oh-my-pi/pi-coding-agent
Added
- Added ExtensionAPI.registerFileWriteFallback(handler) and ExtensionAPI.registerFileDeleteFallback(handler), letting an extension supply a fallback writer or deleter that is consulted when a native write, edit, or apply_patch byte-write or unlink is denied with a permission error (EPERM/EACCES/EROFS) — for hosts that embed the agent inside a sandbox that denies direct filesystem access but exposes a privileged channel. The brokered path is symlink-resolved so a handler's allowlist sees the real destination, a destination that cannot be resolved is not brokered at all, and req.sessionId names the session that issued the mutation so a handler sharing the process-wide registry can enforce policy per session. See docs/extensions.md.
Changed
- Send the omp/<version> User-Agent on xAI chat (xai and xai-oauth) unless the request already set its own (#8840 by @Jaaneek).
- Updated the default model for XAI_API_KEY (xai) and SuperGrok OAuth (xai-oauth) to grok-4.6. Automatic model selection continues to prefer paid xai/grok-4.6 when only XAI_API_KEY is set, with xai-oauth/grok-4.6 still available explicitly.
Fixed
- Fixed omp stats and /stats dashboards being unreachable from container hosts by accepting an explicit --host bind address while preserving the 127.0.0.1 default.
@oh-my-pi/omp-stats
Fixed
- Fixed the stats dashboard being unreachable from container hosts by accepting an explicit --host bind address while preserving loopback-only binding and same-origin API access by default.
What's Changed
- fix(ai): send omp User-Agent on xAI chat only by @Jaaneek in https://github.com/can1357/oh-my-pi/pull/8840
New Contributors
- @Jaaneek made their first contribution in https://github.com/can1357/oh-my-pi/pull/8840
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.3.6...v17.3.7
17.3.5 16 d’agost del 2026 — 64 bits Portable
17.3.4 14 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed Codex-compatible V2 remote compaction with an explicit v2Endpoint by sending the required feature-negotiation header (#8524).
@oh-my-pi/pi-ai
Fixed
- Fixed omp usage invalidate to discard stale OAuth and API-key usage snapshots, then force a cache-bypassing, per-provider serialized refresh with a broker request budget sized for the full unfiltered account batch, so upgraded subscriptions do not silently retain pre-change quota data.
- Fixed quota reporting and Cookie capture guidance for China (Beijing) Alibaba Token Plan credentials (#8509).
@oh-my-pi/pi-catalog
Added
- Added wire constants for Codex V2 remote-compaction feature negotiation.
Fixed
- Fixed raw COPILOT_GITHUB_TOKEN credentials skipping plan-specific endpoint discovery, which routed GitHub Copilot Business model requests to the personal endpoint and returned HTTP 403. The GitHub Copilot model cache is now scoped per credential, so switching the token no longer serves another account's stale endpoint for the cache TTL (#8507).
- Fixed the OpenRouter deepseek/deepseek-v4-pro-0813 route silently clamping the reasoning effort to high: the dated SKU advertises (and accepts) the wire-exact low/high/max ladder, so its effort override no longer collapses to high-only. The undated deepseek/deepseek-v4-pro OpenRouter route stays high-only. (#8517)
@oh-my-pi/pi-coding-agent
Changed
- Replaced the MuPDF-WASM PDF document backend with pdf-inspector through @oh-my-pi/pi-natives, preserving cached text conversion and PDF line selectors while reporting pages that need OCR.
- Restored read <pdf>: and read <pdf>:<image>.png page rendering by automatically capturing PDF pages through the headless Chromium browser tool.
Fixed
- Fixed Streamable HTTP MCP sessions being invalidated by opening the optional GET SSE stream before sending notifications/initialized, which prevented Figma Dev Mode MCP from connecting (#8514).
- Fixed the /hotkeys table describing Ctrl+D (app.exit) as "Exit (when editor is empty)" when it actually exits unconditionally and saves the current prompt as a resumable draft (#8530).
- Fixed Ctrl+G external editors failing to launch on Windows because Bun re-quoted the embedded cmd.exe /c command line (#8544).
@oh-my-pi/pi-mnemopi
Fixed
- Fixed recall() silently dropping scope='global' rows whenever a channelId filter was active: buildWhere() appended a redundant hard channel_id = ? clause on top of the (session_id = ? OR scope = 'global' OR channel_id = ?) visibility clause, so global rows whose channel_id didn't match (e.g. imported rows with channel_id NULL) were excluded. Channel isolation is preserved by the visibility clause alone. This made imported/global episodic memory permanently unrecallable through callers that always pass a channel (such as the coding-agent memory backend). (#8525)
@oh-my-pi/pi-natives
Added
- Added the async pdfToMarkdown native API backed by pdf-inspector, with page numbering, page-count, OCR-needed-page, and encoding-issue metadata.
Changed
- Docker images (Dockerfile, scripts/install-tests/*.dockerfile) build the native addon through the cargo/napi-rs backend (OMP_NATIVE_BUILD_BACKEND=cargo) instead of Bazel: a single fixed host target gains nothing from hermetic cross toolchains, and none of those images shipped bazelisk. OMP_NATIVE_CARGO_PROFILE picks the profile for that path (images use ci, local default stays local).
Fixed
- Fixed the root Cargo workspace failing to load when a stale directory exists under crates/ — e.g. a deleted crate whose directory survived git reset --hard. members no longer globs crates/pi-*, so a directory without a Cargo.toml can no longer break every cargo and Bazel build.
- Fixed Docker build contexts shipping nested build output: .dockerignore patterns are anchored at the context root, so bare target/ and dist/ matched neither go-port/*/target (~1.4 GB) nor packages/*/dist (~600 MB).
- Fixed deviceCheckGenerateToken aborting the whole process with SIGTRAP when
17.3.3 14 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Distinguished Gemini thought-only STOP responses from empty transports, avoiding repeated identical reasoning requests and duplicate Antigravity endpoint streams while surfacing the missing final output for session-level recovery.
@oh-my-pi/pi-coding-agent
Fixed
- Automatically continued Gemini turns that stopped after thinking without final output, using a bounded final-answer reminder instead of exhausting generic retries.
- Retried Gemini MALFORMED_FUNCTION_CALL failures when every emitted tool call was proven unexecuted, while preserving real tool-result and visible-output replay guards.
- Kept current terminal retry errors in one pinned banner with attempt context while surfacing local continuation failures instead of stale provider errors.
@oh-my-pi/hashline
Fixed
- Recovered dangling range separators in hunk headers (PUT 244.=:, CUT 5.=) as single-line ranges (N.=N) instead of rejecting the header as an orphan payload line.
@oh-my-pi/pi-tui
Fixed
- Fixed Gemini reports rendering their final headings and tables as one raw code block when the model emitted a lone closing Markdown fence without its opener.
What's Changed
- fix(gemini): recover malformed, empty, and broken Markdown turns by @usr-bin-roygbiv in https://github.com/can1357/oh-my-pi/pull/8493
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.3.2...v17.3.3
17.3.2 13 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Dropped unsigned thinking blocks from Antigravity Claude requests instead of sending them without a signature, preventing HTTP 400 responses when resuming sessions or switching models.
- Classified Antigravity HTTP 429 responses from structured google.rpc.ErrorInfo reasons (QUOTA_EXHAUSTED, RATE_LIMIT_EXCEEDED, and INSUFFICIENT_G1_CREDITS_BALANCE), using retry delays of five minutes or longer to distinguish rotatable quota windows from transient throttling instead of relying only on message regexes.
Removed
- Removed the Antigravity identity-prompt injection (ANTIGRAVITY_SYSTEM_INSTRUCTION and shouldInjectAntigravitySystemInstruction): Cloud Code Assist accepts arbitrary system instructions on gemini-3.x and Claude routes (verified live), and the injected stub never matched the real client's system prompt anyway. User system prompts are now sent unmodified (still tagged role: "user").
- Fixed Antigravity auto mode not failing over to the sandbox endpoint when the daily endpoint returned a thinking-only STOP, which caused Advisor turns to be falsely recorded as empty-response failures (#8480).
@oh-my-pi/pi-catalog
Added
- Added support for the deepseek-v4-pro:preview model
- Added support for the gemini-3.7-flash model
- Added dynamic Antigravity client-version discovery from the official update manifest (darwin/arm64 channel), so version-gated models appear without a code change; PI_AI_ANTIGRAVITY_VERSION remains available as an override.
Fixed
- Fixed Antigravity discovery missing Gemini 3.7 Flash: Cloud Code Assist gates newer models on the client version in the User-Agent, and the pinned antigravity/hub/2.1.4 was too old. The user-agent now matches the captured 2.8.0 client format (antigravity/hub/2.8.0 (aidev_client; os_type=darwin; arch=arm64; cl=963137146)); os_type/arch stay pinned to the darwin/arm64 reference client. Overridable via PI_AI_ANTIGRAVITY_VERSION / PI_AI_ANTIGRAVITY_CL / PI_AI_ANTIGRAVITY_OS / PI_AI_ANTIGRAVITY_ARCH.
Removed
- Removed ANTIGRAVITY_SYSTEM_INSTRUCTION from wire/gemini-headers; the Antigravity transport and web search no longer inject a fake identity prompt.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed the parent TUI stalling after a subagent submits its result until terminal focus or resize wakes the event loop (#8462).
- Fixed omp update misclassifying foreign npm/bun bin aliases while preserving package-manager ownership for globally linked checkouts (#8468).
- Fixed read hashline headers collapsing nested in-workspace paths to the bare basename, which let a same-basename file at the session cwd capture a verbatim follow-up edit and deterministically reject it with hash is not from this session. Headers now retain the workspace-relative path (e.g. [src/settings.json#0063]) (#8482).
@oh-my-pi/pi-utils
Fixed
- Fixed fetchWithRetry() aborts during retry backoff to preserve the documented "Request was aborted" error contract (#8450).
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.3.1...v17.3.2
17.3.1 13 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Added
- Added dynamic Antigravity and Gemini CLI discovery support for Gemini 3.7 Flash, with low/medium/high thinking-level routing.
Changed
- Updated model metadata, context windows, pricing, and configurations in the catalog
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Claude Code user discovery ignoring CLAUDE_CONFIG_DIR for configuration, plugins, MCP servers, and imported sessions.
- Fixed the status-line git branch display freezing after switching branches.
- Fixed Pi extension contexts omitting the runtime mode, which caused TUI guards to silently disable extension UI.
- Fixed extension-registered tool names being rejected by the --tools flag before extension discovery, which prevented least-privilege sessions from allowlisting plugin tools.
- Fixed omp plugin install failing with cloning errors for legacy Pi extensions whose tool schemas use legacy-typebox builders.
- Fixed omp update aborting with chmod ENOENT when concurrent update runs overlapped by using unique download temporary paths.
- Fixed the browser tool executable probe launching the user's installed GUI Chromium on Windows: the --version version probe from ecb22957 was Linux-scoped but ran for every platform candidate, so on Windows it could hand off to a running chrome.exe, open a normal browser window, then reject the candidate and fall back to cached Chrome for Testing. The probe is now confined to Linux (#8445).
@oh-my-pi/pi-natives
Fixed
- Fixed omp failing to start on a clean Windows install with Failed to load pi_natives native addon for win32-x64 ... The specified module could not be found (LoadLibrary error 126). The shipped win32-x64 addon linked the dynamic MSVC CRT (/MD) and imported VCRUNTIME140.dll from the Visual C++ Redistributable, which is absent on a fresh Windows install. The addon now statically links the CRT (+crt-static for rustc plus the static_link_msvcrt cc feature for its C dependencies), so the .node imports only core Windows system DLLs (#8439).
@oh-my-pi/omptype
Fixed
- Fixed TypeBox adapter omitting pattern, non-URL format, and multipleOf constraints from the emitted JSON Schema.
@oh-my-pi/pi-tui
Fixed
- Fixed screen flashing in Herdr panes during transcript streaming.
What's Changed
- fix(coding-agent): watch git dir so branch display tracks switches by @roboomp in https://github.com/can1357/oh-my-pi/pull/8415
- fix(extensions): populate runtime mode in context by @roboomp in https://github.com/can1357/oh-my-pi/pull/8423
- fix(cli): allow extension tools in --tools allowlists by @roboomp in https://github.com/can1357/oh-my-pi/pull/8424
- fix(extensions): lower embedded omptype schemas in Type.Unsafe by @roboomp in https://github.com/can1357/oh-my-pi/pull/8425
- fix(tui): stop direct Herdr pane flicker by @roboomp in https://github.com/can1357/oh-my-pi/pull/8433
- fix(update): unique temp path for concurrent self-updates by @roboomp in https://github.com/can1357/oh-my-pi/pull/8435
- fix(discovery): honor Claude config directory by @roboomp in https://github.com/can1357/oh-my-pi/pull/8440
- fix(natives): static-link win32 MSVC CRT so addon needs no VC++ redist by @roboomp in https://github.com/can1357/oh-my-pi/pull/8443
- fix(coding-agent): confine browser executable probe to Linux by @roboomp in https://github.com/can1357/oh-my-pi/pull/8446
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.3.0...v17.3.1
17.3.0 13 d’agost del 2026 — 64 bits Portable
17.2.15 12 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed an issue where AWS_BEDROCK_SKIP_AUTH failed to expose Amazon Bedrock models when AWS credential files were unavailable.
- Fixed an issue where forceReasoningOff was ignored by Anthropic and Google transports, which allowed native thinking alongside a caller-supplied external scratchpad.
@oh-my-pi/pi-catalog
Fixed
- Fixed classification of model IDs with large minor versions (e.g., claude-opus-5-11) or three-part versions, ensuring they no longer fall back to stale default configurations.
@oh-my-pi/pi-coding-agent
Added
- Added --external-thinking CLI flag to force external thinking tool activation.
- Added omp compress command, which uses an isolated, two-tool agent loop to rewrite single or multiple text files (supporting glob patterns and concurrent processing) into dense prompt registers.
- Expanded tool discovery in omp cleanse to support staticcheck and golangci-lint (Go); mypy, pylint, flake8, ty, and basedpyright (Python); oxlint, deno lint, stylelint, and vue-tsc (JS/TS); and actionlint (GitHub Workflows).
- Added support for natural language requests in omp cleanse "<request>", which launches a discovery subagent to automatically inspect the project, determine the correct commands, and map outputs.
- Added an interactive picker to omp cleanse when run without arguments on a TTY, allowing users to run all checkers, select a specific checker, or describe what to fix.
Changed
- Restricted the think tool to GPT, Claude, and Gemini transports that support native reasoning replacement.
- Increased the default subagent cap for omp cleanse from 8 to 32.
Fixed
- Fixed a hang in headless omp -p runs when plan.defaultOnStartup: true is enabled by disabling the startup default in print mode.
- Fixed display.hideToolActivity failing to hide certain activity blocks, such as reminders, diagnostics, and completions.
- Fixed several issues in the MCP Streamable HTTP transport, including updating the negotiated protocol version to 2025-11-25, resolving connection drops and SSE resumption gaps, and preventing double-execution of tools during auth refreshes.
- Fixed /handoff losing local artifacts (plans, scratch files, research notes) by copying them across the handoff session boundary.
- Replaced libarchive-based tar parsing with a hardened, in-process tar reader to prevent crashes and safely handle complex archive structures, symlinks, and sparse metadata.
- Fixed Ctrl+O tool-output expansion failing to reach launch-completion messages wrapped in the hidden tool activity container.
@oh-my-pi/hashline
Added
- Added a post-apply parse advisory warning that alerts users when an applied edit fails to parse (despite the pre-edit content parsing successfully), helping catch balance-neutral misplacements that previously failed silently.
Fixed
- Fixed a bug where Rust lifetimes (e.g., 'static) were incorrectly parsed as starting a string literal, which blinded the delimiter-balance scanner and could lead to silent signature deletions. Single-quote lexing on .rs files is now language-aware and correctly distinguishes lifetimes from character literals.
- Fixed an issue where terminal newlines in files were incorrectly exposed as editable blank rows.
@oh-my-pi/snapcompact
Fixed
- Fixed Anthropic model ID parsing to be case-insensitive and extended the high-resolution 1932px frame tier to Claude Opus 5 and later, preventing sessions from falling back to lower-resolution 1568px frames and preserving full history per compaction.
@oh-my-pi/pi-utils
Changed
- Extended parsed Server-Sent Events (SSE) to include optional id and retry fields, enabling reconnecting transports to retain stream cursors and respect server-requested retry intervals.
What's Changed
- fix(snapcompact): extend high-res frame tier to opus 5+ by @roboomp in https://github.com/can1357/oh-my-pi/pull/8257
- fix(coding-agent): copy local artifacts across handoff session boundary by @roboomp in https://github.com/can1357/oh-my-pi/pull/8262
- fix(mcp): sen
17.2.14 11 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added forceReasoningOff and disableReasoning options to disable reasoning in OpenAI and Azure OpenAI models
@oh-my-pi/pi-coding-agent
Added
- Added externalThinking setting for private scratchpad reasoning via the new think tool
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.2.13...v17.2.14
17.2.13 11 d’agost del 2026 — 64 bits Portable
17.2.12 9 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed account-scoped Codex cyber-policy denials bypassing sibling credential rotation; replay-safe requests now try every configured account before surfacing the error.
@oh-my-pi/pi-catalog
Fixed
- Fixed dynamically discovered alibaba-token-plan/qwen3.8-max metadata so thinking controls and image input are available (#8019).
- Routed opencode-go/deepseek-v4-flash through the OpenAI responses API — the OpenCode Go gateway does not serve this model at /zen/go/v1/chat/completions, only at /zen/go/v1/responses (deepseek-v4-pro keeps chat completions).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed shell minimization replacing meaningful rustc --print output with OK.
- Fixed shell minimization altering outputs shorter than 1,000 characters; these now pass through unchanged.
- Fixed primary and advisor Codex sessions falling back to another provider before trying sibling accounts when an account lacks Trusted Access for Cyber approval.
- Fixed task subagent assistant turns being omitted from the per-model TPS/TTFT aggregates shown by /models. (#8022)
- Fixed terminal-title spinner writes consuming CPU during WSL/ConPTY agent waits by using the same static working separator as native Windows (#8012).
- Fixed long-running sessions leaking memory for every completed keep-alive task/scout subagent: a disposed (parked) subagent's AgentSession stayed pinned through the lifecycle adoption record's reviver closure, and dispose() never released the message array, append-only provider transcript, session-manager entries, or the raw-SSE debug buffer, so heavy transcripts and captured provider wire frames accumulated for the process lifetime (#8003).
- Fixed Z.AI web search dropping sources and exposing raw JSON when MCP responses double-encode content text (#8000).
- Fixed /handoff masking empty/whitespace-only generation and harness-initiated aborts as "Handoff cancelled"; manual empty generation now surfaces a logged failure, harness aborts preserve their reason (or report "Handoff aborted by session"), and auto-handoff still falls back to context-full compaction (#7993).
@oh-my-pi/hashline
Breaking Changes
- PUT N.=M @name over a span now throws when @name was never captured, instead of warning and deleting the range. Pasting a never-captured register over a span wrote nothing back, so a mistyped or hallucinated register name silently destroyed content. Gap pastes (PUT >N @name) keep the warned no-op behaviour from 17.2.11.
Added
- applyEdits now takes a path and uses the native tree-sitter parser to decide every boundary repair that depends on delimiter semantics. The authored edits are materialized first: if that result parses, it is returned untouched, so a } inside a regex literal, a string, or Markdown prose is never mistaken for a block closer. A closer-spare repair lands only when the repaired result is shown to parse — never on delimiter arithmetic alone — so an unrecognized language or an unprovable candidate leaves the edit exactly as authored. Wired through the patcher, recovery, section apply, and the edit tool's preview.
- Auto-repair for replacement ranges that start one line early on a structural closer (the } of the construct above): the closer is spared and the payload lands after it, gated on the same parse proof.
- Warning for balanced payloads over ranges that end mid-block (deleting opener(s) whose closer(s) survive below), pointing at the block-op remedy (PUT N*:). Raised only when the baseline parsed and the authored result does not, so it cannot fire on prose or an unknown language.
- Warning when a + body row is itself a valid hunk header (+CUT 5.=9). Such a row is literal content by definition and is inserted into the file as text; naming it at the moment it happens turns a silent source-file corruption into an actionable diagnostic.
Fixed
- Rejected patches whose pasted N:TEXT read-output rows repeat a source line number. Each such row is recovered as a single-line PUT N.=N:, so a body written as 
17.2.11 7 d’agost del 2026 — 64 bits Portable
17.2.10 6 d’agost del 2026 — 64 bits Portable
17.2.9 5 d’agost del 2026 — 64 bits Portable
17.2.8 4 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Changed
- Upgraded the bundled omptype schema engine: intersection and pipe operators, bigint and RegExp literals in the string DSL, Standard Schema V1 interop, JSON Schema import via fromJsonSchema(), and richer union/collection error reporting.
@oh-my-pi/omptype
Added
- Added io: 'input' and io: 'output' options to toJsonSchema(), supporting input validation shapes and piped .to() target types
- Added Standard Schema V1 interop: every schema exposes ~standard with synchronous validation, enabling direct use with @t3-oss/env, tRPC, and other Standard Schema consumers.
- Added fromJsonSchema(), rebuilding callable schemas from JSON Schema documents (draft-07 / draft-2020-12 structural keywords, string formats, $defs recursion, enums, and anyOf/oneOf/allOf composition) — the inverse of Type.toJsonSchema().
- Added $defs/$ref emission for recursive alias schemas in toJsonSchema() (draft-07 converts to definitions), preventing unbounded recursion on cyclic scopes.
- Added AnyType, a minimal structural constraint for generic functions accepting any schema without descending the recursive fluent surface.
- Root .default() values now materialize for undefined input in direct calls and at the Standard Schema boundary (factories run per call).
- .narrow()/.filter() boolean overloads accept OmpErrors returns, so cond || ctx.reject(...) recipes typecheck.
Changed
- Restored low-overhead schema construction by lazily activating advanced normalization and compatibility machinery.
- .default() is typed input-side (i | (() => i)) and marks the schema's input as optional (i | undefined).
- Parse keywords (string.integer.parse, parse.number, ...) now infer their morph output inside union strings, and input-side inference is union-aware.
- Object-literal inference for .merge()/.or()/.and() unwraps embedded schema values (output and input sides).
Fixed
- Alias intersections defer through memoized lazy nodes, so cyclic scope schemas no longer overflow the stack in .and() or morph-union determinism checks.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.2.7...v17.2.8
17.2.7 3 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Replaced arktype with @oh-my-pi/omptype for schema validation, delivering up to 100x faster schema construction and 60-100x faster validation while maintaining full compatibility with existing type/Type exports and the isArkSchema contract.
Fixed
- Fixed OpenAI-Codex (ChatGPT OAuth) requests failing with an Unsupported service_tier: auto error on default or legacy sessions by omitting the implicit auto service tier on the wire.
- Fixed an issue where Cursor kimi-k3 sessions would break permanently when a same-model assistant turn was persisted without thinking blocks, replacing hard errors with graceful warnings.
@oh-my-pi/pi-catalog
Fixed
- Fixed an issue where setting thinking-level: off failed to disable reasoning on direct DeepSeek V4 requests.
@oh-my-pi/pi-coding-agent
Changed
- Replaced arktype with @oh-my-pi/omptype for tool parameter and config schemas, significantly improving startup performance with ~100x faster schema construction. Config schema errors are now reported via OmpErrors using the same path/problem structure.
Fixed
- Fixed an issue where custom, extension, or hook tool wrappers stripped schema methods off parameters, causing wire-schema detection failures and status-line token estimator crashes.
- Fixed a bug where agent() calls in evaluation cells ignored turn cancellation and continued running indefinitely.
- Fixed the built-in tail command to exit silently with code 141 (SIGPIPE) instead of failing with a "Broken pipe" error when a downstream pipeline reader exits early.
- Fixed the in-process ps shell builtin to support common procps/BSD format specifiers, including tpgid, pri, flags, real/effective user/group columns, wchan, fault counters, sz, and the STAT + foreground flag.
- Fixed install.sh falsely reporting success on musl-based systems (such as Alpine Linux) when the binary fails to start; the installer now smoke-tests the binary, exits non-zero on failure, and provides remediation steps.
- Fixed Codex config.toml discovery incorrectly importing MCP servers that are configured with enabled = false.
- Fixed bash.patterns allow rules rejecting valid commands when quoted arguments contained shell metacharacters (such as Cargo benchmark regex filters).
@oh-my-pi/pi-natives
Added
- Added missing procps/BSD output format specifiers and aliases to the in-process ps shell builtin, including support for columns like tpgid, pri, flags, wchan, and various user/group/time fields.
- Updated ps -j to include the TPGID column, ps -l to display the single-character S column, and the STAT column to support the + foreground process group flag.
@oh-my-pi/omptype
Added
- Introduced omptype, an ArkType-compatible schema validation library featuring a lazy JIT runtime that compiles specialized validators on the third call for ultra-fast hot-path validation and low construction overhead.
- Added support for a rich string definition DSL (primitives, literals, unions, arrays, bounds, inline defaults, and optional keys), object definitions (including index signatures and strict key rejection/deletion), and comprehensive composition methods (.or, .and, .array, .pipe, .narrow, .describe, .default, .allows, .assert).
- Added TypeBox-style (@oh-my-pi/omptype/typebox) and Zod-style (@oh-my-pi/omptype/zod) authoring adapters that produce native omptype schemas.
- Added support for recursive named scopes, modules, runtime generics, fixed/optional/variadic tuples, Date literals/bounds, disjointness-aware intersections, separate input/output inference, and draft-2020-12 JSON Schema emission.
- Shipped transpiled ESM and TypeScript declarations in the npm package to support plain Node.js environments, while preserving TS source resolution for Bun consumers.
Changed
- Optimized the lazy JIT compiler to support tuples, refinements, morphs, intersections, instances, and recursive aliases, while reducing schema construction overhead.
Fixed
- Fixed a TypeScript compiler error (TS2589: "type instantiation is 
17.2.5 3 d’agost del 2026 — 64 bits Portable
17.2.4 1 d’agost del 2026 — 64 bits Portable
17.2.3 1 d’agost del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added the ai& (aiand) provider registry entry with API-key paste login validated against https://api.aiand.com/v1/models.
Fixed
- Fixed Anthropic OAuth (Claude Pro/Max subscription) requests hard-429ing (Usage credits are required for long context requests) on every beta-gated 1M model — e.g. claude-sonnet-4-6, which the default task/smol/scout subagent roles resolve to — regardless of prompt size, breaking all subagents. The 17.2.1 cowork request profile reintroduced the context-1m-2025-08-07 beta for any model with a 1M catalog window, but subscription credentials have no long-context credit balance so Anthropic rejects the request outright. The beta is no longer advertised on OAuth requests; subscription accounts transparently get the standard 200k window. (#7238)
- Fixed OpenAI Codex Responses ignoring disabled cache retention when deriving prompt_cache_key, while preserving transport session identity (#7219).
@oh-my-pi/pi-catalog
Added
- Added support for the ai& provider (aiand), an OpenAI-compatible inference API with dynamic model discovery (context windows, capabilities, reasoning efforts, and USD pricing from /v1/models) and API-key authentication via the AIAND_API_KEY environment variable.
@oh-my-pi/pi-coding-agent
Changed
- Tightened the system prompt notation: the legend now defines ⟺, ≠, ∉/∌, and operator binding order; replaced undefined symbols (⊭, ≢) in prompt bodies; removed delegation guidance duplicated between the eager-tasks preamble and the delegation gates.
Fixed
- Fixed headless browser launch storms and orphaned Chromium process trees: omp processes now attach to one project-shared Chromium owned by the daemon broker (tabs per session; Chrome dies with the last omp client in the project), concurrent browser opens in one process share a single launch, and concurrent daemon start requests for one name can no longer spawn duplicate untracked processes.
- Fixed Bash auto-background leaving a live Bun.sleep threshold timer scheduled after a command completes (or abort/steering wins) first, which could keep the event loop alive and delay SDK/headless shutdown until the threshold expired (#7235).
- Fixed ephemeral side turns and native compaction bypassing an explicit or fork-inherited prompt cache key (#7218).
- Fixed the live Ask dialog crashing the whole session with a replaceTabs TypeError when a question reached AskDialogComponent without a string question field; questions are now normalized at dialog entry, mirroring the transcript renderer (#7211).
- Fixed Codex web search collapsing backend errors to Codex error (): Unknown error; the SSE error parser now preserves the backend code and message from top-level, nested error, and response.error envelopes (#7200).
@oh-my-pi/pi-mnemopi
Fixed
- Stripped <think>…</think> reasoning blocks from remote LLM output in cleanOutput, so reasoning-model responses no longer leak into consolidated memories or corrupt fact extraction (the reasoning wrapper previously survived parsing and every stored fact became reasoning prose). (#7231)
What's Changed
- fix(coding-agent): preserve Codex web search backend error diagnostics by @roboomp in https://github.com/can1357/oh-my-pi/pull/7201
- fix(coding-agent): normalize ask dialog questions to prevent render crash by @roboomp in https://github.com/can1357/oh-my-pi/pull/7212
- fix(coding-agent): reuse effective prompt cache key by @roboomp in https://github.com/can1357/oh-my-pi/pull/7220
- fix(ai): honor disabled cache retention in Codex Responses by @roboomp in https://github.com/can1357/oh-my-pi/pull/7221
- fix(mnemopi): strip reasoning blocks from summaries by @roboomp in https://github.com/can1357/oh-my-pi/pull/7234
- fix(coding-agent): clear bash auto-background threshold timer by @roboomp in https://github.com/can1357/oh-my-pi/pull/7236
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.2.2...v17.2.3
17.2.2 31 de juliol del 2026 — 64 bits Portable
17.2.1 30 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added exact OAuth credential-row resolution by durable credential id. The targeted path refreshes only that row and never ranks, rotates, or falls back to sibling accounts.
Changed
- Anthropic OAuth requests now reproduce Cowork's current claude-desktop request profile, including client/runtime metadata, beta selection, system and billing attestation, the 64K output cap, and stable HTTP/1.1 header ordering.
@oh-my-pi/pi-catalog
Fixed
- Fixed Ollama model-manager caches being reused after the configured base URL changed by scoping cache namespaces to the normalized native discovery endpoint, including reverse-proxy path prefixes (#7087).
@oh-my-pi/pi-coding-agent
Added
- Added --from-claude and --from-codex session imports, also available from /resume @claude and /resume @codex.
- Added an opt-in OMP-native software-security workflow (security.enabled, default off) with immutable scan plans, exact-account Codex subscription affinity, native task-worker review, canonical findings/coverage/SARIF publication, project-scoped history, explicit dispositions, producer-differential comparison, and the read-only security:// resource namespace. Generic SARIF and official Codex Security bundles normalize into the same OMP-owned store.
- Added explicit Codex Security cloud operations to the opt-in security workflow: list and start account-pinned cloud scans, inspect their progress, and import current findings into OMP's canonical store and security:// namespace without changing the native scan engine or spoofing official runtime attribution.
Changed
- Reserved security:// from RPC host URI shadowing so vendor adapters cannot replace OMP's canonical security-analysis namespace.
Fixed
- Fixed remote or LAN local-engine endpoints being ignored during model discovery: the llama.cpp and Ollama probes used timeouts tuned for loopback, so a host reached over the network could exceed them and return no models, while changing OLLAMA_BASE_URL/OLLAMA_HOST could keep reusing a fresh cache from the previous endpoint. Non-loopback hosts now get a generous discovery timeout, and Ollama cache rows are scoped to the normalized endpoint (#7087).
- Fixed omp install failing extension validation for pi extensions that import createEditTool or createWriteTool (e.g. gentle-pi) — the legacy @oh-my-pi/pi-coding-agent shim exported the read/bash/grep/find/ls tool factories but omitted the edit and write ones, so a named import threw Bun's static "Export named X not found" error. Added createEditTool/createEditToolDefinition and createWriteTool/createWriteToolDefinition to match the upstream pi surface (#7094).
- Fixed Python eval's loopback tool bridge being routed through macOS system HTTP proxies, which caused parallel() tool reads to fail with ConnectionRefusedError after a local proxy stopped.
@oh-my-pi/pi-natives
Fixed
- Fixed the computer tool advertising Wayland support that never worked: on the default rootless XWayland (GNOME/KDE/sway) the X11 root window has no readable pixmap, so root GetImage failed on every screenshot with a raw BadMatch protocol dump. Monitor::all now probes root drawability at initialization and fails fast with an actionable DESKTOP_BACKEND_UNAVAILABLE message naming the rootless-XWayland constraint, and docs/computer-use.md now lists rootless XWayland as unsupported (#7085).
@oh-my-pi/pi-utils
Added
- Added a postmortem.quit configuration option to safely handle shutdown paths when the terminal output has already disconnected.
- Added project-keyed OMP security-state directory helpers under the user state root.
What's Changed
- feat(security): add OMP-native security scan subsystem by @kmccleary3301 in https://github.com/can1357/oh-my-pi/pull/7048
- fix(natives): fail fast on uncapturable rootless XWayland root by @roboomp in https://github.com/can1357/oh-my-pi/pull/7086
- fix(coding-agent): honor remote llama.cpp/ollama discovery base urls by @roboomp in https://github.com/can1357/oh-my-pi/pull/7090
- fix(c
17.2.0 30 de juliol del 2026 — 64 bits Portable
17.1.8 28 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed an HTTP 400 error when resuming or replaying OpenAI history after an interrupted native Computer Use turn.
- Fixed connection 404 errors when using Google Vertex AI in multi-region locations (eu and us) by correctly resolving regional endpoint (REP) hosts.
- Fixed a resource leak in SqliteAuthCredentialStore.close() where unclosed prepared statements kept the SQLite connection alive, preventing database file cleanup (especially on Windows where files remained locked).
@oh-my-pi/pi-catalog
Added
- Added resolveVertexEndpointHost(location) utility to resolve the correct Vertex AI API endpoint hostnames for global, multi-region, and regional locations.
Fixed
- Fixed an issue where calculateCost under-reported Anthropic cache-write costs by honoring the usage.cttl breakdown to correctly price 1-hour retention writes at 2x the base input rate.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Changed tab.screenshot() to no longer accept a per-call save path; it now saves screenshots under browser.screenshotDir (or the OS temp directory if unset) and returns the saved path.
Added
- Added omp cleanse, a new command that automatically detects language-ecosystem checkers, parses diagnostics (such as Cargo Clippy JSON), distributes repair workloads across concurrent subagents, and runs verification checks with a live progress bar.
Changed
- Reworked the /guided-goal command from a modal-based popup flow into a natural, conversational chat interface where the agent asks follow-up questions directly in the session.
- Reduced startup memory usage by lazy-loading HTML session export assets only on their first use.
Fixed
- Fixed Advisor notes appending stale-review-window warnings when newer primary turns are queued during a review.
- Fixed layout padding alignment issues in bordered output blocks and web-search result panels.
- Fixed excluded web search providers remaining visible in the Web Search Provider Order settings list.
- Fixed internal Hub peer messages being exposed as ordinary tool-call updates in clients like Paseo.
- Fixed compatibility issues when installing legacy pi extensions by updating the legacy shim to correctly bridge missing runtime symbols and exports (such as isContextOverflow, isRetryableAssistantError, and JSON parsing utilities).
- Fixed an issue where routine daemon operations (like list, logs, stop, or describe) could inadvertently trigger a restart loop for detached daemons in a backoff window.
- Fixed marketplace plugin MCP discovery to correctly honor the mcpServers manifest field in plugin configuration files.
- Fixed user-initiated shell executions (! and $) being misattributed as agent actions in advisor transcripts.
- Fixed unnecessary prompt-cache invalidations by preserving the active auto-thinking effort level when per-turn classification fails.
- Fixed the omp process name showing up as bun in Linux process managers (like ps and top).
- Fixed agent shell commands inheriting environment variables from the launch directory's .env file, ensuring they only receive the parent environment and explicit tool overrides.
- Fixed the /new command retaining completed or failed async jobs from the previous session.
- Improved error handling in omp update to display a friendly timeout message if the download times out while streaming the binary.
- Fixed the write tool incorrectly treating semicolon-joined read selectors as filesystem paths and creating unintended directory structures.
- Fixed omp worktree clear prematurely deleting active task-isolation sandboxes owned by running subagents.
- Fixed /vibe mode preventing the director from completing parent tasks after verifying worker results by keeping the built-in todo tool active.
- Fixed numeric GitHub issue and pull request autocomplete being suppressed inside skill slash-command arguments.
@oh-my-pi/pi-natives
Fixed
- Fixed an issue on macOS (darwin) where the native addon delivered zero AudioCapture callbacks, which prevented microph
17.1.7 28 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- beforeToolCall now runs during arg-prep in a pre-dispatch prepare phase — on the streamed path before the assistant message's message_start/message_end are emitted, and always ahead of concurrency resolution, tool_execution_start, telemetry span start, and tool.execute — instead of inside the already-scheduled execution slot. It receives the resolved tool in its context and may return args to replace the call's arguments; a replacement is revalidated against the tool schema, written back to the assistant message's tool-call block, and re-resolves argument-dependent interruptibility, making it the single source of truth for history, persistence, provider replay, scheduling, execution events, and tool.execute. Argument validation moved into the same prepare phase, so functional concurrency resolvers now see validated (and possibly revised) arguments rather than raw pre-validation ones. The hook now receives the run's request abort signal rather than the per-tool signal.
@oh-my-pi/pi-ai
Changed
- Upstream 403 Forbidden responses (e.g. Anthropic permission_error plan/model denials, Copilot model-policy rejections) now rotate through sibling credentials like usage limits do, instead of failing the session on the first denied account. The denied credential is soft-blocked for 60s and re-validated — never removed — and the original 403 surfaces only once every sibling has been tried.
- Usage report filtering in the auth-broker remote store is memoized per (reports, snapshot) with a precomputed per-provider OAuth credential map, replacing an O(reports × credentials) scan on every credential-selection and status refresh
- Cursor and Devin Connect-frame readers no longer copy every stream chunk through Buffer.concat when the pending buffer is empty
@oh-my-pi/pi-catalog
Added
- Added support for moonshotai/Kimi-K3 and kimi-k3-fast models
- Added umans-kimi-k3 prerelease model configuration
Changed
- Updated pricing and token limits for selected models
@oh-my-pi/pi-coding-agent
Added
- Added the bundled ts-no-local-is-record TTSR rule, which catches local isRecord function and lambda definitions and directs agents to shared guards plus explicit shape validation.
- A tool_call handler (extension or hook) can now return input to revise the arguments a tool executes with, not just block it. The returned object is the raw execution input passed to the tool (ignored when block is set, and not applied to computer tool calls), enabling wrappers that normalize or rewrite a built-in's arguments without reimplementing the tool. For model-issued calls the event fires at arg-prep time in the agent loop, so a revision is revalidated against the tool schema and is what concurrency scheduling, tool_execution_start/transcripts, the persisted assistant message, and the approval gate all observe — the user approves exactly what runs, and a revision that changes a tool's functional concurrency (e.g. bash pty) schedules correctly. A revised nested write xd:// device dispatch forfeits the outer write gate's approval and faces the full prompt again (#6681 by @psyrendust).
- Added a parser for macOS sample(1) call-tree reports to the read tool: *.sample.txt reads now return a compact bottleneck summary — per-thread hot paths with on-CPU sample counts (blocked syscall time excluded), demangled Rust v0/legacy symbols, flattened direct recursion, merged call-site siblings, idle-thread classification, and a process-wide top-functions-by-self-samples table. :raw still reads the original report, and files that merely carry the extension fall back to plain text.
- Added V8 .cpuprofile support to the read tool (Node/Bun --cpu-prof, Chrome DevTools, CDP Profiler.stop output): reads now return a compact bottleneck summary — hot-path call tree with on-CPU milliseconds ((idle) time excluded), collapsed pass-through chains, flattened direct recursion, shortened file URLs, and a top-functions-by-self-time table. :raw still reads the original
17.1.6 27 de juliol del 2026 — 64 bits Portable
17.1.5 27 de juliol del 2026 — 64 bits Portable
17.1.4 26 de juliol del 2026 — 64 bits Portable
17.1.3 24 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Cursor sessions exposing ast_edit (and other staged-preview xd:// devices) without a reachable resolver: the built-in write tool — which carries the xd://resolve / xd://reject transport that finalizes a staged preview — was filtered out of Cursor's forwarded catalog, so previews could never be resolved and the session aborted after three forced write turns. write is now re-included in the forwarded catalog whenever pi-agent devices are advertised (#6536).
- Fixed OpenAI Responses and chat-completions streams honoring per-model first-event watchdog policy, allowing local llama.cpp-style backends to process arbitrarily large prompts without a premature client cancellation (#6524).
@oh-my-pi/pi-catalog
Fixed
- Disabled the first-event watchdog for local OpenAI-compatible backends while retaining the 300-second inter-event watchdog, so long llama.cpp prompt prefill is not canceled and retried (#6524).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed the in-process find builtin's -exec/-execdir children inheriting the omp process's real stdout/stderr, so commands like find … -exec ls -ld {} \; spammed their output straight into the terminal (corrupting the TUI) instead of the tool's captured output, and bypassed shell redirects. Exec children now stream stdout/stderr back through the shell's scope streams (matching xargs/ifne) and run with the shell's exported environment (env_clear + scope snapshot) instead of the host process environment.
- Fixed ast_edit erroring with "lang is required" — an argument that no longer exists in the tool schema — when paths resolved to files of multiple languages (e.g. a crate directory with .rs + .toml). Mixed-language paths now rewrite per file: each file is parsed in its own inferred language, patterns are compiled per language, and a pattern that doesn't parse in some matched language simply skips those files.
- Fixed the retain tool's TUI renderer crashing when streaming arguments temporarily expose a non-array items value (#6528).
- Fixed Edit and Write tools failing with Settings not initialized in isolated sessions by using each tool session's settings for generated-file guards, with safe schema defaults for standalone guards and inline-image rendering (#6549).
@oh-my-pi/pi-natives
Changed
- astEdit without an explicit lang now rewrites mixed-language paths per file (each file parsed in its own inferred language, patterns compiled per language) instead of erroring when the path/glob spans multiple languages. A pattern that parses in no discovered language is still reported (or fails the call under failOnParseError); files whose language cannot be inferred surface as per-file parse errors instead of aborting the whole call.
What's Changed
- fix(cursor): forward write transport so xd:// previews resolve by @roboomp in https://github.com/can1357/oh-my-pi/pull/6538
- fix(coding-agent): use session settings in file guards by @roboomp in https://github.com/can1357/oh-my-pi/pull/6551
- fix(ai): disable local first-event watchdogs by @roboomp in https://github.com/can1357/oh-my-pi/pull/6525
- fix(coding-agent): guard retain renderer streaming args by @roboomp in https://github.com/can1357/oh-my-pi/pull/6529
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.1.2...v17.1.3
17.1.1 24 de juliol del 2026 — 64 bits Portable
17.1.0 24 de juliol del 2026 — 64 bits Portable
17.0.9 23 de juliol del 2026 — 64 bits Portable
17.0.8 23 de juliol del 2026 — 64 bits Portable
17.0.7 21 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Portkey/gateway custom models whose ids start with @ (e.g. @modal/GLM-5-2-FP8) being rewritten to unrelated bundled wire ids (e.g. glm-5-2), which caused 400 responses requiring x-portkey-config or x-portkey-provider.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v17.0.6...v17.0.7
17.0.6 20 de juliol del 2026 — 64 bits Portable
17.0.5 18 de juliol del 2026 — 64 bits Portable
17.0.3 17 de juliol del 2026 — 64 bits Portable
17.0.2 17 de juliol del 2026 — 64 bits Portable
17.0.1 16 de juliol del 2026 — 64 bits Portable
17.0.0 15 de juliol del 2026 — 64 bits Portable
16.5.2 14 de juliol del 2026 — 64 bits Portable
16.5.1 14 de juliol del 2026 — 64 bits Portable
16.5.0 14 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added an automated image-dropping rescue tier to compaction dead-end recovery.
- Added visual warnings and detailed recovery instructions to the session timeline when compaction fails to free sufficient space.
@oh-my-pi/pi-ai
Added
- Added diagnostic response headers to auth-gateway inference endpoints, including request IDs (x-request-id/request-id), LiteLLM model metadata (x-litellm-model-id/x-litellm-model-api-base), and performance/cost metrics (x-litellm-response-cost, x-litellm-response-duration-ms, openai-processing-ms) on non-streaming responses.
Changed
- Updated Google and Google Vertex providers to always use streamGenerateContent requests.
Fixed
- Fixed empty provider responses (such as from Cloud Code Assist API) being classified as non-retryable, allowing session retries and model-fallback chains to engage instead of failing the turn.
Removed
- Removed automatic /interactions chaining for follow-up turns in Google provider calls, along with the useInteractionsApi, storeInteraction, and previousInteractionId stream options.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Replaced the --reasoning-slide-* flag family with a unified --prewalk mechanism (--prewalk, --prewalk-into <model>, and --no-prewalk) to manage model handoffs during execution.
Added
- Added a new --prewalk execution flow (with --prewalk-into <model> and --no-prewalk overrides) that starts tasks on a strong model for planning and todo initialization before handing off to a faster, cheaper model for implementation.
- Added a status line annotation for the active prewalk phase (armed or active).
- Added the tui.scrollbackRebuild setting to gate the erase-and-replay native scrollback rebuild mechanism (defaults to off).
- Added a display setting to toggle between collapsing or keeping compacted history inline in live session displays.
- Added a compact session-only model picker (Alt+P) for quick model switching, featuring @ search to quickly list and apply configured quick roles.
- Redesigned Agent Hub entries into a cleaner two-line card layout showing identity, active model, reasoning level, age, and task description.
- Added a project-scoped launch tool (gated by launch.enabled) for managing shared long-running services and debuggers, featuring readiness probes, bounded logs, PTY input, restart policies, and automatic teardown.
- Added support for detached launches, allowing standalone services to survive broker shutdowns and reconnect to subsequent sessions.
Changed
- Updated JSON logs (--mode json) to include provider payloads in auto-compaction events.
- Updated tangential agent forks (/tan) to ignore parent session history and focus exclusively on the new request, hardening isolation with cleared todo lists and concurrent editing warnings.
- Added visual markers in the transcript for elided tool calls that have no corresponding result.
- Updated the status event log to prioritize the most recent entries in the display window.
- Upgraded @agentclientprotocol/sdk to version 1.2.1.
Fixed
- Fixed terminal scrollback duplication issues with expanded streaming edit previews (Ctrl+O) by using a viewport-sized tail window.
- Fixed custom model role resolution and alias parsing, ensuring canonical role selectors (@role) and thinking suffixes resolve correctly across all configuration surfaces.
- Fixed quadratic growth in JSON logs by eliding redundant message snapshots and payloads.
- Fixed prompt cache misses and incorrect cache key pinning for /tan and /fork clones.
- Fixed inconsistent history rendering and scrollback repainting when toggling the display setting for compacted items.
- Fixed retry.fallbackChains failing to engage on non-retryable provider errors, ensuring the agent correctly falls back to the next candidate model.
- Improved auto-compaction to automatically drop images and elide content when context is tight, and added persistent warning badges when manual intervention is required.
- Fixed backgrounded Ba
16.4.8 12 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Enabled Home and End keyboard navigation in the model browser
- Added a c hotkey in the plan-review overlay that copies the current reviewed plan markdown to the system clipboard, including in-overlay edits.
Changed
- Streamlined list view styling by removing inline model role chips from row entries
- Reworked /models hub selection visuals: the background highlight band is reserved for mouse hover, the keyboard position is a cursor glyph drawn only in the pane that owns the arrow keys, and the sidebar's active scope renders as a bold accent label
- Removed the redundant "login" label from inactive (locked) provider entries in the Model Hub sidebar
Fixed
- Improved search reliability for Perplexity provider by forcing retrieval for all queries
- Fixed JS eval cells losing top-level function and var declarations across cells when the defining cell contained top-level await — the async wrapper scoped them to the cell's IIFE instead of publishing them to the worker global
- Fixed PageUp/PageDown in the model browser wrapping past the list edges instead of clamping
- Fixed the hover highlight sticking to the last hovered model row when the pointer moved into the provider sidebar
@oh-my-pi/omp-stats
Fixed
- Fixed a SQLITE_CONSTRAINT_NOTNULL crash (messages.stop_reason) aborting the entire session sync when a persisted assistant message lacks a stopReason. Malformed entries — missing stop reason, token counts, or message timestamp — are now coerced at the parser boundary, and entries with no usage or model attribution are skipped instead of failing the batch insert.
@oh-my-pi/pi-tui
Fixed
- Fixed keyboard navigation paying an extra frame of input latency after idle; the queue-drain grace now applies only to Ctrl+C and Escape double-press gestures.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.4.7...v16.4.8
16.4.6 12 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added asynchronous invalidateUsageCache method to clear cached usage reports
- Added support for cross-service usage cache invalidation between AuthStorage and AuthBroker
Fixed
- Fixed OAuth credential resolution returning "No API key found" when every plan-eligible OpenAI Codex account was rate-limit blocked and the only unblocked account failed the model's plan gate: resolution now runs a last-resort ladder that first yields a plan-fitting account regardless of usage blocks (so callers get real usage-limit retry semantics), then tries every account with the plan filter dropped before reporting no credential
@oh-my-pi/pi-coding-agent
Added
- Added invalidate action to the usage command to clear cached usage reports
- Added model-oriented keys and wildcard entries to retry.fallbackChains: a provider/model-id key attaches a fallback chain to that exact model, a provider/* key covers every current or future model of a provider, and a provider/* chain entry keeps the failing model's id while swapping the provider (google-antigravity/x → google/x) — so fallbacks survive role and model reassignments without config edits. Keys resolve by specificity: exact model, then provider wildcard, then role, then default.
- Added fallback-chain editing to the /models Roles view: each role's retry.fallbackChains entries render as indented rows beneath it, f picks a fallback model to append, Enter on an entry replaces it, x/backspace removes it, and [/] (or shift+↑/↓) reorder the chain.
- Added model-keyed fallback management to the /models Roles view: model and provider/* chains render as a separate section below the roles (divider + "+ New fallback…" row for creating one by picking the protected model, then keying it by model or provider), with the same replace/remove/reorder editing as role chains; the model strip gains fallbacks:<model> and fallbacks:<provider>/* chips as shortcuts.
- Added /queue <message> plus -> / => composer shorthand for follow-up messages that wait until the agent yields. The shorthand opens a dim Queueing header and splits sequential numeric, Roman-numeral, or alphabetic lists into separately highlighted queue entries.
- Added per-model TPS/TTFT tracking: every completed assistant turn folds its timing into recency-weighted aggregates in ~/.omp/agent.db, and the /models browser shows measured speed — a right-aligned 118t/s column on wide terminals (plus TTFT, e.g. 0.9s 118t/s, when wider) and ~118t/s · 0.9s ttft facts in the selection detail line — with no dependency on the omp stats session scan.
Changed
- Retain completed and abandoned tasks in session history for improved context on resume
- Changed the Model Hub retry-fallback strip chip to append the model to the default fallback chain instead of prepending it, matching the chain-building order of the Roles view (already-registered models are a no-op).
- Changed per-model perf recording (recordModelPerf) to be deferred like prompt history: samples are batched and written to agent.db in one transaction ~100ms later, keeping SQLite writes off the turn-completion hot path.
Fixed
- Fixed failure to trigger model fallback when the retry budget is exhausted by credential rotation
- Fixed uncontrollable mouse-wheel scrolling in the /models hub: the wheel moved the selection (one step per wheel event, so a single trackpad flick skipped many rows) and wrapped from the bottom back to the top. Wheel scrolling now pans the list viewport only, clamps at the ends, and leaves the selection where it is; keyboard navigation still scrolls the selection into view. Likewise, the wheel over the provider sidebar no longer switches the active scope (or triggers provider refreshes) — it just scrolls the sidebar.
- Fixed TPS being inflated several-fold when a provider hides reasoning tokens until late in the stream (e.g. google/gemini-3.5 vs google-vertex/gemini-3.5 reporting 648 vs 186 TPS for identical durations): omp bench, the per-turn usage row, and the /mode
16.4.5 11 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added a process-global pause gate (agentPauseGate) to safely pause agent loops before model calls or tool executions, allowing them to be resumed later or aborted cleanly.
@oh-my-pi/pi-ai
Fixed
- Fixed an issue in GLM tool calling where missing or malformed argument closers (such as <arg_value> mistyped as </arg_key>) caused subsequent arguments to be swallowed or merged into a single field, affecting both in-band and native tool calling.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Reworked the task tool wire schema: moved the top-level agent field into individual task items, renamed assignment to task and id to name, and removed the role and description fields. UI labels are now automatically generated from the task text.
Added
- Introduced a fullscreen, mouse-supported Model Hub (via /model) featuring a sidebar of scopes, metadata-aligned model tables, inline role/thinking assignment strips, custom role creation, quick-switch cycle editing, and manual provider refreshing.
- Added a /pause command to freeze all active agents (main, subagents, and advisor) at their next safe step, allowing manual repository edits mid-run before resuming.
- Added support for per-ID bulk conflict directives via write({ path: "conflict://*", content: "..." }) to resolve multiple conflicts in a single call.
- Added auto as a valid thinking-level in agent frontmatter, which is now the default for the bundled task subagent.
- Added rich, interactive, fixed-height ask dialogs featuring question tabs, option previews, notes, and multi-select toggles.
Changed
- Redesigned OAuth logins to run inside a cancellable dialog (aborted via Esc) rather than an inescapable pairing prompt.
- Reworked the subagent soft request budget to gracefully steer child agents to wrap up and yield partial findings instead of silently terminating them, and raised the default budget to 200 requests.
- Updated task rendering to retain the agent type badge on live progress and finished result rows.
Fixed
- Fixed issues where in-flight tool calls or task blocks would disappear from the chat during mid-turn transcript rebuilds or when background jobs settled.
- Fixed mixed blocking and non-blocking task batches degrading all spawns to synchronous execution; execution mode is now correctly handled per item.
- Fixed budget-stopped subagents becoming unreachable, allowing them to remain adopted and resumable via irc with full context.
- Fixed code duplication in write conflict:// when models pasted lines adjacent to the marker block.
- Fixed visible per-keystroke lag when searching in the /resume session picker by caching search targets and debouncing SQLite lookups.
- Fixed compiled Linux binary extension loading failures related to bundled web-search header generation data paths.
- Fixed job list and empty-poll snapshots returning empty output, ensuring running subagents without backing jobs are properly listed.
- Fixed agents getting stuck waiting for messages from peers that have already stopped running.
- Fixed compiled Linux binary extension loading when bundled web-search header generation cannot read header-generator data files from the build-time path. (#5178)
- Fixed plugin custom tool loading to skip and report invalid feature entries instead of crashing startup when a plugin dependency tree leaves one feature unresolved. (#5189)
@oh-my-pi/pi-natives
Added
- Added context-safe, in-process shell builtins for common utilities including base64, basename, dirname, cut, tee, tr, paste, comm, sed, xargs, jq, and the md5sum/sha/b2sum checksum family. These builtins run without spawning external binaries, support pipelines, respect shell-relative paths and environment variables, and honor abort/timeout cancellation.
@oh-my-pi/pi-tui
Added
- Added FuzzyText, a prepared fuzzy-match handle that builds the search index once and matches many queries against it, optimizing performance for large corpora like session or transcript searches.
Fixed
- Fixed a
16.4.4 11 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Changed
- Optimized session title generation and auto-thinking classification for sub-billion-parameter tiny models (such as LFM2) by rewriting system prompts, improving input truncation to preserve message context, and unifying preprocessing to filter out noise like ANSI codes, XML tags, and long commit hashes.
Fixed
- Fixed an issue where the Windows binary exited silently without running the CLI, which also caused omp update to roll back.
- Fixed native Windows binary compatibility on older Windows 10 CPUs by building the omp-windows-x64.exe release asset with a baseline x64 runtime instead of AVX2. (#5172)
- Fixed GenerateImage rejecting OpenAI Codex-compatible proxy bearer keys when the token does not expose a chatgpt-account-id. (#5174)
- Fixed context promotion documentation to accurately reflect the contextPromotionTarget runtime behavior and contextPromotion.enabled default. (#5163)
@oh-my-pi/pi-natives
Fixed
- Fixed fuzzyFind tie-breaking logic to prefer shallower paths first, preventing deeply nested matches from ranking above shallow ones on score ties.
- Fixed macOS installation issues for pi-natives by statically linking PCRE2, removing the runtime dependency on Homebrew's dynamic libpcre2-8.0.dylib library.
What's Changed
- docs(coding-agent): correct context promotion target docs by @roboomp in https://github.com/can1357/oh-my-pi/pull/5165
- fix(setup): use baseline Windows binary target by @roboomp in https://github.com/can1357/oh-my-pi/pull/5173
- fix(tools): allow opaque Codex image keys by @roboomp in https://github.com/can1357/oh-my-pi/pull/5176
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.4.3...v16.4.4
16.4.3 11 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed an issue where skipped sibling tool results incorrectly reported that a queued user message caused the skip.
@oh-my-pi/pi-ai
Fixed
- Fixed auth database upgrades from schema v5 by creating the OAuth credential refresh-lease table before lease statements are prepared.
- Fixed an issue in the Responses API where empty tool results were incorrectly serialized with a "(see attached image)" placeholder, causing models to look for non-existent attachments.
- Fixed OpenAI Responses server non-streaming envelopes to always include the required "incomplete_details" field, using null for completed responses.
- Preserved Cloud Code Assist tool schemas when mixed-type unions carry branch-local validation descriptions.
@oh-my-pi/pi-catalog
Fixed
- Fixed parsing of SAP AI Core Claude model IDs in version-first format (e.g., anthropic--claude-4.8-opus), restoring adaptive thinking metadata and capability gates.
- Fixed GitHub Copilot Business and Enterprise model discovery to correctly preserve vision capabilities instead of downgrading models to text-only.
@oh-my-pi/pi-coding-agent
Added
- Added /vibe mode, allowing the model to act as a director driving persistent background worker sessions (fast and good tiers) with dedicated session tools (vibe_spawn, vibe_send, vibe_wait, vibe_kill, vibe_list) and a live TUI "TV wall" showing active worker activity, tool traces, and streamed output.
- Added multiple credential-free web search providers (Google, Bing, Yahoo, Startpage, Ecosia, Mojeek) with stealth-browser escalation, bot challenge detection, and recency filters, alongside a parallel public ("Public Web") provider that aggregates and deduplicates results across all engines.
- Added PCRE2 fallback support for grep lookaround and backreferences when the default Rust regex engine rejects a pattern.
- Added a helpful stderr hint when launching omp acp from an interactive terminal to clarify that the command communicates via JSON-RPC over stdout.
Changed
- Reduced browser action timeout from 15s to 8s to improve agent iteration speed.
- Refined agent delegation logic to prioritize top-level planning by the primary agent, discourage single-agent delegation, and handle prerequisite work inline.
- Optimized credential-free web search engine ordering and routing, prioritizing Startpage and Ecosia, and using randomized desktop Chrome profiles with stealth-browser escalation for blocked requests.
Fixed
- Fixed advisor config preserving an explicit empty tool list so /advisor config can disable all advisor tools. (#5155)
- Fixed npm bundle generation failing on Linux with E2BIG by building dist/cli.js in-process via Bun.build instead of passing the embedded docs payload as a CLI --define argument.
- Fixed compiled-binary extensions failing to load native .node FFI dependencies by resolving platform-specific packages against the extension's own node_modules.
- Fixed a hang in omp search and omp q CLI commands by ensuring the AuthStorage connection is properly closed upon completion.
- Fixed write blocking for the full LSP diagnostics poll by deferring slow diagnostics to a late-diagnostics channel.
- Fixed multiple browser and puppeteer tool issues, including locator action timeouts, missing console/print output buffering, missing fill() method on element handles, and incorrect viewport screenshots on multi-tab setups.
- Improved browser selector error diagnostics to report match counts and abort early on empty matches instead of waiting for the full timeout.
- Fixed silent failures, duplicate error messages, and unhandled provider errors in ACP mode when errors occurred before streaming assistant text.
- Fixed glob reporting contradictory "no files found" messages on timeout by explicitly stating the scan was incomplete.
- Fixed read adding unwanted context padding to raw range selectors (e.g., raw:31-31).
- Fixed first-run interactive startup rendering the entire changelog when the last-seen marker is missing or
16.4.2 10 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed serialization of BigInt tool arguments to prevent data loss during remote compaction.
@oh-my-pi/pi-ai
Fixed
- Fixed compatibility with xAI by automatically downgrading OpenAI-specific tool calls and image detail settings during message history replays.
- Fixed a race condition in shared SQLite OAuth token refreshes by implementing durable credential ownership and compare-and-set persistence to prevent stale refresh failures.
- Fixed OpenAI Codex requests to include the required version header for newly gated models.
@oh-my-pi/pi-catalog
Fixed
- Fixed OpenAI Codex model discovery to include the Codex version header alongside the client_version query parameter.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed an issue where BigInt values in tool arguments failed to serialize during session compaction.
- Resolved an issue where GPT-5.6 over-delegated tasks by refining task fan-out and concurrency policies in the system prompt.
- Fixed a race condition in concurrent MCP OAuth token refreshes across processes, ensuring rotating refresh tokens are only refreshed once and preventing stale token errors from clearing valid credentials.
@oh-my-pi/omp-stats
Fixed
- Fixed a crash during stats synchronization on legacy session entries that lack a cost breakdown by falling back to catalog pricing when available.
@oh-my-pi/pi-utils
Added
- Added stringifyJson utility with support for BigInt serialization.
What's Changed
- fix(codex): send version header with Codex requests by @roboomp in https://github.com/can1357/oh-my-pi/pull/5108
- fix(auth): serialize MCP OAuth refreshes by @roboomp in https://github.com/can1357/oh-my-pi/pull/5116
- fix(ai): adapted xai responses replay shapes by @any-victor in https://github.com/can1357/oh-my-pi/pull/5109
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.4.1...v16.4.2
16.4.1 10 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Enabled reasoning encryption content for all Responses Lite compaction requests
@oh-my-pi/pi-ai
Changed
- Enforced all_turns reasoning context for all Responses Lite requests
@oh-my-pi/pi-catalog
Added
- Added GPT-5.6 Luna, Sol, and Terra models
- Added perplexity-academic-researcher model
Changed
- Updated context windows for multiple GPT-5.6 models
- Increased max tokens for several models
- Updated cache write costs for GPT-5.6 variants
- Reduced pricing for select models
Removed
- Removed the generated GPT-5.6 pro-reasoning aliases (gpt-5.6-{luna,sol,terra}-pro) from the openai-codex subscription provider — pro reasoning is not offered on subscriptions; the openai API-key aliases remain
@oh-my-pi/pi-coding-agent
Changed
- Reduced agent bias against large diffs and refactors in advisor prompts
- Updated advisor blocker criteria to prioritize explicit user instructions over plan size
Fixed
- Fixed MCP OAuth dynamic client registration omitting discovered scopes on the RFC 7591 registration body. Providers such as Clerk bind DCR-created clients to only the scopes declared at registration, then reject the subsequent authorize request when it asks for openid (from scopes_supported). Registration now includes config.scopes when present, matching Claude Code and the scopes already sent on authorize.
@oh-my-pi/pi-tui
Added
- Added full 2-D layout support for display LaTeX math (fractions, matrices, radicals, limits), modeled on the layout approach of txm (Terminal TeX Math) by @thatmagicalcat
- Added support for \left, \right, and \middle stretchy delimiters in display blocks
- Added rendering for cases, matrix, pmatrix, bmatrix, and vmatrix environments
- Added support for block-level scripts and big-operator limits (e.g., \sum, \int\limits)
- Added cross-box styling for \color, \textcolor, and math font commands (e.g., \mathbf)
Changed
- Improved row alignment and spacing for align, gather, and array environments
- Updated matrix environments to render as baseline-aligned grids with stretched brackets
What's Changed
- fix(mcp): included OAuth scopes in dynamic client registration by @any-victor in https://github.com/can1357/oh-my-pi/pull/5107
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.4.0...v16.4.1
16.4.0 10 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added the ThinkingLevel.Max ("max") configuration option, mapping to the Effort.Max tier for supported models.
Fixed
- Fixed remote compaction behavior for Codex Responses Lite (GPT-5.6 family) models across both V1 and V2 endpoints to ensure correct formatting and routing.
- Fixed an issue where aborted tool-result hooks could trigger subsequent provider calls before the abort signal fully settled.
@oh-my-pi/pi-ai
Added
- Added "max" as a first-class reasoning effort option across providers (including Anthropic, Google, Bedrock, and OpenAI), supporting a maximum reasoning budget of 32,768 tokens.
- Added and standardized the "Responses Lite" wire contract and transport, enabling automatic activation via model-level catalog flags, moving tools and instructions into developer input items, disabling parallel tool calls, and stripping image detail instead of falling back to the full transport.
- Added support for concurrent reasoning summaries on Codex Responses using the sequential-cutoff streaming contract.
- Added Novita API-key login with authenticated key validation and automatic NOVITA_API_KEY environment variable discovery.
Changed
- Recognized Pro Lite as a paid plan tier for OpenAI Codex models.
Fixed
- Fixed xAI SuperGrok multi-account rotation to correctly treat HTTP 403 credit exhaustion and spending limit errors as usage limits, triggering a credential rotation to a sibling account.
- Fixed error classification for AWS credential-resolution failures (AwsCredentialsError) to correctly map them as authentication failures.
- Fixed OpenAI-compatible chat-completions streams to preserve vLLM-style trailing cached-token usage chunks, ensuring accurate cacheRead and billable input session statistics.
- Fixed xai-oauth/grok-4.5 Responses requests to omit the unsupported reasoning.summary field while preserving the reasoning.effort payload.
- Fixed Codex OAuth credential selection to re-check blocked accounts during ranking and clear stale usage-limit blocks once live usage indicates recovery.
- Fixed sequential-cutoff reasoning summaries duplicating section headers across Codex reasoning items by tracking the cumulative summary response-globally, so replayed sections and replay-only items no longer re-emit text earlier thinking blocks already streamed.
@oh-my-pi/pi-catalog
Breaking Changes
- Redesigned reasoning effort ladders to be wire-exact, removing the shifted five-tier effort mapping. Models now expose exactly the effort tiers their upstream APIs accept, mapped 1:1. Removed SHIFTED_FIVE_TIER_EFFORT_MAP, ANTHROPIC_ADAPTIVE_EFFORT_MAP_4_TIER, and per-host xhigh-to-max alias maps. Selecting an unsupported tier now automatically clamps down via clampThinkingLevelForModel. Devin effort routing is now mapped 1:1 onto per-tier siblings.
Added
- Added support for new models: Grok 4.5 family, Dolphin Mistral 24b Venice Edition, GLM5.2-Fast, and Zenmux variants for GPT-5.6 (Luna, Sol, and Terra).
- Added Novita as a model provider, including public catalog discovery, pricing, limits, modality, reasoning, and tool metadata.
- Added useResponsesLite to Model and ModelSpec to support the Responses Lite transport, enabled by default for the GPT-5.6 family.
- Added Effort.Max ("max") as a first-class user-facing thinking level above xhigh.
Changed
- Enabled reasoning effort controls for Grok 4.5 and updated support flags for additional Grok variants
- Standardized reasoning effort levels to use a wire-exact max tier across all model providers, including Devin routing and Ollama configurations.
- Updated costs and context windows for various models in the catalog.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Renamed the bundled agent explore to scout, including its configuration keys, prompt files, and task definitions. Any configurations, allowlists, or invocations referencing explore must now use scout.
- Changed the public selectLaunchAdapter() result from DapResolvedAdapter | null to LaunchAdapterSe
16.3.15 9 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Breaking Changes
- Renamed OpenAIResponsesCacheOptions, normalizeOpenAIResponsesPromptCacheKey, and getOpenAIResponsesPromptCacheKey to the endpoint-neutral OpenAICacheOptions, normalizeOpenAIPromptCacheKey, and getOpenAIPromptCacheKey.
Added
- Added automatic prompt-cache affinity header injection for OpenAI-family chat completions
- Added support for explicit prompt-cache affinity headers in OpenAI-family chat completions
- Added OpenAI pro reasoning mode support: models carrying the catalog reasoningMode: "pro" marker (GPT-5.6 Pro aliases) send reasoning: { mode: "pro" } on OpenAI Responses and Codex Responses requests, alongside the configured effort. The Codex request body now honors requestModelId so catalog aliases request the base upstream model id.
Changed
- Updated xAI OAuth to use a dedicated device-code flow instead of redirect/loopback server
Fixed
- Improved account routing for GPT-5.6 models to better respect paid tier requirements
- Refined account selection logic to correctly identify plan types from account metadata
- Fixed OpenAI Codex multi-account routing for GPT-5.6: Sol and Luna requests now prefer Plus-or-higher accounts while Terra remains available to Free/Go accounts; local pro-mode aliases inherit their base model's Codex plan eligibility.
- Fixed xAI Grok OAuth login to use xAI's device authorization flow: /login now opens the verification URL, displays the device code, and polls for approval instead of asking for a pasted redirect or linking to Hermes Agent documentation.
@oh-my-pi/pi-catalog
Added
- Added support for Grok 4.5 model
- Added gpt-5.6 base models and gpt-5.6-{luna,sol,terra}-pro variants
- Added meta/muse-spark-1.1 model support
- Added support for thinking modes on poolside/laguna models
- Added generated GPT-5.6 Pro aliases (gpt-5.6-{luna,sol,terra}-pro) on the openai and openai-codex providers: each alias sends the base model id on the wire (requestModelId) with the new reasoningMode: "pro" marker, and re-derives from the current base rows on every catalog regeneration.
Changed
- Updated cache read costs for Grok models
- Reduced max token limit for Grok 4.3 model
- Enabled prompt cache affinity for Grok models via the x-grok-conv-id header in OpenAI compatible endpoints
- Enabled prompt cache affinity for Grok models via the x-grok-conv-id header
- Marked direct xAI Grok Chat Completions models for x-grok-conv-id prompt-cache affinity.
@oh-my-pi/pi-coding-agent
Changed
- Integrated testing guidance directly into the main system prompt for improved workflow cohesion
- Moved testing guidance into the main system prompt and removed the bundled Tester subagent.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.14...v16.3.15
16.3.14 9 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Updated Codex reasoning effort mapping to support shifted wire tiers for newer models
Fixed
- Fixed the Codex Responses request transformer bypassing catalog/compat reasoning effort maps: the clamped user effort is now remapped to the provider wire tier (GPT-5.6's shifted five-tier scale sends max for user xhigh and xhigh for high), failing loudly if a map produces a value outside the Codex wire vocabulary.
@oh-my-pi/pi-catalog
Added
- Added support for GPT-5.6 (Luna, Sol, Terra) model variants
- Enabled expanded five-tier reasoning effort scale (minimal to xhigh) for GPT-5.6 models
- Added GPT-5.6 (Terra/Luna/Sol) support for the new max reasoning tier: on wire-effort APIs (OpenAI Responses, Codex, Azure, openai-compat/OpenRouter models that advertise reasoning) user efforts shift up one notch — xhigh sends max, high sends xhigh — mirroring the Claude Fable/Opus 4.7+ five-tier mapping, and the exposed ladder becomes minimal..xhigh with minimal reaching the native low tier. Devin's per-tier GPT-5.6 sibling rows now collapse into gpt-5-6-{luna,sol,terra} logical models with the same shifted routing (xhigh → -max), plus -fast families that keep the direct low..xhigh -priority scale since Devin serves no -max-priority tier.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed issue where unfinalized tool blocks could incorrectly pin the live-region scroll seam
- Improved rendering of raw thinking blocks by stripping empty HTML comment noise
- Fixed display of thinking blocks consisting entirely of hidden comment noise
- Fixed gpt-5.6 reasoning summaries rendering literal <!-- --> sentinel lines in thinking blocks; empty HTML comments (and the unterminated <!-- tail while streaming) are now dropped from the thinking display, and blocks reduced to pure comment noise are hidden entirely.
@oh-my-pi/pi-tui
Fixed
- Fixed race condition where scrollback rows could be incorrectly retracted between render frames
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.13...v16.3.14
16.3.13 9 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Changed the xAI Grok OAuth (xai-oauth) provider to use manual code-paste login by default. /login now accepts a pasted authorization code or full http://127.0.0.1:56121/callback?code=... redirect URL without starting a local callback listener (#3277 by @Jaaneek).
- Renamed the xAI Grok OAuth provider in login and credential prompts to "xAI Grok OAuth (SuperGrok or X Premium+)" (#3277 by @Jaaneek).
Fixed
- Fixed the generic lazy-stream idle watchdog aborting healthy cursor-agent streams with "Provider stream stalled while waiting for the next event" while a Cursor exec-channel local tool (shell/read/grep/write/MCP/…) legitimately ran longer than the idle budget. Provider streams now advertise consumer-side local work in flight and the watchdog slides its deadline instead of aborting; genuinely silent streams still time out. (#4593)
- Fixed OpenAI Codex/Responses reasoning streams so streamed thinking content is preserved when the final output_item.done reconstructs to an empty summary (#4918).
- Fixed Anthropic streams hanging forever when generation wedges mid-stream (notably long write tool calls on Opus 4.8 high/xhigh) while the server keeps sending ping keepalives: pings now extend the idle watchdog only within a bounded window (3x the idle timeout) since the last real stream event, so a stalled tool-call stream times out and recovers instead of hanging with no retry path (#4900).
@oh-my-pi/pi-catalog
Added
- Added support for Grok 4.5 across multiple providers
- Added support for GPT-5.6 series models (Luna, Sol, Terra)
- Added Aion 3.0 and 3.0 Mini models
- Added Kuaishou KAT-Coder v2.5 models
- Added Nex-N2-Mini and SWE-1.7 series models
- Added Hy3 models and free variants
Changed
- Updated cost and token configurations for various models across providers
- Renamed several models for consistency (e.g., MiniMax M3, Gemma 4 31B, Qwen variants)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed read and grep treating empty optional selector fields emitted by models as invalid selectors instead of behaving like omitted selectors. (#4879)
- Fixed grep explicit line selectors on directory searches so they filter each matched file by line number instead of aborting with a single-file-only error (#4898).
- Fixed Read tool previews dropping explicit selector arguments, so line ranges and raw modifiers render in terminal read call titles again (#4899).
- Fixed named profiles dropping default user keybindings from ~/.omp/agent/keybindings.*; profile keybindings now inherit those defaults and override only the keys they define (#4867).
- Fixed pi extensions calling ctx.ui.addAutocompleteProvider(...) crashing at load with TypeError: ... is not a function — a failure that, for extensions guarding init in one try/catch (e.g. @ff-labs/pi-fff), aborted the extension's entire initialization. ExtensionUIContext now implements pi's autocomplete-provider API: interactive mode stacks each registered factory on top of the built-in editor provider (re-applied on every slash-command refresh, with throwing or malformed factories skipped), while RPC/ACP/headless contexts accept the factory as a no-op. (#4919)
- Fixed bundled reviewer and plan subagents to inherit their model roles' explicit thinking effort suffixes instead of pinning high (#4761).
- Built-in provider model discovery now refreshes an expired stored OAuth credential before an online refresh needs it, instead of silently skipping the provider. The refresh is scoped to the providers actually being discovered (refreshProvider cannot rotate unrelated credentials), fires under online-if-uncached only when the model manager will actually fetch, and offline discovery stays peek-only (#4893).
- Fixed Escape during an active TUI prompt requiring a second press before canceling; the first Escape now aborts the streaming turn immediately. (#4921)
- Fixed the streamed write tool's collapsed pending tail preview leaving stale rows above the first partial-result frame in the TUI;
16.3.12 8 de juliol del 2026 — 64 bits Portable
16.3.11 6 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed openai-codex-responses fresh plan execution requests that contained only system/developer guidance by mirroring the final instruction as user input so Codex accepts the first turn. (#4714)
- Fixed Codex WebSocket compact/resume delta diagnostics to record request shape and raw-vs-displayed usage buckets, so persistent server-reported uncached suffixes without orchestration_* fields are visible in debug stats. (#4707)
@oh-my-pi/pi-catalog
Added
- Added Claude Haiku 4.5 (JP) model support
- Added tencent/hy3 model support via ZenMux
Changed
- Updated naming format for various synthetic models to include provider prefix
- Adjusted context window limit for MiniMax-M3 model
- Updated pricing for select models
@oh-my-pi/pi-coding-agent
Changed
- Improved session title generation reliability by moving to marker-based parsing for all models
Fixed
- Fixed session titles occasionally showing raw {"title": "..."} JSON. Online title generation now always uses the <title>...</title> marker prompt instead of a forced set_title tool call — hosts that ignored or rejected forced tool_choice echoed the prompt's JSON example verbatim as the title — and JSON-shaped responses (bare, code-fenced, marker-wrapped, or truncated) are unwrapped to the bare title.
- Fixed Linux startup prompt construction to read the CPU model from /proc/cpuinfo instead of os.cpus(), avoiding per-core sysfs frequency probes on many-core hosts (#4712).
- Fixed llama.cpp model discovery to honor per-model architecture.input_modalities from /v1/models, so router presets that advertise image input are no longer treated as text-only (#4719).
What's Changed
- fix(tool): include ignored files for empty grep glob discovery by @roboomp in https://github.com/can1357/oh-my-pi/pull/4710
- fix(ai): record Codex compact usage diagnostics by @roboomp in https://github.com/can1357/oh-my-pi/pull/4713
- fix(cli): avoid os.cpus during startup prompt by @roboomp in https://github.com/can1357/oh-my-pi/pull/4717
- fix(ai): mirror Codex instruction-only plan input by @roboomp in https://github.com/can1357/oh-my-pi/pull/4718
- fix(coding-agent): detect llama.cpp per-model vision metadata by @roboomp in https://github.com/can1357/oh-my-pi/pull/4721
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.10...v16.3.11
16.3.10 6 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Ollama/Ollama Cloud EOS-only completions to retry empty stops with a single output token before the agent loop can halt silently. (#4659)
- Fixed Claude Sonnet 5 failing every request on feature-gated gateways (Azure Foundry, OpenAI-compatible relays) that reject strict tools with "structured_outputs not supported" — the rejection is now classified as a strict-tool rejection, so the request retries without strict tools and the session remembers the downgrade.
@oh-my-pi/pi-catalog
Fixed
- Fixed LiteLLM rich discovery to ignore unusable sentinel placeholders and continue to /v2/model/info for real models. (#4655)
@oh-my-pi/pi-coding-agent
Changed
- Limited subagent HUD display to 8 items with a summary for additional running subagents
- Improved TUI performance by coalescing agent registry and observer UI updates
Fixed
- Fixed high-subagent sessions overwhelming the TUI by bounding the running-subagent HUD and coalescing subagent progress repaint bursts.
- Fixed browser run cleanup crashing or wedging the whole omp session: run-end and tab-close aborts could reject fire-and-forget wait()/facade/tab promises with no consumer, and the global unhandled-rejection handler then exited the process, killing every subagent sharing it. Run-scoped promises are now observed at creation and routine browser teardown aborts are downgraded to log lines (#4499, #4672).
- Reduced CPU use while many task subagents stream progress by disabling the obsolete task partial-result spinner repaint loop (#4424).
- Capped collapsed nested subagent trees at the same per-level agent limit as top-level task rendering (failures prioritized, elided rows summarized), so deep many-subagent sessions no longer render unbounded progress trees on every repaint.
- Fixed skill card headers to render a single space between the skill tag and skill name (#4662).
- Fixed get_session_stats RPC responses to include context-window usage so RPC clients can render context meters.
- Fixed startup of cached llama.cpp vision models so the initial default/restored model refreshes /props metadata before the session exposes it as text-only.
- Fixed IRC-woken yielded subagents skipping empty-stop retry because stale yield-termination state carried into the wake turn (#4658).
- Fixed irc wait skipping replies that arrived between wait calls by draining pending IRC asides before honoring queued-interrupt aborts (#4657).
@oh-my-pi/pi-tui
Fixed
- Registered emergency terminal restore with the postmortem fatal path whenever a real terminal starts, so fatal exits restore raw mode/alternate screen on the normal CLI graph (previously the registration was a side effect of a barrel module the CLI never imported).
@oh-my-pi/pi-utils
Added
- Added postmortem.markExpectedCleanupError() / postmortem.isExpectedCleanupError() to tag errors thrown by routine resource teardown; the global uncaughtException/unhandledRejection handlers downgrade marked errors (walking the cause chain) to warnings instead of exiting the process.
Fixed
- Bounded postmortem cleanup with a 10s deadline so a hanging cleanup callback can no longer wedge the process indefinitely after a fatal error or signal; the process now always reaches process.exit.
What's Changed
- fix(catalog): ignore LiteLLM all-team-models placeholders by @roboomp in https://github.com/can1357/oh-my-pi/pull/4661
- fix(ai): retry empty Ollama completions by @roboomp in https://github.com/can1357/oh-my-pi/pull/4663
- fix(agent): reset IRC wake maintenance state by @roboomp in https://github.com/can1357/oh-my-pi/pull/4664
- fix(agent): surface pending irc replies to wait by @roboomp in https://github.com/can1357/oh-my-pi/pull/4665
- fix(tui): correct skill card header spacing by @roboomp in https://github.com/can1357/oh-my-pi/pull/4666
- fix(rpc): include context usage in session stats by @roboomp in https://github.com/can1357/oh-my-pi/pull/4671
- fix(agent): refresh startup llama.cpp vision metadata by @roboomp in https://git
16.3.9 6 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Fixed
- Fixed compatibility with OpenCode Go DeepSeek V4 models by sending max_tokens instead of max_completion_tokens to match the provider's API requirements.
@oh-my-pi/pi-coding-agent
Added
- Added a --file flag to the say command to read input text directly from files.
- Enabled streaming text synthesis in the say command for gapless, long-form audio generation.
- Added voice selection validation to the say command.
Changed
- Improved the say command to display the segment count and total duration upon completion.
Fixed
- Fixed an issue where local llama.cpp vision models remained text-only after a model refresh, ensuring they are correctly recognized as image-capable when configured as the default or vision role.
- Fixed omp commit split plans aborting when lock files (such as bun.lockb) were staged alongside their manifests by correctly pairing lock files with their corresponding commit groups and properly handling binary files during split execution.
- Fixed skill loading to ensure that disabling a higher-priority provider does not drop same-named skills from enabled lower-priority providers.
@oh-my-pi/pi-mnemopi
Fixed
- Fixed extractor JSON parsing to correctly unwrap object-shaped facts, instructions, preferences, and timeline items from known text fields instead of persisting literal [object Object] rows.
@oh-my-pi/omp-stats
Changed
- Refined behavior metrics to significantly reduce false positives in profanity, yelling, and anguish detection by excluding technical terms (e.g., "dummy", "trash", "garbage"), neutral punctuation (e.g., dot runs), and single-word capitalization (e.g., filenames or environment variables).
- Re-categorized frustration interjections (such as "ugh", "argh", and "grr") from profanity to anguish.
- Improved negation and blame detection to exclude determiners (e.g., "no auto start") and compounds (e.g., "no-op") while adding support for phrases like "why did you" and "makes no sense".
- Added sad emoticons as a signal for anguish while excluding code-like patterns.
- Triggered a one-time automatic re-ingestion of sessions on the next database sync to apply the updated metrics.
@oh-my-pi/pi-tui
Fixed
- Fixed autocompletion for absolute paths (such as /tmp/... or /Users/...) at the start of a prompt, ensuring they fall back to file-path completion instead of being incorrectly treated as slash commands.
- Updated absolute path autocompletion behavior so that accepting a suggestion inserts the path without submitting the prompt.
What's Changed
- fix(commit): pair staged lock files with the split-plan commit that owns them by @roboomp in https://github.com/can1357/oh-my-pi/pull/4634
- fix(catalog): use max_tokens for OpenCode Go DeepSeek V4 by @roboomp in https://github.com/can1357/oh-my-pi/pull/4651
- fix(mnemopi): drop object-shaped fact coercion by @roboomp in https://github.com/can1357/oh-my-pi/pull/4653
- fix(coding-agent): preserve enabled duplicate skills by @roboomp in https://github.com/can1357/oh-my-pi/pull/4652
- fix(coding-agent): refresh llama vision input metadata by @roboomp in https://github.com/can1357/oh-my-pi/pull/4656
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.8...v16.3.9
16.3.8 5 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Fixed the browser tool silently launching without its Puppeteer stealth patch: the patch was keyed to puppeteer-core@25.1.0 while the resolved dependency had drifted to 25.3.0, so Bun skipped it and Chrome ran with the Runtime.enable automation tell re-enabled. Regenerated the stealth patch against 25.3.0 and pinned puppeteer-core so the exact-version patch cannot silently deactivate on future drift.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.7...v16.3.8
16.3.7 5 de juliol del 2026 — 64 bits Portable
16.3.6 4 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Persisted credential rate-limit blocks across processes: auth_credential_blocks (auth schema v5) stores per-credential blocks keyed by row id + provider key + block scope with MAX-upsert semantics, AuthStorage merges persisted and in-memory blocks on read, and auth-broker snapshots/SSE carry per-entry blocks with POST /v1/credential/:id/block and DELETE /v1/credential/:id/blocks endpoints so gateway and sibling omp processes stop re-discovering exhausted accounts by burning a 429 each.
Fixed
- Fixed Anthropic credential selection sampling Fable/Mythos-exhausted accounts on every new session: a Fable/Mythos weekly cap now proactively hard-blocks the credential when confirmed exhausted (server exhausted status or used fraction >= 1) with a live resetsAt, and a live Fable 429 extends the reactive block to the confirmed tier reset instead of the 60s default. Unconfirmed rows (missing/expired reset, below cap) remain ranking hints only, preserving the false-100% guard.
- Fixed Ollama/Ollama Cloud tool requests failing with HTTP 400 by rewriting boolean subschemas (true/false) into a value-widening anyOf union of primitive types, stripping boolean additionalProperties/unevaluatedProperties, and flattening nullable type arrays before serializing tool parameters, so unconstrained fields still advertise "any JSON value" to grammar-constrained samplers (llama.cpp) instead of collapsing to an empty object. (#4488)
@oh-my-pi/pi-coding-agent
Changed
- Hided abort labels for user-interruptions in the transcript to match silent abort behavior
- Updated transcript streaming to declare settled rows for scrollback exactness: TranscriptContainer reports one boundary (finalized blocks plus the first live block's getTranscriptBlockSettledRows(), fed by markdown's frozen-token prefix — completed content blocks in a streaming reply render final and settle in full, so long replies and visible thinking reach terminal scrollback as exact bytes mid-stream). The heuristic commit machinery (deriveLiveCommitState append-only detection, stable-prefix ratchet, volatile cooldowns, rewrite floors), isTranscriptBlockCommitStable, and the per-renderer provisionalPendingPreview/provisionalPartialResult flags are all removed. Ephemeral block retraction (displaceable todo/job cards, IRC cards) is now gated on TranscriptContainer.isBlockUncommitted() — cards whose rows already entered scrollback seal in place as history instead of being removed.
- Recovered auto-retry errors now compact in the transcript: once a retry succeeds, the superseded error rows (e.g. Anthropic 429 rate_limit_error during account rotation) render as a single dim note like rate-limited; switched account; retried across live view, rebuilds, resume, and subagent/collab transcript views, and are excluded from model context on session resume. Terminal (unrecovered) errors keep full error rendering. The persisted marker is retryRecovery on the assistant message and auto_retry_end success events carry additive recoveredErrors data.
Fixed
- Fixed raw read ranges not contributing to edit seen-line provenance, so re-reading an anchor range with :raw now unblocks hashline edits without adding non-raw line prefixes.
- Fixed replan-driven session title refresh updating the statusline but not the terminal window title: terminal-title updates now fire from the session-name-changed listener, so every setSessionName path (first-input titling, /rename, plan seeding, replan refresh) sets the OSC title consistently.
- Fixed user-interrupt aborts rendering the persisted Interrupted by user label in assistant transcripts; replay and live views now suppress that redundant line again while preserving generic/custom abort labels.
- Fixed streamed assistant text and thinking disappearing from terminal scrollback while a long turn ran, and Ctrl+O-expanded tool results showing up half-rendered above the viewport: everything that scrolls off the window now reaches native scrollback (exact bytes f
16.3.5 4 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- OAuthCallbackFlow now serves a GET /launch route on its loopback callback server that 302-redirects to the pending authorization URL, and exposes that short URL as OAuthAuthInfo.launchUrl. UIs can advertise it as a truncation-safe copy target (~30 chars) instead of the full authorize URL, so terminals narrower than the composed row cannot silently drop OAuth query parameters like code_challenge_method=S256 (#4418).
- Preserved explicit tool.strict === false on OpenAI-family function tool payloads (openai-responses, openai-codex-responses, openai-completions) so backends that distinguish strict: false from an omitted flag stop over-filling optional arguments (#4336).
Fixed
- Fixed tool-call validation to strip stray trailing line terminators on schema-matching enum values and on well-known identifier fields (path, paths, file, file_path, url, uri, title, label) before dispatch, keeping ordinary trailing spaces and content-carrying fields (content, input, code, command, etc.) intact (#4461).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Mnemopi auto-retention so protocol markers are stripped from embedding and FTS projections while stored transcripts remain readable. (#4395)
- Fixed mnemopi auto-retain storing cumulative full-session transcripts on every retention interval; subsequent retains now store only newly completed user-turn suffixes. (#4396)
- Fixed large artifact:// reads materializing entire MCP/tool artifacts before selector paging, preventing OOM crashes on unbounded raw reads and surfacing bounded read/search/copy guidance (#4482).
- Fixed /mcp reauth and /mcp add dropping OAuth scopes advertised via WWW-Authenticate: Bearer scope="..." challenges and via protected-resource metadata (scopes_supported / scopes / scope), which caused tokens to be issued without the required scopes and reconnects to fail with insufficient_scope (#4467)
- Fixed task-branch merges aborting the whole cherry-pick range when a single commit was empty against HEAD (redundant with an already-applied change, or 3-way merged to HEAD); the sequencer now --skips the empty commit and continues so later non-overlapping work still lands (#4438).
- Fixed write/edit LSP diagnostics for TypeScript files outside any project root by suppressing project-resolution noise while preserving real syntax errors (#4401).
- Fixed /mcp reauth against S256-only OAuth providers (Linear, and OAuth 2.1 flows generally) failing on Windows and other environments where the browser cannot auto-open. Two independent defects converged on Linear's The plain PKCE method is not allowed. Use S256 instead. error page: openPath invoked bare rundll32 and silently swallowed the resulting Executable not found in $PATH when the Windows machine PATH no longer referenced System32, and MCPAuthorizationLinkPrompt composed a Copy URL: line wider than the viewport that TUI#prepareLine then silently truncated — trimming the trailing code_challenge_method=S256 (RFC 7636 §4.3 treats a request with code_challenge and no method as plain). The MCP OAuth fallback, /login, setup wizard, auth-broker CLI, and login-dialog now advertise the short OAuthAuthInfo.launchUrl (loopback /launch route hosted by OAuthCallbackFlow) as the copy target; the OSC 8 hyperlink still carries the full URL for click-through; the MCP flow additionally stages the copy target on the clipboard via OSC 52; and the Windows opener now resolves rundll32.exe through %SystemRoot%\System32\, logging spawn failures and non-zero exits instead of dropping them (#4418).
- Fixed lsp rename_file aborting when an LSP server returns duplicate byte-identical non-empty text edits for the same range, such as tsserver willRenameFiles edits through barrel re-exports (#4458).
- Fixed Linux x64 PI_TINY_DEVICE=cuda tiny-model side runtimes missing ONNX Runtime CUDA provider binaries by repairing the onnxruntime-node CUDA sidecar install and preserving actionable CUDA diagnostics in omp tiny-models download output (#4475).
- Im
16.3.4 3 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added support for Baseten as an AI provider
Changed
- Improved Claude usage reliability by removing proactive hard-blocking for Fable and Mythos tiers
Fixed
- Fixed Anthropic OAuth account rotation to exclude unreliable model-scoped Fable/Mythos weekly caps from proactive hard-blocking, ensuring they act only as ranking priority hints while still allowing reactive 429-fallback to rotate and reach serviceable siblings.
@oh-my-pi/pi-catalog
Added
- Added Baseten as a supported model provider
- Added support for new models from Baseten, including DeepSeek V4 Pro and Kimi series
- Added new Devin agent models: Claude 5 Fable variants
- Added new Github Copilot models: Kimi K2.7 Code and MAI-Code-1-Flash
- Added Poolside Laguna XS 2.1 models via Kilo and OpenRouter providers
- Added support for Claude Fable 5 (Free) via Zenmux provider
Changed
- Updated priority ordering to include Baseten
- Updated pricing and limits for various existing models in the catalog
@oh-my-pi/pi-coding-agent
Fixed
- Fixed omp usage hiding sibling-only limits such as Claude 7 Day (Fable) on accounts whose current report omitted that scoped bucket; the account now renders an explicit not reported row instead of looking like the usage refresh skipped the column.
What's Changed
- fix(mnemopi): preserve empty structured extraction by @roboomp in https://github.com/can1357/oh-my-pi/pull/4391
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.3.3...v16.3.4
16.3.3 2 de juliol del 2026 — 64 bits Portable
16.3.2 2 de juliol del 2026 — 64 bits Portable
16.3.0 2 de juliol del 2026 — 64 bits Portable
16.2.13 1 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed pre-5.4 OpenAI Codex models (gpt-5.1-codex, gpt-5.3-codex, gpt-5.3-codex-spark) rejecting requests with Unsupported parameter: 'reasoning.summary' is not supported with this model by gating reasoning.summary behind the same gpt-5.4 wire floor as reasoning.context: "all_turns".
@oh-my-pi/pi-catalog
Added
- Added support for human-readable reasoning summaries on compatible OpenAI Codex models (v5.4+)
Fixed
- Fixed discovered OpenAI Codex models to advertise V2 streaming remote compaction, avoiding the legacy compact endpoint timeout path for Codex sessions. (#4146)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed models.yml remote compaction schema support for V2 streaming endpoint fields. (#4146)
- Fixed the SSH tool to reject cwd values of ~ and ~/... before sending guaranteed-bad quoted tilde paths to remote POSIX shells. (#4002)
@oh-my-pi/pi-tui
Fixed
- Fixed fuzzy-search filtering for CJK and other non-ASCII queries by preserving Unicode letters and numbers during query normalization (#4114).
What's Changed
- fix(tool): reject ssh tilde cwd by @roboomp in https://github.com/can1357/oh-my-pi/pull/4003
- fix(tui): preserve unicode fuzzy search terms by @roboomp in https://github.com/can1357/oh-my-pi/pull/4119
- fix(catalog): enable codex v2 remote compaction by @roboomp in https://github.com/can1357/oh-my-pi/pull/4147
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.2.12...v16.2.13
16.2.12 1 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Improved streaming performance for Cursor and Devin providers by optimizing mid-stream tool-call argument parsing to prevent UI stalls when handling large payloads.
Fixed
- Fixed issues with tool call streaming where tool call IDs, partial JSON payloads, or late-arriving IDs could be lost, filtered, or incorrectly initialized.
- Fixed an issue where stream healing for leaked thinking blocks could replace live tool-call blocks with empty-id placeholders, breaking streamed tool arguments on Anthropic-compatible streams.
- Fixed an issue where stalled auth-gateway SSE responses could hang indefinitely in pi-native streams by ensuring first-event and idle timeout watchdogs are properly honored.
- Fixed cross-turn tool-call loops going undetected by adding a guard for consecutive identical tool calls. (#3971)
@oh-my-pi/pi-catalog
Breaking Changes
- Removed runtime canonical-equivalence APIs from the identity module, including resolveCanonicalVariant, buildCanonicalModelOrder, CanonicalVariantPreferences, and getBundledCanonicalReferenceData. These utilities have been transitioned to a build-time generator script and are no longer exposed in the runtime bundle.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Removed the canonical-alias grouping and resolution layer. The equivalence key (overrides/exclude) in models.yml/models.json is now inert, and canonical-related methods have been removed from ModelRegistry.
- Removed the "CANONICAL" tab from the interactive model selector and the omp models canonical subcommand.
Changed
- Simplified model selection and matching by resolving bare model IDs in --model, modelRoles, and enabledModels via exact/flat-ID and provider-preference matching instead of cross-spelling canonical coalescing.
- Improved cold start performance by removing the catalog-wide canonical index build from the startup path.
Fixed
- Fixed the write tool not streaming execution progress to the TUI while files are being written.
- Fixed live tool-call argument previews disappearing while arguments stream.
- Fixed the LSP tool ignoring timeouts and abort signals during cold-starts and notification writes.
- Fixed the browser tool leaking Chromium/Puppeteer processes when operations are aborted or when an agent session is disposed.
- Added a configurable 30-second timeout (extensionHandlerTimeoutMs) to extension tool call handlers to prevent hung third-party extensions from blocking execution, and fixed a timer leak keeping the CLI alive.
- Fixed the built-in fd tool ignoring shell cancellation (such as Ctrl-C or timeouts) during directory walks.
- Fixed MCP stdio requests hanging indefinitely past their configured timeouts when the child subprocess stops draining stdin.
- Fixed Python eval shell helpers buffering child-process output by streaming chunks and truncating oversized output.
- Fixed cached model edit variants failing to update when changing project directories.
- Fixed subagents with structured output schemas failing validation by correctly wrapping the schema in the system prompt.
- Fixed MCP Streamable HTTP request and notification timeouts staying unarmed during stalled response body reads.
- Fixed evaluation and task spawn defaults to respect restricted agent spawn lists.
- Fixed timed-out browser.run calls leaving evaluated JavaScript continuations running.
- Fixed performance degradation in session context and branch path reconstruction on deep linear histories.
- Fixed agents repeating the same tool call across turns without corrective steering by wiring the cross-turn tool-call loop guard into sessions.
- Fixed OpenAI-compatible model discovery (including LM Studio) reporting flat default context windows when proxies omit context length metadata, by resolving discovered IDs against the bundled model reference catalog to inherit accurate context windows, output limits, display names, modalities, and reasoning support.
@oh-my-pi/pi-tui
Fixed
- Optimized streaming markdown rendering to
16.2.11 1 de juliol del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added streaming support for keyed parameter argument deltas in XML-family in-band tool call scanners (Anthropic, DeepSeek, XML, Minimax)
Changed
- Improved native tool-call passthrough in wrapInbandToolStream to accurately mirror live streaming IDs, arguments, and partial JSON states from the underlying provider
Fixed
- Fixed streaming UI glitches and resolved an issue where invalid empty tool call IDs were persisted in the chat history.
- Fixed a bug where tool calls with empty or missing IDs were not detected as malformed, causing API validation failures (e.g., 400 errors with Anthropic) on subsequent requests
- Raised Gemini header runaway threshold to prevent premature interruption of complex reasoning loops
- Fixed leaked ```thinking fences with nested language-tagged Markdown code blocks so inner fences remain inside structured thinking instead of leaking as visible reply text.
@oh-my-pi/pi-catalog
Added
- Added Claude Sonnet 3.7, Claude Opus 3, and Claude Sonnet 3 model entries to the Anthropic catalog
- Added Anthropic Claude Sonnet 5 model entry to the Kilo provider catalog
- Added first-party catalog discovery support for the Anthropic provider
- Added Gemini 3.1 Flash Lite Image model entry to the Kilo provider catalog
- Added Anthropic Claude Sonnet 5 model variants with low, medium, high, xhigh, and max thinking efforts to the Devin provider catalog
- Added Claude Sonnet 5 model entry to the Anthropic curated catalog.
Changed
- Updated the base API URL for the Claude Sonnet 5 model in the Anthropic catalog
- Updated pricing metrics for DeepSeek R1 and DeepSeek V3 model entries to reflect new rates
Fixed
- Fixed a potential memory leak caused by dangling timeout timers during model discovery in OpenAI-compatible, vLLM, LiteLLM, and LM Studio catalogs.
- Widened stream watchdogs for local OpenAI-compatible backends (including llama.cpp, LM Studio, vLLM, and Ollama) to prevent premature timeouts during cold model loads.
@oh-my-pi/pi-coding-agent
Changed
- Updated grep warnings to clarify that large files are partially searched rather than entirely skipped
- Renamed the filesystem walker worker count environment variable from PI_GREP_WORKERS to PI_WALK_WORKERS
- Centralized filesystem traversal policy in pi-walker.
- Changed the in-session /resume session picker to open as a fullscreen window on the terminal's alternate screen, matching the startup --resume picker and /settings. It borrows the alt buffer for its lifetime (the transcript is untouched underneath) and enables mouse tracking — the wheel scrolls the list and a left click resumes the row under the pointer — with the keybinding hint and bottom border pinned to the screen bottom. Previously it mounted inline in the editor slot and rendered compactly without mouse support.
Fixed
- Fixed model-discovery requests (Ollama, Llama.cpp, LM Studio, OpenAI, LiteLLM, vLLM) failing to clear timeouts after completion, preventing potential memory and timer leaks.
- Fixed grep and ripgrep built-in tools ignoring shell abort and timeout signals during recursive directory walks, allowing them to be interrupted immediately.
- Fixed omp setup speech returning prematurely before Whisper STT downloads complete, and improved error reporting for worker download failures.
- Fixed high memory usage in multi-target ast_grep searches by only retaining the final page window while preserving exact match and file counts.
- Fixed async job manager disposal and task cancellation handling to properly release session semaphores when aborted.
- Updated and expanded omp:// documentation coverage for managed memory, skill tools, image/speech generation, and package CLIs.
- Fixed Git subcommands ignoring repository paths by stripping ambient Git environment variables
- Fixed concurrent bash commands cross-killing each other on cancel/timeout. Cancellation cleanup previously walked the whole host process tree and signalled every descendant spawned since a per-run baseline,
16.2.9 30 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added OAuthCallbackFlowOptions.allowPortFallback to allow disabling random-port fallback, enabling strict port enforcement and early configuration errors for OAuth flows with static redirect URIs.
Changed
- Improved OAuthCallbackFlow port conflict error messages to include the busy port, configured redirect URI, and actionable remediation steps.
Fixed
- Fixed an issue where malformed tool-call JSON from local Ollama or llama.cpp models was incorrectly retried as generic 500 errors, now surfacing a clear recovery message.
- Fixed a race condition in OAuth callback flows where abort signals triggered before the callback listener was registered were ignored.
@oh-my-pi/pi-catalog
Added
- Added full capability support for Claude Sonnet 5, aligning it with Claude Opus 4.8 and Fable 5. This includes adaptive thinking display, mid-conversation system messages, sampling parameter and thinking omission API restrictions, and 5-tier adaptive reasoning effort mapping (including xhigh and max levels) across direct APIs, OpenRouter, and Bedrock Converse.
Changed
- Updated input and output costs for models in the catalog.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Renamed the built-in quick_task subagent to sonic; update any task spawns or configurations referencing quick_task by name.
Added
- Added the llama3.2:3b local model option for memory and auto-thinking tasks, utilizing a quantized ONNX model.
- Added a built-in Tester subagent designed to write high-signal tests for behavior, invariants, and edge cases while avoiding redundant or low-value tests.
- Added a Speech-to-Text submit trigger setting to auto-submit dictation on release, on complete sentences, or via a spoken submit command.
- Added a loop-guard mechanism that detects thinking/response loops and injects a system notice during auto-retries to guide the model to break the pattern and take a concrete next step.
Changed
- Renamed the Speech-to-Text (STT) setting label from "TTS Submit Trigger" to "Speech-to-Text Submit Trigger".
Fixed
- Fixed an issue where mid-run compaction was incorrectly skipped when a persisted assistant display variant shared a persistence key but differed in content from the live message.
- Fixed duplicate placeholder cards being created when streamed tool blocks started with an empty ID.
- Fixed omp debug --profile failing on Bun by treating the optional --allow-natives-syntax flag as best-effort when v8.setFlagsFromString is unavailable.
- Fixed release binaries missing the compiled tiny-model Transformers.js version pin, preventing runtime resolution issues on certain platforms like Homebrew Darwin arm64.
- Fixed MCP OAuth flows silently falling back to a random redirect port when the preferred port (default 3000) was busy, which caused authentication failures with strict providers. The flow now fails fast with a configuration error when a static client ID is pinned, while dynamic registration flows continue to use fallback ports.
- Fixed /mcp reauth and /mcp add commands ignoring the Escape key during OAuth authentication, allowing users to cancel the flow immediately instead of waiting for the timeout.
Removed
- Removed the built-in oracle subagent.
@oh-my-pi/pi-tui
Added
- Added Editor.submit() to allow programmatic composer submission, enabling integration with speech input and other automated flows.
@oh-my-pi/pi-utils
Added
- Improved resilience in fetchWithRetry() by adding a response-body retry gate to handle deterministic provider failures that return retryable HTTP statuses.
Fixed
- Fixed YAML frontmatter parsing for skill descriptions containing unquoted colons (: ), ensuring typed fields are correctly preserved without triggering unnecessary warnings.
What's Changed
- fix(mcp/oauth): fail fast when callback port is busy instead of advertising a random one by @roboomp in https://github.com/can1357/oh-my-pi/pull/3889
- fix(mcp): made Esc cancel /mcp reauth and /mcp add OAuth flow by @roboomp in https://github.com/can
16.2.8 30 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Added built-in Go coding rules including go-add-cleanup, go-bench-loop, go-exp-promoted, go-ioutil, go-join-hostport, go-new-expr, go-rand-v2, and go-range-int
Changed
- Relaxed strict bash tool constraints regarding the use of search, grep, ls, and find commands
Fixed
- Fixed auto-compaction dead-ends by automatically triggering a shake rescue to elide oversized tails
- Improved compaction warning message to suggest running /shake images for irreducible image tails
- Fixed grep/search direct execution to accept JSON-array string paths for string-or-array inputs. (#3873)
- Fixed auto-compaction dead-ending with "Compaction freed too little context to make progress" when a single recent turn (large tool output, heavy fenced/XML block) is itself bigger than the recovery band — findCutPoint can't cut inside one message, so the summarizer had no lever left. The guard now runs an artifact-backed shake elide pass over the oversized tail and re-tests headroom before pausing, and the remaining warning points at /shake images for image-only tails it can't elide. (#3786)
- Fixed reviewer/task subagents whose incremental yield (type: ["overall_correctness"], type: ["findings"], …) carried a value that mismatched the matching property's sub-schema being silently accepted and then post-mortem rejected with schema_violation — opaquely swapping the agent's accepted output for an error blob. The yield tool now validates each incremental section's data against its top-level property's sub-schema (items schema for array-typed labels) and surfaces the same retry feedback as terminal yields, so models like deepseek-v4-pro that emit "Correct"/"correct."/"approved" for an enum field get up to three corrective retries; the existing MAX_SCHEMA_RETRIES override then accepts the value with SUBAGENT_WARNING_SCHEMA_OVERRIDDEN instead of losing the entire result. Unknown labels stay unconstrained (#3870).
- Fixed streaming tool-call previews (notably write) showing an empty body for the entire streaming phase by surfacing the partial JSON already in hand on the first reveal, then pacing only subsequent growth (#3881).
- Fixed hashline edit mode preserving UTF-8 BOM bytes on edited files. (#3867)
- Fixed slow local LLM streams by forwarding persisted stream timeout settings (providers.streamFirstEventTimeoutSeconds, providers.streamIdleTimeoutSeconds) into model requests, so users can widen or disable watchdogs without environment variables. (#3878)
- Fixed the bash interceptor blocking echo / printf redirects to /dev/null, /dev/tty, /dev/stdout, and /dev/stderr device sinks while still directing real file writes to the write tool. (#3763)
- Fixed long snapcompact sessions re-sending multi-megabyte standing image archives on every provider request by enforcing a per-request frame byte budget, letting auto-compaction fall back to context-full summaries when snapcompact output is too large, and omitting legacy over-budget frames from rebuilt LLM contexts. (#3792)
@oh-my-pi/hashline
Fixed
- Fixed hashline writes preserving UTF-8 BOM bytes when the host text decoder hides the leading U+FEFF. (#3867)
@oh-my-pi/snapcompact
Fixed
- Fixed large snapcompact archives being reconstructed into unbounded per-request image payloads by adding a frame base64 byte budget and omitting over-budget archive frames from prompt blocks. (#3792)
What's Changed
- fix(compaction): cap snapcompact frame payloads by @roboomp in https://github.com/can1357/oh-my-pi/pull/3866
- fix(yield): validate incremental sections per-label to prevent fatal schema_violation by @roboomp in https://github.com/can1357/oh-my-pi/pull/3872
- fix(tools): accept json-array grep paths by @roboomp in https://github.com/can1357/oh-my-pi/pull/3874
- fix(coding-agent): forward stream timeout settings by @roboomp in https://github.com/can1357/oh-my-pi/pull/3879
- fix(coding-agent): seed tool-args reveal with the partial JSON already in hand by @roboomp in https://github.com/can13
16.2.7 30 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added service tier support for Google Gemini and Vertex AI, including model-specific service tier configurations via ServiceTierByFamily.
- Added Google Vertex AI Interactions API support for Gemini 3+ models by default, with automatic fallback to 😒treamGenerateContent and a useInteractionsApi: false option to force standard generation.
- Added support for explicit Vertex bearer access tokens via GOOGLE_CLOUD_ACCESS_TOKEN or CLOUDSDK_AUTH_ACCESS_TOKEN environment variables.
Changed
- Updated service tier logic to use per-provider configurations instead of global scopes.
- Refactored priority request billing and accounting to better align with specific provider capabilities.
- Updated API key resolution precedence so explicit environment variables (e.g., GEMINI_API_KEY) override stored or broker-migrated static API keys, while deliberate OAuth logins still take highest precedence.
Fixed
- Improved Vertex AI reliability by automatically falling back to global endpoints on 404 errors.
- Fixed safety setting application for Google Vertex AI models.
- Fixed Kimi Code's Anthropic-compatible request path to keep thinking enabled and downgrade forced tool choice for Kimi K2.7 Code title generation.
- Fixed leaked reasoning fences (such as ```thinking or ) across all providers by splitting them into structured thinking blocks during streaming.
- Fixed Codex requests failing with unsupported all_turns errors on older models (gpt-5.1 and gpt-5.3) by gating the reasoning.context: "all_turns" default to gpt-5.4+ models.
@oh-my-pi/pi-catalog
Fixed
- Fixed compatibility with Kimi K2.7 Code on native endpoints to ensure thinking mode is preserved and tool choice is not forced.
- Fixed Cerebras gemma-4-31b dynamic discovery to correctly identify the model as image-capable, enabling proper serialization of attached images.
@oh-my-pi/pi-coding-agent
Breaking Changes
- Replaced the global serviceTier and fastModeScope settings with granular, per-family settings (tier.openai, tier.anthropic, and tier.google) to control service tiers, subagents, advisors, and /fast mode targets.
Changed
- Improved binary file detection and terminal handling to prevent corruption from non-UTF-8 content, and updated file summaries to explicitly note skipped binary files.
- Enhanced context compaction (snapcompact) to resolve shapes contextually based on rendered text content.
Fixed
- Improved reliability of DuckDuckGo web searches by updating browser request headers and parameters
- Fixed an issue where CJK (Chinese, Japanese, Korean) history could become unrenderable during repeated context compactions.
- Fixed a memory exhaustion bug in the TUI when using /resume on large previous sessions.
- Fixed an issue where the irc inbox missed messages that arrived while the recipient agent was already running.
- Fixed a startup hang caused by system-prompt GPU detection blocking and repeatedly running failed probes.
- Improved error reporting for omp tiny-models download by displaying the actual worker-side download error.
- Resolved status inconsistencies between /extensions, /mcp list, and the dashboard, ensuring MCP server states, allowlists/denylists, and configuration files (like mcp.json) stay fully synchronized.
- Improved branch-mode task merges to preserve the agent's original commit history (messages and authors) and fixed a bug where merges were rejected due to unrelated dirty changes in the parent checkout.
- Fixed an issue where the Working... loader spinner would prematurely disappear or fail to re-arm after a subagent (task) tool completed or during transient overlays (such as auto-compaction or auto-retry).
@oh-my-pi/pi-natives
Added
- Added embedded Silver TrueType font rendering support to renderSnapcompactPng, featuring automatic per-glyph fallback for missing bitmap characters and anti-aliased scaling for East Asian wide code points.
- Added the snapcompactSupportedChars function to check font capability for specific characters.
16.2.6 29 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Antigravity usage reporting to correctly infer daily and weekly quota windows from unlabeled reset-only rows, preventing Cloud Code Assist payloads from collapsing these counters into the default category.
@oh-my-pi/pi-catalog
Fixed
- Fixed namespaced GLM-5.x model IDs on Z.AI/Zhipu OpenAI-compatible endpoints to inherit the widened stream watchdog, avoiding spurious stalled-stream errors during long thinking phases. (#3819)
@oh-my-pi/pi-coding-agent
Changed
- Optimized argument streaming performance by throttling JSON re-parsing for renderers that do not require raw input.
Fixed
- Fixed memory leaks in the benchmark CLI by ensuring provider sessions are properly closed upon completion.
- Fixed TUI rendering lag and shimmer frame starvation during streamed tool-call argument previews.
- Fixed llama.cpp discovery mapping unlimited output limits (max_tokens = -1 / n_predict = -1) to the generic 32K cap instead of the actual discovered runtime context window.
- Fixed memory exhaustion on Ctrl+C flush for large resumed sessions with repeated compaction summaries, and improved fork session previews starting with developer or assistant turns.
- Fixed omp search applying web search provider filters before resolving its implicit provider chain.
- Fixed Gemini web search to support standard GEMINI_API_KEY developer API credentials for native Google Search grounding, in addition to Cloud Code Assist OAuth.
- Fixed the bash interceptor blocking echo and printf redirects to standard device sinks (such as /dev/null, /dev/tty, /dev/stdout, and /dev/stderr).
- Fixed session file size inflation in the edit tool by pruning extremely large oldText and newText snapshots (over 32 KB) from tool-result details, while preserving visible diffs, paths, lines, and diagnostic metadata.
- Fixed Windows MCP stdio launches for PATH-resolved npx.cmd shims by preserving the cmd.exe wrapper path to keep subprocess stdio attached.
- Fixed the DuckDuckGo web_search provider returning empty results for non-encyclopedic queries by switching from the Instant Answer API to parsing the HTML frontend, and added clear error handling for bot-challenge throttling.
- Fixed Windows --extension paths with spaces or \?\ prefixes being truncated or incorrectly passed to Bun import/spawn APIs.
- Fixed /mcp reauth compatibility with Cloudflare by aligning OAuth prompt behavior with the reference MCP SDK and updating the client label to oh-my-pi.
@oh-my-pi/hashline
Fixed
- Fixed a parser error ("payload line has no preceding hunk header") caused by stray dots before the trailing colon in hunk headers, improving compatibility with GLM 5.2 outputs.
@oh-my-pi/omp-stats
Fixed
- Fixed application crashes and Bun aborts on macOS and when parsing large stats session files, including during omp --smoke-test runs, by utilizing a more resilient serial parser and lenient line scanner.
@oh-my-pi/pi-utils
Added
- Added stripWindowsExtendedLengthPathPrefix() utility to normalize \\?\ and native Win32 path prefixes before Bun import or spawn calls.
What's Changed
- fix(providers): honor llama.cpp unlimited output cap by @roboomp in https://github.com/can1357/oh-my-pi/pull/3782
- fix(session): avoid ctrl-c rewrites for compacted sessions by @roboomp in https://github.com/can1357/oh-my-pi/pull/3790
- fix(mcp): preserve npx cmd shim launching by @roboomp in https://github.com/can1357/oh-my-pi/pull/3796
- fix(ai): infer Antigravity reset windows by @roboomp in https://github.com/can1357/oh-my-pi/pull/3798
- fix(cli): preserve Windows extension paths by @roboomp in https://github.com/can1357/oh-my-pi/pull/3806
- fix(providers): support Gemini API key web search by @roboomp in https://github.com/can1357/oh-my-pi/pull/3811
- fix(mcp/oauth): force login screen before consent on /mcp reauth by @roboomp in https://github.com/can1357/oh-my-pi/pull/3818
- fix(stats): avoid macOS stats sync worker abort by @roboomp in https://github.com/can1357/oh-my-pi/pull/3735
- fix(coding
16.2.5 29 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Google and Cloud Code Assist streams that end without a finish reason (dropped connections or truncated responses) being treated as fatal; they are now classified as transient so the coding agent automatically retries.
@oh-my-pi/pi-coding-agent
Changed
- Status line now collapses a linked git worktree path to the project name with a worktree icon, leaving the git segment to show the branch once instead of repeating it in the path.
Fixed
- Fixed Ollama and llama.cpp discovery overestimating local context windows by using model training metadata instead of the runtime num_ctx / n_ctx, which could let compaction retry prompts that llama.cpp rejects as over context. (#3752)
- Fixed isolated subagent worktrees embedding long task ids and merged in the working path; isolation now uses compact hashed segments with an m mount dir. (#3756)
- Fixed recoverable context-overflow compaction keeping the failed assistant error turn in visible session history after scheduling the retry. (#3747)
- Fixed editing a file read from outside the workspace (e.g. ~/.claude/settings.json) failing with "File not found": the read snapshot header now carries the full out-of-workspace path so the edit resolves it directly instead of against the working directory.
- Fixed subagents spawned with an output schema (agent(..., schema=...), task with structured output) failing with schema_violation: missing required fields since the typed-yield rework: a type: "result" finalize carrying the full object was assembled as a section named result, nesting the payload one level deep. String-typed yields are now treated as terminal finalizers (their data is the complete result), only array-typed yields form accumulating sections, and a data-less finalize keeps accumulated sections instead of collapsing to the last assistant turn.
- Fixed the per-provider concurrency cap (e.g. providers.ollama-cloud.maxConcurrency) bracketing the whole subagent lifecycle, which deadlocked any spawn tree wider than the cap because parents held every slot while waiting for children queued on the same cap. The semaphore now wraps each provider HTTP request, so a parent's slot frees between turns and child subagents can acquire while their parent's tool calls are running. (#3749)
- Fixed Ctrl+Q / Ctrl+Enter follow-up submissions sending the literal [Paste #N] marker instead of the expanded paste body (#3737).
@oh-my-pi/pi-natives
Fixed
- Fixed the in-process grep builtin rejecting GNU-grep's --color/--colour (with or without =WHEN) and --version flags. The shadowing rejection broke bash's near-universal alias grep='grep --color=auto', causing bare grep in any pipeline to fail with exit 2. The builtin now accepts and ignores --color[=WHEN] (its output goes through in-process file descriptors, never a TTY, so ANSI injection would corrupt downstream consumers) and reports its version through the context streams (#3755).
What's Changed
- fix(coding-agent): expand paste markers in Ctrl+Q follow-ups by @roboomp in https://github.com/can1357/oh-my-pi/pull/3738
- fix(session): drop recovered overflow error turns by @roboomp in https://github.com/can1357/oh-my-pi/pull/3748
- fix(task): scoped provider concurrency cap to each LLM turn by @roboomp in https://github.com/can1357/oh-my-pi/pull/3751
- fix(providers): honor local runtime context limits by @roboomp in https://github.com/can1357/oh-my-pi/pull/3754
- fix(grep-builtin): accept --color/--colour/--version by @roboomp in https://github.com/can1357/oh-my-pi/pull/3757
- fix(agent): shorten isolated task worktree paths by @roboomp in https://github.com/can1357/oh-my-pi/pull/3758
- fix(tool): allow bash redirects to dev sinks by @roboomp in https://github.com/can1357/oh-my-pi/pull/3764
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.2.4...v16.2.5
16.2.4 28 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- Improved the reliability of remote compaction by introducing transient error retries, configurable timeouts, and immediate termination upon user-initiated aborts.
Fixed
- Fixed an issue where assistant responses and encrypted reasoning could be lost during local history trimming prior to remote compaction.
- Fixed type compatibility for hosts with title audit entries by adding support for title_change session metadata.
- Fixed an issue where transient stream read failures after a completed tool call were treated as terminal errors, allowing the agent to successfully execute the tool and continue the turn.
@oh-my-pi/pi-ai
Added
- Enabled freeform tool patch support for Azure OpenAI and Codex models
Fixed
- Fixed usage reporting for Antigravity and Z.AI to correctly surface and preserve distinct quota windows (daily, weekly, monthly) instead of collapsing or duplicating them
- Fixed an issue where /usage show returned "No usage data available" when using a custom proxy base URL for Codex
- Fixed OpenAI stream read errors being incorrectly classified as non-transient, enabling the coding agent to automatically retry after recoverable stream failures
@oh-my-pi/pi-coding-agent
Fixed
- Fixed todo-reminder HUD rendering outside durable chat history while preserving native collapsing and auto-clear behavior.
- Fixed goal-mode continuations losing track of persisted todo progress, ensuring autonomous goal turns reconcile stale in-progress items before moving to subsequent tasks.
- Fixed the /move command to correctly relocate the current session and its artifacts to the target directory, allowing /resume to work seamlessly from the new location.
- Fixed omp update leaving behind stale Bun install-cache directories for globally installed packages.
- Fixed a reconnection loop issue in /collab sessions caused by oversized entries (such as large tool outputs) by truncating replicated payloads that exceed 1 MB.
- Fixed autolearn and local memory writes mutating Anthropic prompt-cache prefixes mid-session, ensuring prompt injections remain session-stable.
@oh-my-pi/pi-natives
Fixed
- Fixed a crash in the in-process tail builtin where the host process would abort with a BrokenPipe panic if the stdout consumer closed the pipe early.
What's Changed
- test(ai): add contract tests for resolveUsedFraction by @oldschoola in https://github.com/can1357/oh-my-pi/pull/3326
- test(eval): increase Julia prelude test timeout to fix CI flake by @oldschoola in https://github.com/can1357/oh-my-pi/pull/3350
- fix(omp): keep moved sessions resumable by @riverpilot in https://github.com/can1357/oh-my-pi/pull/3675
- fix(omp): show Antigravity weekly usage limits by @riverpilot in https://github.com/can1357/oh-my-pi/pull/3678
- Recover completed tool calls after transient stream read errors by @cexll in https://github.com/can1357/oh-my-pi/pull/3694
- fix(collab): cap replicated host frames so oversized entries do not loop the relay by @roboomp in https://github.com/can1357/oh-my-pi/pull/3740
- fix(agent): keep autolearn memory prompts cache-stable by @roboomp in https://github.com/can1357/oh-my-pi/pull/3745
- fix(coding-agent): prune stale bun update cache by @dragonbaba in https://github.com/can1357/oh-my-pi/pull/1716
New Contributors
- @dragonbaba made their first contribution in https://github.com/can1357/oh-my-pi/pull/1716
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.2.3...v16.2.4
16.2.3 28 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- Enabled V2 streaming remote compaction by default for compatible AI and OpenAI-compatible models, which forwards full conversation history to the provider and supports session routing, prompt caching, provider-native tool history replay, transient error retries, and configurable timeouts.
Fixed
- Fixed an issue where assistant responses and encrypted reasoning could be lost during local history trimming.
- Added title_change session metadata to the compaction entry type union to maintain type compatibility for hosts with title audit entries.
@oh-my-pi/pi-ai
Changed
- Enabled automatic removal of leaked reasoning tags for all models
- Prevented reasoning text duplication when models emit both structured and inline thinking
- Defaulted reasoning context to all turns for all Codex requests.
Fixed
- Enabled freeform tool patch support for Azure OpenAI and Codex models.
- Fixed an issue where the /usage show command returned "No usage data available" when using a custom proxy base URL for Codex.
@oh-my-pi/pi-catalog
Added
- Added support and configuration parameters for V2 streaming compaction in RemoteCompactionConfig, catalog types, and model/provider metadata.
Changed
- Enabled automatic content markup healing for all OpenAI-compatible streaming models
- Updated pricing and context window limits for several catalog models.
- Disabled reasoning capability for multiple providers in the catalog.
@oh-my-pi/pi-coding-agent
Added
- Added support for multiple configurable advisors via WATCHDOG.yml/WATCHDOG.yaml files, allowing per-advisor models, tool subsets, and instructions.
- Added /advisor configure, a full-screen, mouse-driven TUI to easily manage the advisor roster, configure models, toggle tool permissions, and edit instructions.
- Added full unified edit diffs to advisor transcripts, allowing advisors to see changes directly without re-reading files.
- Added the statusLine.compactThinkingLevel setting to render the model segment's thinking level as a single leading glyph instead of a separate text suffix.
- Added support for tracking reasoning tokens in session and advisor statistics.
- Added Remote Compaction V2 streaming configuration settings (compaction.remoteStreamingV2Enabled and compaction.v2RetainedMessageBudget) to control token budgets and toggle V2 streaming for remote compaction.
- Added the edit.citationTags setting to emit model-facing hashline section headers as OpenAI citation markers with opaque source IDs, along with citation-marker unwrapping for hashline edit parsing, diff previews, and streaming matching.
- Added mutable session titles with automatic replan title refreshes and configurable idle recaps.
- Added support for incremental yield submissions with typed sections and final results for subagents.
Changed
- Reduced session file size by removing redundant thinking signatures already present in payloads
- Advisors can now be granted any built-in agent tool (including edit, write, and bash), removing the previous read-only restriction.
- Improved the debug log and raw SSE stream viewers with a wider, bordered overlay, clearer status indicators, dynamic layouts, and mouse support for scrolling and interaction.
- Updated the idle recap feature to use an LLM-generated summary of where things stand (anchored by the live goal and active todo task) instead of a static status line.
- Refined interrupted thinking system instructions to encourage smoother continuation.
Fixed
- Fixed array-typed output schema validation by correctly assembling incremental yields into lists.
- Fixed OpenAI/Codex compatibility by removing top-level schema combinators from tool parameters.
- Fixed validation errors for untyped final yields in strict-mode providers by allowing null types.
- Fixed Alt+M default-role model configuration being disabled by the current session's context size.
- Fixed MCP type: "sse" servers by adding the legacy HTTP+SSE endpoint handshake and streaming JSON-RPC response p
16.2.2 27 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added optional AgentTool.matcherPaths(args) and AgentTool.matcherEntries(args) hooks to allow tools to surface target file paths and isolate file evaluations for path-scoped stream matchers (e.g., when handling multi-file payloads or embedded paths in streamed arguments).
Removed
- Removed support for Pi dialect integration.
@oh-my-pi/pi-ai
Added
- Added a comprehensive, public-facing error module exported via the "./error" path, featuring structured error classification, provider-specific HTTP error classes (e.g., Anthropic, OpenAI, Gemini), OAuth/Auth-specific errors, rate-limit utilities, and retryability predicates.
Changed
- Updated OpenAI Codex defaults to increase default text verbosity to medium, enable detailed reasoning summaries by default, and include all turns in the reasoning context by default.
- Updated the OpenAI Codex WebSocket transport to resolve its configuration (via PI_CODEX_WEBSOCKET_* environment variables) once at startup rather than re-parsing on every request.
- Enhanced cross-model reasoning recovery and preservation to render demoted reasoning in the target model's canonical inline thinking dialect (such as Gemini's thinking fence or standard think tags) to prevent leaking inert context or control tokens into history.
- Broadened the leaked-thinking stream healer to recover reasoning emitted in any dialect's canonical idiom (including Gemini, Gemma, Harmony, and scratchpads) and route them to thinking events instead of raw markup.
- Implemented automatic retry logic for detected thinking-loop stalls to improve response reliability.
- Hardened stateful delta chaining to ignore transient streaming bookkeeping symbols during structural equality checks, preventing unnecessary full-transcript replays.
Fixed
- Fixed preservation of OpenAI Responses assistant message phase values across auth-gateway parsing, streaming, and history replay, ensuring GPT-5.4/GPT-5.5 intermediate updates and final answers retain their original phase labels.
Removed
- Removed Pi dialect support and related serialization/parsing logic.
@oh-my-pi/pi-catalog
Removed
- Removed 'pi' from the list of supported dialects.
@oh-my-pi/pi-coding-agent
Added
- Added a new tiny model role for consolidated online task handling.
- Added a textVerbosity setting to control OpenAI and Codex response detail.
Changed
- Simplified the status line subagent display by removing running state and hub hint indicators.
- Updated online title, memory, and classification tasks to prioritize the new tiny model role.
Fixed
- Improved reliability of auto-retry logic for aborted requests by standardizing error classification across model adapters and ensuring stale session states are reset.
- Enhanced MCP authentication error detection, header-based server discovery, and 401/403 authorization failure detection during Smithery commands and HTTP RPCs.
- Fixed auto-generated session titles incorrectly preserving user all-caps text, ensuring proper sentence casing while still respecting intentional mixed-case identifiers.
- Fixed Tavily web search with recency filters to automatically retry without a time range if Tavily returns an empty HTTP 200 response.
- Fixed TUI thought stream stalling and ui.loop-blocked warnings during subagent-heavy runs by optimizing the mid-run compaction persistence check.
- Fixed marketplace-installed plugins incorrectly appearing in both the npm plugin list and the extension-package status provider.
- Fixed inconsistent OpenRouter prompt-cache hits on /advisor turns by ensuring advisor agents inherit the same provider-shaping options, hooks, and settings as the main agent.
- Fixed path-scoped TTSR (Targeted Tool Safety Rules) evaluation for hashline and apply_patch edit streams, ensuring rules are correctly applied to file paths parsed from section headers and envelope markers without leaking across file scopes.
@oh-my-pi/pi-mnemopi
Fixed
- Improved resilience during API extraction calls by enhancing 
16.2.1 27 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Included project context files (AGENTS.md, etc.) in the advisor's system prompt to ensure adherence to user-defined project rules
- Added project context files (AGENTS.md and the like) to the advisor's system prompt, so the read-only reviewer judges against the user's standing project rules the same way the main agent does.
Fixed
- Fixed live ACP generate_image updates resolving OMP-internal image blob refs before sending renderable image content to clients. (#3623)
- Fixed Claude marketplace plugin .mcp.json MCP servers to expand environment variables in url and headers before connecting. (#3621)
What's Changed
- fix(mcp): expand Claude plugin MCP header env vars by @roboomp in https://github.com/can1357/oh-my-pi/pull/3622
- fix(acp): resolve generated image blob refs in live updates by @roboomp in https://github.com/can1357/oh-my-pi/pull/3625
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.2.0...v16.2.1
16.2.0 27 de juny del 2026 — 64 bits Portable
16.1.23 26 de juny del 2026 — 64 bits Portable
16.1.22 26 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed llama.cpp / LM Studio / vLLM (and any local OpenAI-compatible server on a loopback or RFC1918 baseUrl) re-processing the full prompt on every assistant continuation when the prior turn produced reasoning_content: the openai-completions encoder dropped the preserved thinking block on re-serialization for compat profiles without requiresReasoningContentForToolCalls / thinkingFormat: "zai", so the chat template re-rendered the assistant turn without <think>…</think> and the rendered tokens diverged from the slot's KV cache state. The auto-learn capture-at-stop nudge made it reproduce on every turn. The encoder now replays preserved thinking as reasoning_content (honoring the streamed signature when it identifies a recognized wire field — reasoning_content / reasoning / reasoning_text — and falling back to the configured reasoningContentField for opaque signatures) whenever the new compat.replayReasoningContent flag is set, and the cross-API transformMessages predicate (openAICompletionsReplaysUnsignedThinking) honors the same flag ahead of the model.reasoning gate so a switch into a discovered local target (where the spec carries reasoning: false because the upstream /models endpoints don't advertise the capability) still preserves the prior turn's thinking block as signature-stripped reasoning instead of demoting it to conversation text. The chat-template-rendered prefix stays byte-stable across turns and llama.cpp's prefix KV cache survives. (#3528)
@oh-my-pi/pi-catalog
Added
- Added OpenAICompat.replayReasoningContent — auto-enabled for the built-in local OpenAI-compatible providers (llama.cpp, lm-studio, vllm, ollama on openai-completions) and for any provider pointed at a loopback / RFC1918 / *.local baseUrl. NOT gated on spec.reasoning: the runtime discovery paths for llama.cpp / lm-studio / openai-models-list hardcode reasoning: false because the upstream /models endpoints don't advertise the capability, while the stream parser still records incoming reasoning_content deltas as thinking blocks — gating on the spec flag would leave every discovered local Qwen / DeepSeek model re-triggering #3528. The encoder only writes reasoning_content when a thinking block actually exists on the turn, so the flag is a no-op on pure-text histories. Built-in proxy providers (currently litellm) are excluded from both checks because they forward to an unrelated upstream that gains no KV-cache benefit and may 400 on the extra field; users running a custom proxy in front of a llama.cpp-style backend can opt in via the sparse compat.replayReasoningContent: true override. Signals to the openai-completions encoder that preserved thinking blocks must be re-emitted as reasoning_content on every assistant turn so chat templates that reconstruct <think>…</think> from the field (Qwen3, DeepSeek-R1, GLM-5.x) keep the prior turn's tokens byte-stable and llama.cpp's prefix KV cache survives. (#3528)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Windows stdio MCP server launches showing a separate cmd.exe window for direct executable servers; MCP subprocesses now set windowsHide on every Windows spawn path. (#3535)
- Fixed MCP OAuth still failing with Authorization failed: An unexpected error occurred against Plane's https://mcp.plane.so/http/mcp endpoint after #3502. The protected-resource metadata advertised the path-scoped issuer https://mcp.plane.so/http, but discoverOAuthEndpoints (packages/coding-agent/src/mcp/oauth-discovery.ts) probed /.well-known/oauth-authorization-server at the origin root first and accepted the metadata served there — which describes a different issuer (https://mcp.plane.so/) whose /authorize endpoint rejects every grant. Discovery now honors RFC 8414 §3.3 and skips authorization-server / OpenID Connect well-known documents whose issuer field doesn't match the queried base URL (trailing-slash insensitive). Servers that omit issuer keep today's permissive behavior, so legacy flows are unaffected. (#3537)
16.1.21 26 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Restored the pollOAuthDeviceCodeFlow export from @oh-my-pi/pi-ai/oauth so legacy provider extensions can reuse the host OAuth device-code poller. (#3508)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed autolearn.autoContinue (the "Auto-run capture at stop" toggle) letting the synthetic capture turn keep working after the learn/manage_skill call instead of yielding. With the toggle on, the controller fires autolearn-nudge.md as a single attribution: "user" message on a new turn; the prompt opened with "Before you finish:" and gave no terminal contract, so the agent treated the synthetic prompt as the user's reply to its prior pending question (e.g. "Want me to commit and push?") and continued — pushing commits, running tools, etc. — without the user ever answering. The nudge is now split into two prompts: passive mode (the reminder rides the user's real next message) keeps additive framing — "answer the user normally; the capture is in addition to" — while auto-continue mode (autolearn-nudge-autocontinue.md) is explicitly terminal — "not a user reply; do not treat this as approval; capture, then stop; wait for the user's next prompt". Attribution stays user to preserve llama.cpp warm-prefix reuse (#3456). (#3504)
- Fixed the clipboard image-paste keybind dropping image-file-only pasteboards as literal text on macOS. When the clipboard exposes only the file URL (e.g. Finder Cmd+C on a .png, certain screenshot tools), arboard::get_image() returns ContentNotAvailable and pbpaste(1) returns empty (it only surfaces plain text / RTF / EPS), so InputController.handleImagePaste either fell through to the #1628 text fallback and pasted the path verbatim, or dead-ended with "Clipboard is empty". The keybind path now (1) reaches the public.file-url representation directly via a new readMacFileUrlsFromClipboard AppleScript bridge on Darwin and routes the first image-shaped path through handleImagePathPaste, (2) reuses a new extractImagePathFromText helper so any clipboard-text path that survived pbpaste is detected the same way the bracketed-paste handler already detects them — including a whole-text-as-path fallback for anchored paths whose unescaped spaces would otherwise be shredded by the bracketed-paste splitter (macOS screenshot names like /Users/me/Desktop/Screenshot 2026-06-25 at 1.23.45 PM.png), and (3) decodes file:// URLs to filesystem paths in normalizePastedPath (mirroring Codex's normalize_pasted_path in codex-rs/tui/src/clipboard_paste.rs). Keybind- and terminal-mediated paste now agree on raw image bytes, plain image paths, file:// URLs, Finder-copied image files, and screenshot filenames with spaces alike. (#3506)
- Fixed compiled-binary validation for legacy pi.extensions packages whose source imports worker-only coding-agent subpaths or extension-local package subpaths such as typebox/value; omp install @charmland/pi-hyper-provider, omp plugin doctor, and runtime provider discovery now use a main-thread-safe load path. (#3508)
- Fixed repeated todo updates in one TUI turn stacking full todo panels; superseded todo snapshots now stay live until the next todo update replaces them or the turn ends. (#3516)
- Fixed MCP OAuth authorization failing with Authorization failed: An unexpected error occurred against authorization servers (Plane is the live example) that reject redundant fallback resource indicators. OMP now drops same-origin resources only when it synthesized them from the server URL fallback (e.g. https://mcp.plane.so/http/mcp). Provider-advertised resources from OAuth/protected-resource discovery or an embedded authorization-URL resource query parameter are preserved even when they are same-origin or origin-only, so gateway-hosted MCP services can still request the audience they advertised. The refresh-token path uses the same policy, filtered against the authorization-server origin persisted on the credential as authorizationUrl, with tokenUrl's origin as the legacy fallback when that field is ab
16.1.20 25 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Ollama/Ollama Cloud native chat responses that finish with done_reason: "length" and no assistant content surfacing as a normal empty stop; they now become a context-window error instead of entering empty-stop retry recovery. (#3464)
- Fixed direct Anthropic Claude Sonnet/Haiku 4.5 requests serializing output_config.effort. The catalog classification (packages/catalog/src/model-thinking.ts) drove the anthropic-budget-effort branch in buildParams, which Anthropic's first-party Messages API rejects on Sonnet/Haiku 4.5 with HTTP 400 This model does not support the effort parameter. Sonnet/Haiku 4.5 now use plain thinking.budget_tokens; Opus 4.5 still emits output_config.effort because Anthropic supports it there. (#3497)
@oh-my-pi/pi-catalog
Fixed
- Fixed direct Anthropic Claude Sonnet/Haiku 4.5 advisor/agent turns crashing every call with HTTP 400 This model does not support the effort parameter. The catalog classified the whole Claude 4.5 family on anthropic-messages (and bedrock-converse-stream) as anthropic-budget-effort, which made the Anthropic provider serialize output_config.effort alongside thinking.budget_tokens. Anthropic only honors output_config.effort on Opus 4.5 and adaptive (4.6+) Messages-API models, so Sonnet 4.5 / Haiku 4.5 rejected the field. inferThinkingControlMode now gates anthropic-budget-effort to parsedModel.kind === "opus" && semverGte(version, "4.5") on both Anthropic-routed APIs, so Sonnet 4.5 / Haiku 4.5 on direct Anthropic + Cloudflare-AI-Gateway + Vertex + GitLab-Duo + Copilot + Bedrock fall through to plain mode: "budget" (thinking budget still scales with the selected effort tier). Opus 4.5 keeps anthropic-budget-effort. anthropic-budget-effort also stays in use for Anthropic-compatible third-party backends that natively support the field (Umans GLM 5.2). (#3497)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Ctrl+Z hanging the terminal after any tool call had run: the TUI tore down (ui.stop()) but the process kept running in Sl+ state, leaving the user with a dead terminal recoverable only via kill -9. The embedded brush-core shell behind every bash tool call installs a tokio SIGTSTP listener on Process::wait (crates/brush-core-vendored/src/sys/unix/signal.rs::tstp_signal_listener → tokio::signal::unix::signal(SIGTSTP)); per tokio's contract, the first call for a SignalKind permanently replaces the kernel-default handler for the lifetime of the process. So the first bash invocation — even /usr/bin/true — silently overrode SIGTSTP's "stop" default, and InputController.handleCtrlZ's subsequent process.kill(0, "SIGTSTP") was swallowed by tokio. The handler now sends SIGSTOP (uncatchable, unblockable, unignorable) to the foreground process group, so the kernel parks omp regardless of installed handlers and the shell sees the whole job stop even when omp runs behind a wrapper (npx, pnpm exec, bunx, …) or as one stage of a pipeline. MCP stdio servers now spawn detached into their own session — they're insulated both from terminal job-control signals (which used to stop their process trees and leave the JSONL read loop blocked on silent pipes) and from the new pgid=0 suspend itself (#3461).
- Fixed image-only composer submissions while the agent is streaming being treated as empty input, which dropped the image or aborted the active turn when another message was queued. Pending pasted images now count as submit content for Enter and Ctrl+Enter follow-ups. (#3467)
- Fixed omp gallery --state accepting lifecycle tokens that did not match displayed state labels and rendering unknown state values as · undefined; displayed labels now work as aliases, invalid values fail with a valid-token list, and failed gallery fixtures visibly render failures. (#3473)
- Fixed the bash tool's snapshotted mise() shell function dying with command: command not found: because $__MISE_EXE was empty in the replay shell. generateSnapshotScript captured the function via declare -f/typeset -f but only ever 
16.1.19 25 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed Ollama/llama.cpp chat payloads serializing user-attributed mid-conversation developer messages (auto-learn capture nudge, advisor cards, file-mention companions) as system turns; they now serialize as user so llama.cpp can reuse the warm prompt prefix instead of forcing full re-processing. Agent-owned developer reminders (attribution: "agent" — empty/unexpected-stop retries, checkpoint rewind warning, todo reminders) keep their system priority. (#3456)
- Fixed prior-turn reasoning being lost on cross-API provider switches: when a session moved from an Anthropic-compatible 3p endpoint to an OpenAI-compatible one (Z.AI Anthropic → Z.AI OpenAI, Kimi Anthropic → Kimi OpenAI, DeepSeek, OpenCode-hosted reasoning models, or any custom models.yaml switch that crosses API types), the cross-API path of transformMessages text-demoted every prior thinking block, so the next request shipped the reasoning chain as plain conversation content instead of structured reasoning_content — losing it as reasoning context and re-billing it. convertMessages now threads the request-time resolved compat into transformMessages, which preserves the prior reasoning as a native, signature-stripped thinking block whenever that resolved target accepts reasoning_content as a continuation hint (requiresReasoningContentForToolCalls — including the whenThinking policy OpenCode reactivates for thinking-on requests, #1071/#1484 — or thinkingFormat: "zai"); the openai-completions encoder surfaces those blocks via reasoningContentField, with a new branch for Z.AI-format hosts (Z.AI, Zhipu, Moonshot Kimi, Xiaomi MiMo) that accept but don't require the field. Targets that can't replay unsigned reasoning (encrypted reasoning blobs, signed thought parts, non-reasoning models, thinking-disabled OpenCode) still text-demote so the reasoning survives as conversation context. (#3437, #3439 by @roboomp; #3433, #3434)
- Fixed Bedrock cross-region inference profiles routing to us-east-1 regardless of their geo prefix: a profile such as eu.anthropic.claude-… (or apac./au./jp.) sent to the hardcoded us-east-1 endpoint returned HTTP 400 The provided model identifier is invalid. streamBedrock now derives the runtime region from the profile's geo prefix — honoring an ambient AWS_REGION/AWS_DEFAULT_REGION only when it can serve that geo and falling back to the geo's default region otherwise — while explicit per-request and ARN-embedded regions still win and region-agnostic global. profiles stay unchanged.
- Fixed malformed tool calls (empty name) wedging entire sessions in HTTP 400 loops: when a model occasionally emits { "name": "", "arguments": "{}" } (observed: GLM-5.2 + thinking on long turns), the agent rejected the call at execution time with Tool not found, but the malformed block plus its error toolResult stayed in conversation history and every subsequent request 400'd on tool_use.name/tool_calls[i].function.name validation until the user ran /clear. transformMessages — the canonical sanitize boundary every provider passes through — now drops toolCall blocks with empty/whitespace name, pairs them with their toolResult messages only inside the same assistant→tool-result window (per-id FIFO queue cleared at non-result boundaries, so stale malformed calls without a result cannot consume later valid duplicate-id outputs), and drops the assistant turn when it has no replayable content left. Defensive (provider-agnostic, fires regardless of model), idempotent (no-op on a clean history), and self-healing (one round-trip after the fix lands sanitizes an already-poisoned session). (#3458)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed omp install <plugin> failing extension validation in compiled-binary mode with Cannot find module '@(scope)/pi-ai/oauth' from '<plugin>/src/oauth.ts' (and any other non-wildcard pi-* subpath import like @oh-my-pi/pi-coding-agent/tools). The bundled-registry override map seeded by __buildLegacyPiPackageRootOverrides only covered the
16.1.18 25 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed AppendOnlyContextManager.syncMessages clearing the entire log on any in-place rewrite of an already-synced message. Per-turn tool-output pruning, image stripping, or any transformContext re-render that touched a single message used to drop every prior turn out of the append-only log and re-send the conversation from scratch, forcing local backends (llama.cpp / Ollama / LM Studio) to re-prefill tens of thousands of tokens every few turns. syncMessages now finds the longest byte-stable prefix between the previously-synced messages and the new ones, truncates the log to that prefix, and only re-appends the diverged tail — so the provider's KV cache stays warm up to the divergence point. (#3406)
@oh-my-pi/pi-ai
Added
- Added listOAuthAccounts for retrieving a read-only list of stored OAuth account identities
- Added getOAuthAccessAt to resolve an OAuth token exclusively for a specific account position
Changed
- Refactored OAuth token persistence and disable logic to use stable credential IDs instead of positional indices to prevent race conditions during concurrent updates
- Updated OAuth failure classification to treat 403 status codes, rate limits, and network errors as transient, preventing unnecessary credential invalidation
Fixed
- Fixed Codex Responses Lite staying enabled for image prompts, which caused GPT/Codex image turns to be rejected as Invalid value: 'input_image'; image-bearing Codex requests now fall back to the full Responses transport. (#3421)
- Fixed the auth-broker background refresher disabling OAuth credentials unconditionally (disableCredentialById) on a definitive refresh failure, so a credential another process or a fresh login rotated mid-refresh could be torn down even though the stored row already held a valid token. The definitive-failure teardown now happens inside AuthStorage.refreshCredentialById via the same compare-and-set the in-stream and usage-probe paths use — it disables only when the persisted row still matches the credential the refresh actually attempted, and reloads on a CAS loss; the refresher now only logs.
- Fixed OAuth refresh persisting the rotated token by a positional index captured before the refresh await. A concurrent disable could reorder or shrink a provider's credential array while the refresh was in flight, landing the new token on the wrong row (or silently dropping it) and leaving accounts with a stale refresh token that failed — and was then disabled — on the next cycle. Refresh persistence, selection-index resync, and CAS-disable now address the row by id across forceRefreshCredentialById, candidate preflight, and in-stream selection (#replaceCredentialById / #disableCredentialByIdIfMatches).
- Fixed isDefinitiveOAuthFailure treating a bare HTTP 403 (and generic unauthorized / access-token-expired wording) as a definitive credential failure, which permanently disabled healthy OAuth accounts on WAF, egress rate-limit, permission, and account-verification responses. Bare 403, rate limits (429), gateway/5xx, and more network errors (ECONNRESET, ETIMEDOUT, EAI_AGAIN, …) are now classified transient; only explicit dead-grant errors (invalid_grant, invalid_token, unauthorized_client, revoked, refresh token … expired) or a bare 401 tear the credential down.
@oh-my-pi/pi-coding-agent
Added
- Added --list flag to view stored OAuth accounts for a provider
- Added --account flag to select a specific OAuth account by index for token retrieval
- Added mouse support for navigation, selection, and toggling in the extensions dashboard
Changed
- Enabled fullscreen alternate-screen mode for the extensions dashboard
- Improved tab navigation to mute empty enabled providers while keeping disabled ones selectable
- /extensions (the Extension Control Center) now opens as a fullscreen window on the terminal's alternate screen, matching /settings: it borrows the alt buffer for its lifetime (the transcript is untouched underneath) and uses the same modern ch
16.1.17 24 de juny del 2026 — 64 bits Portable
16.1.16 23 de juny del 2026 — 64 bits Portable
16.1.15 22 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed API-key /login providers replacing sibling credentials instead of appending new keys for the same provider. (#3265)
- Fixed OpenAI Codex OAuth account rotation for quota failures that surface as bare HTTP 429 or insufficient_quota, so pre-content failures temporarily block only the exhausted credential and retry a healthy sibling. The 429 status-only fallback applies only to absent/opaque bodies; informative transient bodies (Too many requests, Service overloaded 529, Please retry in 5s, …) defer to parseRateLimitReason and stay in the provider's own backoff layer instead of burning sibling credentials. (#3231)
@oh-my-pi/pi-coding-agent
Added
- Added a share.store setting (blob "Encrypted Blob" | gist "GitHub Gist", default blob) controlling where /share uploads the encrypted session blob — the share server (blob) or a secret GitHub gist with share-server fallback (gist).
Changed
- Changed /share to upload the encrypted session blob to the share server by default instead of a secret GitHub gist. Gist-hosted shares are fetched through the unauthenticated GitHub gist API from the viewer's browser, which GitHub rate-limits to 60 requests/hour per IP, so shared NATs and repeated reloads surfaced Failed to load session: Gist fetch failed: HTTP 403 in the viewer. Opt back into gist hosting with share.store: "gist".
Fixed
- Fixed nohup/& background processes started inside an auto-backgrounded (:async:) bash turn being killed when that turn's shell was torn down. The per-job shell is now retained while a background process is still running (and reaped once it exits), so backgrounded commands survive across turns while still dying with the harness process.
- Fixed openai-completions provider session state surviving /model switches across different providers or base URLs. AgentSession.#closeProviderSessionsForModelSwitch only evicted openai-codex-responses and openai-responses:<provider> keys; entries keyed openai-completions:<provider>:<resolvedBaseUrl>:<modelId> (cached strict-tools disable scopes and reasoning-effort fallbacks for the old transport) lingered indefinitely. Moving away from an openai-completions backend now evicts every cached entry for the previous provider, including entries whose base URL was resolved at request time rather than read from the catalog, while same-backend model toggles keep their cached state (#3260)
- Fixed MCP stdio servers failing on Windows when the launcher's PATH walk can't pin down a bare npx/yarn/pnpm-style shim (empty Bun.env.PATH under a restricted parent process, UNC mounts that reject fs.access, locked-down shells). resolveStdioSpawnCommand used to fall back to the bare command name, which Bun.spawn → CreateProcess can only resolve as <name>.exe — never .cmd/.bat — so npx -y … died ~140ms after spawn with ENOENT/EINVAL. The Windows resolver now routes any unresolvable bare command through cmd.exe /d /s /c so Windows's own PATHEXT search picks up the shim. The reporter's diagnosis ("process.env not merged") was incorrect — the merge happens at transports/stdio.ts:316-319 — but the symptom they hit is real (#3250).
- Fixed bracketed paste treating a bare .png filename as an image attachment path instead of normal prompt text. (#3253)
- Fixed Perplexity web-search provider hijacking the auto fallback chain when only OpenRouter auth was configured. PerplexityProvider.isAvailable() accepted hasAuth("openrouter") as a credential, so any user with an OpenRouter key (for LLM access) had every webSearch: auto request silently routed through OpenRouter's perplexity/sonar-pro — bypassing downstream providers like Gemini and producing unexpected OpenRouter billing. Auto-chain admission now requires a direct Perplexity credential (PERPLEXITY_COOKIES, OAuth, or PERPLEXITY_API_KEY); users who want the OpenRouter-backed Perplexity path can still opt in explicitly with webSearch: perplexity (#3251)
- Fixed configured model discovery caches to refresh when models.yml/models.js
16.1.14 22 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added proxy support for model providers via PI_PROXY and PI_PROXY_<PROVIDER> variables
- Added NO_PROXY environment variable support for bypassing proxy configuration
- Added support for Sakana AI provider
- Added Sakana AI login and request base URL support for SAKANA_* / FUGU_* environment variables
Changed
- Consolidated API key authentication logic across registry providers
- Disabled parallel tool calls for Devin provider requests
Fixed
- Improved proxy bypass logic to correctly handle private IP ranges and local metadata services
- Enhanced memoization for proxy environment variable lookups to improve performance
@oh-my-pi/pi-catalog
Added
- Added Sakana AI provider support with Fugu model integration
- Added Sakana AI/Fugu provider catalog entries with Fugu model discovery and Responses API metadata
- Added support for "xhigh" reasoning tier across model configurations
- Added configuration for new models GCP-5.4 Mini, GPT-5.5, and variants
- Added devin variant collapse table to streamline model tiering
Changed
- Updated reasoning label pattern to include "minimal" and "max" efforts
- Simplified model identification logic for Devin-powered reasoning models
- Refactored variant routing to consolidate and standardize tier definitions
@oh-my-pi/pi-coding-agent
Added
- Added a Ruby eval backend (language: "rb"): a persistent subprocess Ruby kernel that shares the standard agent tool bridge and evaluation model, supporting IRB-style auto-display of results in a persistent session
- Added a Julia eval backend (language: "jl"): a persistent subprocess Julia kernel that shares the standard agent tool bridge and evaluation model, supporting Julia REPL-style auto-display of results in a persistent session
- Added configuration options eval.rb and eval.jl to enable or disable the Ruby and Julia backends (both opt-in, disabled by default), as well as ruby.interpreter and julia.interpreter paths for explicit runtime control
- Added a Ruby eval backend (language: "rb"): a persistent subprocess Ruby kernel modeled on the Python kernel, speaking the same NDJSON protocol with isolated frame/stdout channels, clean SIGINT cancellation that preserves kernel state, and per-owner kernel cleanup. Ships the full prelude helper surface (display/read/write/append/tree/diff/env/output/sort/uniq/counter, the tool.<name> bridge proxy, and completion/agent/parallel/pipeline/log/phase/budget) over the shared loopback tool bridge, honors local:// roots, and auto-displays the last expression IRB-style (suppressing assignments/definitions). Gated by the eval.rb setting and PI_RB env flag, with an optional ruby.interpreter override.
- Added a Julia eval backend (language: "jl"): a persistent subprocess Julia kernel with the same NDJSON/result-display/tool-bridge model as the Python and Ruby eval runtimes, including eval.jl / PI_JL gating, julia.interpreter override support, persistent per-session state, clean cancellation, owner-based kernel cleanup, local:// path roots, and the current Julia prelude helper surface (display/read/write/append/tree/diff/env/output, the tool.<name> bridge proxy, and bridge-backed completion/agent/parallel/pipeline/log/phase/budget, including structured schema parsing and return_handle DAG shaping).
- The eval tool advertises only the runtimes enabled for the session: a disabled backend (rb/jl are off by default) is omitted from the tool's language enum, schema field descriptions, discovery summary, and prelude documentation, so disabled backends are never surfaced to the model
Changed
- Refactored internal evaluation logic to consolidate shared kernel management and IPC handling
- Refactored internal draft state to consolidate images and text into the editor component
- Unified transcript rendering logic for background jobs, IRC traffic, and file mentions
- Unified subprocess lifecycle management for mnemopi, speech, tiny-model, and TTS workers
- Softened the system-prompt and tool-prompt guidance t
16.1.13 22 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added support for Devin as a provider
Changed
- Updated tool call arguments to use Record<string, unknown> and unknown for tool results
Fixed
- Fixed OpenAI Responses native history replay dropping failed/incomplete image generation calls instead of resending their transient ig_... item IDs, preventing follow-up requests from failing with 404 Item with id ... not found. (#3225)
- Fixed /login fireworks rejecting valid fw_… keys with Fireworks API key validation failed (500): Error listing deployed models. The validator pinged /inference/v1/models, which Fireworks serves from the per-account deployment registry and 500s for accounts without active deployments. Login now hits the static control-plane List Models catalog (GET /v1/accounts/fireworks/models?filter=supports_serverless=true&pageSize=1) — the same endpoint discovery already uses — so authentication no longer depends on the caller's deployment state. (#3219)
@oh-my-pi/pi-catalog
Added
- Added support for Devin as a model provider
- Added capability to fetch dynamic models from the Devin model manager
@oh-my-pi/snapcompact
Fixed
- Fixed the Umans provider image budget to match its 10-image request cap.
What's Changed
- fix(ai): validated Fireworks login against control-plane catalog by @roboomp in https://github.com/can1357/oh-my-pi/pull/3220
- fix(ai): drop stale OpenAI image generation replay by @roboomp in https://github.com/can1357/oh-my-pi/pull/3226
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.12...v16.1.13
16.1.12 21 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Changed
- Refined secret obfuscation to only target message roles and fields containing operator secrets
- Restricted obfuscator to ignore secrets and regex matches shorter than 8 characters
- Optimized obfuscation to skip static system prompts and tool schemas in provider contexts
- Ensured image data bytes are never modified to prevent corrupted data URL payloads
Fixed
- Fixed secret obfuscation corrupting Codex image reads (and other provider requests) with Invalid 'input[N].content[].image_url'. Expected a base64-encoded data URL ... but got an invalid base64-encoded value. The obfuscator deep-walked every string in the outbound request — including inline image base64 and opaque provider replay/signature fields — so a configured secret that happened to be a substring of the base64 (or of an ordinary word like response) injected #HASH# placeholders mid-payload. Obfuscation is now opt-in and fully typed: only user messages, tool-result messages, and user-attributed developer messages (@file mentions) are redacted; system prompts and tool schemas pass through untouched; image bytes and signature/encrypted-reasoning fields are never rewritten; and tool-call arguments are the only JSON walked. Configured plain secrets and regex matches shorter than 8 characters are now ignored to stop false matches on short words.
- Fixed RPC/ACP startup clobbering explicit caller/project/global configuration for task.isolation.{mode,merge,commits}, task.eager, task.batch, task.maxConcurrency, task.maxRecursionDepth, task.disabledAgents, task.agentModelOverrides, memory.backend, memories.enabled, advisor.{enabled,subagents,syncBacklog,immuneTurns}, plus the RPC-only async.{enabled,maxJobs} and bash.autoBackground.{enabled,thresholdMs}. applyDefaultSettingOverrides re-asserted the schema default as a runtime override after settings load, regressing the isConfigured() guard added for #2598 and ignoring every explicit value the embedder, project, --config overlay, or global config had set. The guard is restored, so the host default now only fills holes (#3207).
@oh-my-pi/pi-natives
Added
- Added Julia syntax highlighting to highlightCode/supportsLanguage via a vendored Julia.sublime-syntax folded into syntect's default set (jl/julia aliases); syntect ships no Julia grammar.
What's Changed
- fix(rpc): preserved explicit caller config across host-defaulted paths by @roboomp in https://github.com/can1357/oh-my-pi/pull/3209
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.11...v16.1.12
16.1.11 21 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed OpenAI Responses native history replay leaking image generation provider-only fields into the next request, which made OpenAI-compatible proxies reject pi tool-calling sessions with Unknown parameter: input[1].action. (#3201)
- Fixed a stream thought-leakage issue for gemini-3.5-flash where the model's internal reasoning JSON could leak into the visible text stream. The stream parser now uses a brace-balanced counting algorithm to accurately slice and discard the leading thought JSON block, with a robust fallback for unescaped double quotes, dynamic tool-name derivation, and preservation of subsequent text deltas without triggering empty-response retries.
@oh-my-pi/pi-catalog
Fixed
- Fixed Umans umans-glm-5.1 / umans-glm-5.2 advertising native image input. The models/info endpoint reports supports_vision: "via-handoff" for the GLM models, meaning vision routes through a separate handoff pre-analysis step instead of accepting raw image blocks; umansSupportsVision treated any non-empty string as native vision support, so image prompts went directly to GLM and were rejected with 400 This model does not support image inputs. The helper now requires supports_vision === true, the bundled GLM 5.1/5.2 rows are corrected to text-only, and stale mismatched Umans cache rows for those ids are dropped so the vision-handoff path runs even before a successful refresh. (#3184)
@oh-my-pi/pi-coding-agent
Added
- Added tab.waitForSelector for more robust element-wait behavior
- Added tab.waitForNavigation to monitor and await page transitions
Changed
- Standardized public HTML exports to use the brand web palette instead of the local TUI theme
- Updated export logic to allow choosing between brand-web or local TUI color palettes
- Clamped browser tool timeouts to fail fast with named errors instead of opaque cell timeouts
- Added strict validation to block unsupported Playwright-only selector engines in browser tools
Fixed
- Fixed Codex image reads re-encoding to WebP and then failing the next request with an invalid input_image.image_url; Codex-bound images now stay in PNG/JPEG-compatible formats.
- Fixed prompt-template and file slash-command submissions rendering both the raw slash invocation and the expanded prompt in the interactive transcript. (#3199)
- Fixed the /model thinking picker labeling the OpenAI GPT-5.5 top effort as max instead of the catalog-declared xhigh (#3194).
- Fixed session-title generation silently falling back to the online smol model (and billing whatever provider held the resolved API key — OpenRouter in the reporter's case) when the user had explicitly configured a local providers.tinyModel: generateSessionTitle raced local against online with a 10s timeout and fired the online request immediately whenever the local worker returned null (unknown key, model not downloaded, transformers.js failure). Now an explicit local-model choice is honored end-to-end — on local failure the session is left untitled with a logger.warn instead of billing the smol fallback (#3187)
- Fixed OpenCode MCP discovery so array commands are normalized into a stdio executable plus arguments, environment is accepted as the OpenCode env key, and argument lists are omitted when empty. (#3180)
- Fixed plan mode leaving write hidden under tools.discoveryMode: "all", which made the agent attempt to create the plan file via edit and stall. Plan-mode entry now force-activates write whenever the registry built it, matching the write+edit instructions in the plan-mode prompt (#3165)
- Fixed edit and patch tool writes bypassing the ACP client's open buffer when Zed (or another ACP client) advertises the fs.writeTextFile capability: all three write-mode tools (edit, patch, replace) now route through the client bridge when it is available, so the editor's TypeScript diagnostics panel updates immediately instead of requiring a workspace reload.
- Fixed the agent dashboard new-agent description and the hook editor (default hook
16.1.10 21 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Fixed labeled user interrupts retaining incomplete streamed tool calls before toolcall_end, which could persist malformed tool-call IDs into replay.
@oh-my-pi/pi-ai
Changed
- Improved JSON robustness by replacing external dependency with a custom, high-performance parser
- Strengthened streaming JSON parsing to prevent non-finite numbers from surfacing as undefined/NaN
- Configured JSON parser to reject JS-specific NaN and Infinity values for tool arguments
- Replaced the JSON repair/parse helpers (parseJsonWithRepair, parseStreamingJson) with a single from-scratch tolerant parser (RelaxedJson) that accepts single-quoted strings, unquoted object keys, trailing/stray commas, // and /* */ comments, Python True/False/None, raw control characters, invalid escapes, and unescaped apostrophes ('it's'). Final parsing still throws on truncated/garbage input (so a malformed tool call is skipped rather than executed with half-formed args) and rejects JS-only NaN/Infinity; streaming parsing stays non-throwing and rolls back incomplete trailing tokens instead of surfacing undefined/NaN. The Cursor provider's ad-hoc regex + JSON5 tool-argument parser now routes through the shared parser.
Fixed
- Fixed tool call ID normalization for Anthropic-compatible models
- Fixed Anthropic Messages replay sanitizing malformed tool-call IDs, including aborted native tool calls with empty IDs, so retries no longer send invalid tool_use.id / tool_result.tool_use_id pairs.
- Fixed the Codex Responses WebSocket transport attributing a prior turn's output to the current one on a reused connection: a trailing/duplicate frame from a cleanly-completed previous response that slipped past the queue drain could be consumed as this request's terminal (ending the turn with empty output) or as a stale tool call. Frames are now keyed by response.id — a frame carrying the previous response's id is dropped, and one carrying a third id (or a regressed sequence_number) fails closed so the turn retries instead of mixing two responses' streams. Idless frames (deltas, the rate-limit/metadata preamble, response.created-less streams) still pass through, matching upstream codex-rs.
- Fixed transformMessages pulling an earlier, orphaned tool result onto a later tool call that reused the same id (left behind when compaction folded the originating tool_use into a summary). The pending-call flush now pairs each call with a result positioned after its assistant turn, so a reused id surfaces its own output rather than a prior turn's.
- Fixed DashScope 429 rate-limit messages that mention authorization being classified as credential failures, preventing valid API keys from being invalidated after throttling. (#3172)
- Fixed OpenCode Go 401 Insufficient balance quota errors being treated as unknown failures instead of usage-limit errors, restoring credential rotation and fallback chains. (#3169)
Removed
- Removed the partial-json dependency; streaming JSON parsing now uses the in-house RelaxedJson parser.
@oh-my-pi/pi-coding-agent
Added
- Added tab.ariaSnapshot(selector?) to the browser tool for Playwright-format ARIA-tree YAML
- Added tab.ref("e5") and support for aria-ref=e5 selectors in all tab action methods
Changed
- Made write and find tools essential so they are always available initially (surviving tools.discoveryMode === "all" hiding) to ensure instructions to write/find files are immediately executable (#3165)
Fixed
- Fixed streamed tool-call previews freezing on their placeholder body ($ …, Write: …, an empty args tree) even after the tool finished: the pending card was created while arguments streamed, but when the closing full-arguments message_update never arrived (smooth-streaming disabled leaving the throttled arguments stale, an owned-dialect projector, or a superseded/aborted turn that still ran the call) nothing re-applied the final args. tool_execution_start — the one event every execution path emits with validated full arguments right
16.1.9 21 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added llama.cpp to the interactive /login provider list, accepting an optional API key while defaulting to local no-auth mode.
Changed
- Optimized generated AI tool schemas by collapsing verbose anyOf unions into standard enum types
Fixed
- Fixed tool-call argument validation dropping nested keys that were accidentally double-encoded
- Fixed the moonshot provider being locked to the international Kimi host (api.moonshot.ai): OpenAI-completions requests now honor a MOONSHOT_BASE_URL override so users can reach the Kimi China platform (api.moonshot.cn), which rejects keys issued for the international endpoint. (#2883)
- Fixed tool-call argument validation dropping fields whose object keys were accidentally JSON-encoded a second time (e.g. { "\"op\"": "done" }), which surfaced as spurious missing-required errors. A schema-agnostic pre-validation pass now recursively unwraps such double-encoded keys — through arrays and nested objects, and again after a JSON-string container is parsed — before the unrecognized-key repair can delete them.
Removed
- Removed the setNextRequestDebugPath, clearNextRequestDebugPath, and getNextRequestDebugPath utility functions for request debugging, as request/response recording now relies exclusively on the PI_REQ_DEBUG environment variable.
- Removed Wafer Pass (wafer-pass) login support; Wafer Serverless remains available as wafer-serverless.
@oh-my-pi/pi-catalog
Fixed
- Fixed the moonshot provider with no path to the Kimi China API: model discovery now honors a MOONSHOT_BASE_URL override (redirecting to api.moonshot.cn), and KIMI_API_KEY resolves as a fallback for MOONSHOT_API_KEY. (#2883)
- Fixed LiteLLM model discovery preserving colliding models.dev transport metadata (for example ollama-cloud deepseek-v4-flash) instead of keeping the LiteLLM openai-completions provider transport. (#3162)
Removed
- Removed bundled Wafer Pass (wafer-pass) catalog entries and generation support; Wafer Serverless remains available as wafer-serverless.
@oh-my-pi/pi-coding-agent
Added
- Added LLM request JSON export functionality to /dump
Changed
- Improved browser stealth by suppressing common automation flags and patching property descriptors
- Enhanced stealth for WebGL, Worker, IFrame, Screen, and Audio APIs to evade detection
- Updated toString patching to register native function sources for improved fingerprint protection
Fixed
- Fixed omp list and omp remove silently starting an interactive agent session (forwarding the bare verb to the model as a prompt) instead of surfacing the real omp plugin list / omp plugin uninstall <name> commands (#2935)
- Fixed lazy-initialized LSP servers (basedpyright/pyright, and likely gopls/rust-analyzer) hanging on the first request: the message reader matched incoming messages against pending client requests by id before checking for a method, so a server-originated workspace/configuration pull whose id collided with an in-flight request was swallowed as a bogus response, leaving the pull unanswered and the server wedged. The reader now routes any message carrying a method as a server request before id-matching (#3001)
- Fixed omp --approval-mode=yolo acp and other global option flags placed before a subcommand being rewritten to launch with the subcommand swallowed as prompt text; the CLI resolver now skips leading global flags (using the launch parser's value-consumption contract) and dispatches the real subcommand with the flags applied, so ACP mode honors the configured approval policy. (#2970)
- Fixed /mcp enable and /mcp disable reconnecting unrelated MCP servers by scoping toggle reconnect/disconnect work to the named server. (#3157)
- Stopped the local llama.cpp no-auth login placeholder from being sent as a discovery bearer token.
Removed
- Removed /debug dump-next-request command
- Removed Wafer Pass from CLI credential help; Wafer Serverless remains available.
What's Changed
- fix(cli): scope MCP toggles to one server by @roboomp in https://
16.1.8 20 de juny del 2026 — 64 bits Portable
16.1.7 20 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-catalog
Fixed
- Fixed MiniMax-M3 catalog context for the MiniMax Coding/Token Plan providers minimax-code and minimax-code-cn to report the documented 1M long-context tier instead of the upstream 512K pricing boundary; the previous patch only covered minimax/minimax-cn, so the Coding Plan picker still showed 512K in the status bar (#3097).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed custom models.yml providers rejecting the compat.supportsImageDetailOriginal override, so Responses-compatible proxies that reject snapcompact's native-resolution image hint can clamp frames to detail: "auto". (#3092)
@oh-my-pi/pi-tui
Fixed
- Fixed slash command autocomplete, inline hints, and Enter completion when the slash command is preceded by leading whitespace (#3095).
- Fixed empty / autocomplete burying user skill commands below every built-in command, so installed skills appear in the initial slash popup (#2875).
What's Changed
- fix(tui): surface skill commands in slash menu by @roboomp in https://github.com/can1357/oh-my-pi/pull/2878
- fix(config): allow Responses image detail compat in models.yml by @roboomp in https://github.com/can1357/oh-my-pi/pull/3093
- fix(tui): complete slash commands after leading spaces by @roboomp in https://github.com/can1357/oh-my-pi/pull/3096
- fix(catalog): pin MiniMax-M3 contextWindow to 1M on minimax-code(-cn) by @roboomp in https://github.com/can1357/oh-my-pi/pull/3098
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.6...v16.1.7
16.1.6 20 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Enabled inline prompts with /loop commands (e.g., /loop 10 fix the bug)
- Added support for compound duration formats in /loop (e.g., 1h30m)
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.5...v16.1.6
16.1.5 19 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Wire-encoded normalizeTools parameters unconditionally so tools whose intent resolves to "omit" (function intent or intent: "omit", e.g. builtin eval / resolve) no longer leak raw arktype/zod schema objects in parameters (#3074)
@oh-my-pi/pi-coding-agent
Changed
- Removed the legacy AgentSession.nextToolChoice() method. The per-turn tool-choice directive now flows solely through nextToolChoiceDirective() (which folds in the hard-choice dequeue plus active-tool filtering as a private helper), eliminating the dual entry point that let callers consume the queue while bypassing the soft pending-preview lifecycle. The underlying ToolChoiceQueue.nextToolChoice() is unchanged.
What's Changed
- fix(agent): wire-encode normalizeTools parameters for omit-intent tools by @roboomp in https://github.com/can1357/oh-my-pi/pull/3076
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.4...v16.1.5
16.1.4 19 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added bounded auto-retry for empty assistant completions specifically to the OpenAI Responses provider
- Added bounded auto-retry for empty assistant completions across the OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages providers. A benign terminal stop that streamed no content and billed no output tokens — the signature of a flaky OpenAI-/Anthropic-compatible gateway that intermittently 200s with an empty body — is now retried up to twice with exponential backoff (honoring providerRetryWait) before being surfaced, instead of silently stalling the agent loop. Retries fire only before any content streams, so live streaming (including thinking) is never delayed, retried, or duplicated.
Fixed
- Fixed the Antigravity (google-antigravity) request builder dropping labels.model_enum when the wire profile does not declare one. Required for Claude 4.6 ids whose AntigravityModelWireProfile carries only maxOutputTokens (no captured model_enum); the label is now emitted only when the catalog defines it. (#3067)
@oh-my-pi/pi-catalog
Fixed
- Fixed Claude 4.6 routing on the google-antigravity (and google-gemini-cli) Cloud Code Assist providers, whose backend exposes the models asymmetrically: claude-sonnet-4-6 has no -thinking twin and claude-opus-4-6 has only the -thinking twin. The shared thinkingPair family was routing thinking efforts on claude-sonnet-4-6 to a non-existent claude-sonnet-4-6-thinking wire id (404 Requested entity was not found); replaced both 4.6 entries with bespoke single-wire families so every effort and off resolve to the live wire id. Added claude-sonnet-4-6 and claude-opus-4-6-thinking entries to ANTIGRAVITY_MODEL_WIRE_PROFILES capped at the backend's 64000-output-token limit (over-cap requests 400'd with Request contains an invalid argument); modelEnum is now optional on AntigravityModelWireProfile since the Claude wire ids are accepted without a captured labels.model_enum. (#3067)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed omp bench and omp dry-balance failing to resolve models from extension providers
- Improved error reporting for omp bench runs that return no output or tokens
- Cache-miss marker no longer fires on a cold turn whose predecessor only wrote the prompt cache (never read it back). The session's opening request always writes the prefix with cacheRead 0, so a long-running first tool call (e.g. gh run watch) that outlived the provider's cache TTL surfaced a spurious ⊘ cache miss divider right under the opening message. The marker now requires the previous turn to have actually read a warm prefix, so it flags only a demonstrably working cache going cold — and collapses a run of consecutive cold turns to a single marker at the moment the cache broke.
- Fixed omp bench and omp dry-balance failing to resolve models from extension-registered providers (pi.registerProvider(...)). Both commands build a one-shot ModelRegistry that previously only knew built-in catalog providers, so a provider/model selector for a provider contributed by an extension under ~/.omp/agent/extensions/ errored with "Model not found". A new loadCliExtensionProviders helper loads the session's extensions, drains their provider registrations into the registry, and discovers dynamic provider catalogs before resolving selectors — mirroring the interactive session and omp models paths.
- omp bench now reports a run that streamed no content and measured no output tokens as a failure ("provider returned no output") instead of a misleading green check with tokens 0 / TPS 0.0.
- Cleared stale pending-preview gates when resolve finds no runnable handler, and forwarded the missing peekPendingInvoker and clearPendingInvokers hooks on the production session so staged previews actually reach the resolve tool and the gate drains on dispatch failure. (#3061)
- Deferred status-line usage refreshes off the render path and timeboxed the startup fetch so slow Anthropic quota lookups no longer block interactive sta
16.1.3 19 de juny del 2026 — 64 bits Portable
16.1.2 19 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Prevented sensitive raw JSON payloads from leaking into agent events during tool validation
- Ensured tool validation errors are handled correctly for malformed JSON parse inputs
- Ensure deep-cloning of tool-call arguments respects own enumerable properties
- Prevent direct object references between agent message snapshots and streaming events
@oh-my-pi/pi-ai
Added
- Added improved JSON repair capabilities for Anthropic tool arguments
- Added authentication broker discovery to sync credentials between local SQLite and remote state
Fixed
- Improved error feedback and transparency for malformed Anthropic tool call arguments
- Added automatic fallback for unsupported OpenAI reasoning effort levels
- Improved reliability when handling invalid reasoning parameter errors across OpenAI-compatible APIs
- Fixed OpenAI-compatible Chat Completions, Responses, and Azure Responses requests to retry once with the nearest provider-supported reasoning effort when an endpoint rejects xhigh/minimal-style effort values.
@oh-my-pi/pi-catalog
Added
- Added support for Gemini 2.5 Flash-Lite, 3.1 Flash-Lite, and 3.5 Flash models
- Added support for Moonshot V1 model family
Changed
- Updated context window and token limits for various Claude, Gemini, and GPT-OSS models
- Refined thinking mode behaviors and routing for supported LLM families
Fixed
- Fixed GLM-5.2 reasoning_effort so the top thinking tier reaches each host's genuine maximum instead of 400ing, mapping the internal xhigh tier per host dialect (verified against live endpoints): Z.ai/Zhipu collapse onto the model's none/high/max scale (xhigh → max); Fireworks, resellers, and Ollama Cloud keep their distinct lower tiers and remap only the top xhigh → max (merged over host quirks such as Fireworks' minimal → none); and OpenRouter — whose API rejects max and treats xhigh as its own max tier — now exposes the xhigh tier and forwards it verbatim. Dialect detection keys off resolved compat.thinkingFormat, so custom OpenRouter/Z.ai-format providers are covered too.
- Maintained thinking effort routing when discovery only returns the base model ID
- Improved credential retrieval logic for Antigravity and Codex providers via auth discovery
@oh-my-pi/pi-coding-agent
Added
- Added a welcome-screen tip for the /advisor runtime. Tips ending in a [NEW] marker now render a bold rainbow NEW! tag (it shimmers across the welcome intro's animation frames, then settles into a still rainbow) and are weighted to surface more often in the random tip rotation.
Changed
- Renamed the search tool's i parameter to case and inverted its semantics to represent case-sensitive search.
- Improved session history to export empty objects as {} instead of empty strings
- Refined system prompt and tool documentation to improve conciseness and clarity
- Simplified tool input descriptions for browser, eval, find, and memory-edit operations
- Refactored authentication storage discovery to share logic with other pi-ai tools
Fixed
- Fixed omp bench resolving an ambiguous model selector — a bare or canonical id shared by several providers (e.g. gpt-oss-20b or openai/gpt-oss-20b) — to a provider you have no credentials for. Bench resolves against the full catalog (credentials are ignored), so the default pick was decided by provider-priority order alone. It now redirects such selectors to an equivalent model under a provider with configured auth (honoring modelProviderOrder and canonical cross-provider variants), while an explicit provider/id selector is still benchmarked verbatim so forced/unauthenticated runs keep working.
- Resuming a session whose project directory no longer exists (deleted or renamed worktree) no longer crashes with an unhandled ENOENT … chdir rejection. The resume now keeps the current working directory instead of trying to chdir into the missing path, across the in-session selector, the --resume startup picker, and SessionManager.open/continueRecent.
- Fixed streaming reflowin
16.1.1 19 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Changed
- Migrated legacy macOS power settings to a single sleep prevention enum
- Defaulted display.cacheMissMarker setting to false to suppress the cache-miss marker by default.
- Replaced the four separate power.preventIdleSleep, power.preventSystemSleep, power.declareUserActive, and power.preventDisplaySleep boolean settings with a single cumulative power.sleepPrevention enum (off → idle → display → system). Each level adds the caffeinate flags of all lower levels. Existing configs are migrated automatically.
Fixed
- Cache-miss marker no longer fires on providers with implicit, best-effort prompt caching (Google/Gemini, OpenAI, Fireworks). Those report cacheWrite: 0 and drop cacheRead to zero intermittently as routine propagation noise that self-heals the next turn. Only explicit, prefix-controlled caches (Anthropic/Bedrock cache_control), which re-create the prefix on a cold turn as cacheWrite > 0, now surface a marker — where a zero cacheRead genuinely means the prefix broke.
- Fixed advisor dependent settings staying visible while advisor.enabled is off (#3027).
- Fixed LSP servers that dynamically register capabilities, including Expert, hanging semantic requests after client/registerCapability was rejected (#3029).
What's Changed
- fix(ux): hide advisor sub-options unless advisor is enabled by @roboomp in https://github.com/can1357/oh-my-pi/pull/3028
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.1.0...v16.1.1
16.0.11 19 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Changed
- Updated the display format for truncated file operation summaries
@oh-my-pi/pi-coding-agent
Added
- Added __advisor.jsonl transcript persistence for advisor model usage attribution and visibility in the Agent Hub
- Added defensive reservation against naming a task __advisor to prevent filesystem collisions with internal transcripts
- Added mode property to the CompactOptions interface for extension-based compaction control
- Updated /compact slash command to support soft, remote, and snapcompact subcommands for per-run strategy overrides
- Added /compact mode subcommands so a manual compaction can override the configured compaction.strategy/remoteEnabled for that one run: /compact soft (summarize locally, skip remote endpoints), /compact remote (summarize via the remote endpoint / provider-native compaction; warns and falls back to a local summary when no remote path is available), and /compact snapcompact (archive history onto dense bitmap images, no LLM call). Bare /compact and /compact <focus text> keep their existing behavior; soft/remote still accept trailing focus instructions, snapcompact rejects them. The subcommands are advertised to ACP clients and surface in TUI autocomplete. Extensions can pass the same selection via compact(instructions, { mode }).
- The advisor now persists its own turns to a subagent-style transcript (<session>/__advisor.jsonl) so the advisor model's usage is attributed in omp stats and its transcript is observable in the Agent Hub (read-only). The transcript follows session switches (/new, resume, branch) and is drained/released before a /drop deletes the old artifacts dir. The advisor remains a non-peer: it is hidden from agent-facing rosters (irc list, history://, subagent peer prompt, broadcast targets), is not messageable (irc send / collab chat), and is not revivable/killable from the Hub or collab.
Changed
- Standardized file elision markers to a compact […Nln elided…] format
- Updated transcript viewer to clear cached content if the session file is deleted while opened
- Refactored agent hub to use a fullscreen overlay for agent transcripts instead of inline chat
- Improved transcript rendering stability by using a non-incremental builder for viewer replays
- Advisor transcripts are now excluded from agent-facing surfaces like irc, history://, and peer rosters
- Advisor transcripts are read-only and cannot be messaged, revived, or killed via the Agent Hub or IRC
- Refined /compact argument parsing to reject focus instructions for modes that do not support them (e.g., snapcompact)
- Protocol hosts (RPC/rpc-ui/ACP) now host-default the full advisor settings group — advisor.syncBacklog and advisor.immuneTurns in addition to advisor.enabled/advisor.subagents — so a host that opts the advisor in gets the default tuning instead of inheriting the user's local advisor preferences.
Fixed
- Fixed memory-leaking stale transcripts in the agent viewer when underlying files are deleted
- Fixed the Agent Hub transcript viewer rendering the transcript body one column right of the "Agent Hub" title (and the title appearing to shift when scrolled to the top): the fullscreen viewer added its own outer gutter on top of the transcript rows, which already carry a 1-column left pad, so the header and body no longer shared a gutter. The viewer now renders the scroll body at full width without the extra gutter, and the file-mention row carries the same 1-column pad as every other row.
- Fixed the bash tool failing with pi-natives:command: syntax error at end of input on a valid &&/; chain whose later pipeline stage is a compound command, e.g. echo x && git log | while read h; do …; done | head. The minimizer's segmented-chain runner rebuilds each chain segment from the brush AST via pipeline.to_string() and re-executes that string, but simple_segment only validated the first pipeline stage — so a compound later stage (while/for/if/subshell) was re-serialized without its terminat
16.0.10 18 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Replaced the old legacy XML-ish pi owned tool-calling dialect with the new sigil-delimited format (§ call header with inline key=value scalars, «…» verbatim body fence for the dominant string argument, ¤ reasoning, ‡‡ tool result) using single-token markers that never occur in source code. Verbatim fences escalate Markdown-style (««…»») so re-rendered history never collides with payload content, and the scanner gates a bare § on an exact known-tool name to avoid swallowing prose. Round-trips and streams through the existing scanner contract at ~46% fewer tokens than the legacy format on typical calls; selectable via tools.format or PI_DIALECT=pi.
Changed
- Updated pi dialect formatting to use a token-frugal, sigil-delimited format (§, ¤, ‡‡)
- Updated pi dialect body fences to automatically escalate when content contains fence markers
- Changed pi dialect tool results response format to ‡‡ blocks
Fixed
- Fixed Bedrock application inference profile ARNs to route requests to the ARN's region instead of the default Bedrock runtime region. (#3004)
@oh-my-pi/pi-coding-agent
Added
- Updated the pi option of the tools.format setting to use the new compact sigil-delimited owned tool-calling dialect (§/«…»/¤/‡‡) that uses ~46% fewer tokens than the legacy format on typical calls.
- Integrated terminal QR codes directly into /collab and /collab view to display both deep links and scannable codes, and added a collab.webUrl setting for separately hosted collab web clients.
Changed
- Updated description of Pi dialect in tool-calling settings to clarify compact sigil format
Fixed
- Fixed legacy settings.json, models.json, and keybindings.json migration/loading to accept comments in JSON config files.
- Support quoted file paths with spaces (such as default macOS screenshots) in file mentions and @-prefixed CLI arguments (#2909).
- Fixed the first steering/follow-up message typed as auto-compaction begins being dropped instead of queued; the compaction abort controller (which backs isCompacting) is now installed before the auto_compaction_start event fires, so input routed during that window lands in the compaction queue rather than the core agent queue.
- Fixed queued steering/follow-up messages being silently cleared by the agent reset during a handoff; they are now captured and restored across the new-session reset.
- Fixed the /login provider list still offering providers listed in disabledProviders (and their credential-alias logins, e.g. openai-codex-device → openai-codex) as login targets (#2906).
- Fixed an owned MCP manager not being disconnected on AgentSession.dispose(), orphaning stdio MCP subprocesses across session teardown (/exit, print mode, subagent teardown); the dispose-time disconnect is now bounded so a slow HTTP/SSE transport close cannot stall exit (#2839).
- Fixed inspect_image resolving pasted image labels such as Image #1, [Image #1, WxH], and attachment://1 against current chat attachments instead of falling back to <cwd>/Image #1 (#2787).
- Accepted max as an alias for the top thinking level (xhigh) in selectors and --thinking, so DeepSeek V4 Pro can be selected with its provider-facing maximum effort (#2727).
- Fixed --provider=amazon-bedrock --model <application-inference-profile ARN> being rejected as an unknown model before the Bedrock provider could send the ARN to Converse Stream. (#3004)
@oh-my-pi/collab-web
Added
- Added support for collab browser wrapper links whose web UI host differs from the relay host, so the connect screen joins the relay encoded in the URL fragment.
@oh-my-pi/omp-stats
Changed
- Updated description of moderated content categories to use more inclusive terminology
Fixed
- Wide data tables (Requests, Errors, Overview, Projects) overflowed the page horizontally at narrow-desktop widths (768-1023px): the .stats-table-desktop-only wrapper used for mobile-card tables lacked the overflow-x: auto containment that .stats-table-container already has. They now scroll wi
16.0.9 18 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed OAuth login replacing all other active accounts for the same provider, allowing multiple OAuth accounts to coexist concurrently.
- Fixed legacy api_key credentials not being replaced/disabled atomically upon upgrading to OAuth login.
- Fixed a logic issue where AuthStorage lost session-to-credential stickiness upon CLI restarts, causing cold-starts for server-side prompt cache (KV cache) and wasting tokens.
- Fixed GitHub Copilot Responses requests rejecting image inputs that carry the detail: "original" hint with an HTTP 400 by degrading the hint to "auto" for hosts that do not support it; other hosts still preserve native-resolution frames (snapcompact). (#2822)
@oh-my-pi/pi-catalog
Fixed
- Fixed GitHub Copilot's anthropic-messages proxy being misclassified as a non-signing reasoning endpoint (replayUnsignedThinking: true). It forwards to signature-enforcing Anthropic, so replaying a stripped/unsigned historical thinking block as signature: "" — most visibly an end_turn-bound checkpoint/branch-return turn whose signature the transform must strip — caused a 400 Invalid signature that corrupted the session and re-tripped on every full history re-send (e.g. after toggling MCP servers). Copilot now degrades such blocks to text like the official API. (#2851)
- Added a supportsImageDetailOriginal compat flag that resolves to false for GitHub Copilot, whose Responses endpoint rejects the detail: "original" image hint with a 400, and true for every other host. (#2822)
@oh-my-pi/pi-coding-agent
Added
- Added a --print-thoughts flag for single-shot print mode that includes sanitized thinking blocks in the text output; default print output stays the final answer only. When thinking is hidden by the hideThinkingBlock setting, --print-thoughts un-hides it for the print run so the flag is not a silent no-op (an explicit --hide-thinking still wins).
Fixed
- Fixed active /goal mode being paused by internal compaction and session-switch lifecycle aborts, and made those switches persist wall-clock goal usage without charging time spent in another session to a preserved goal.
@oh-my-pi/pi-tui
Fixed
- Fixed bottom-anchored fullscreen overlays keeping their body rows but clipping off footer actions when terminal-height clamping is applied, restoring plan-mode approval options on short or stale-size terminals (#2957).
What's Changed
- fix(ai): degrade Copilot Responses image detail "original" to auto by @wolfiesch in https://github.com/can1357/oh-my-pi/pull/2902
- fix(auth-storage): atomic legacy api_key cleanup within sqlite transaction during oauth login by @cagedbird043 in https://github.com/can1357/oh-my-pi/pull/2929
- Add print-mode thinking output flag by @oldschoola in https://github.com/can1357/oh-my-pi/pull/2960
- fix(tui): preserve plan approval actions when overlay is clamped by @roboomp in https://github.com/can1357/oh-my-pi/pull/2961
- fix(auth): persist session-sticky credentials in SQLite cache across restarts by @cagedbird043 in https://github.com/can1357/oh-my-pi/pull/2972
- fix(session): prevent goal mode pause during compaction/switch by @usr-bin-roygbiv in https://github.com/can1357/oh-my-pi/pull/2986
New Contributors
- @wolfiesch made their first contribution in https://github.com/can1357/oh-my-pi/pull/2902
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.0.8...v16.0.9
16.0.8 18 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Stopped the compaction <files> summary from tracking scheme:// URLs — internal URIs (conflict://, artifact://, local://, history://, …) and web URLs are no longer recorded as files, and legacy entries rehydrated from older compaction summaries are dropped.
@oh-my-pi/pi-ai
Fixed
- Improved reliability of auth-broker snapshot loading by implementing a robust manual schema check
- Fixed MCP tool argument validation to drop optional empty-string parameters before schema validation, matching the existing optional null handling and avoiding pattern/type failures for omitted model-filled fields. (#2981)
- Fixed API-key credential replacement to hard-delete superseded disabled api_key rows so auth_credentials does not grow indefinitely after key rotation. (#2941)
- Fixed Cursor provider streaming to close text blocks before tool calls so post-tool text opens a new content block and TUI transcript cards render inline instead of grouped near the bottom. (#2924)
@oh-my-pi/pi-catalog
Changed
- Refactored model family ID predicates and capability checkers to use a shared, uniform process-lifetime memo utility to eliminate caching boilerplate.
Fixed
- Fixed LM Studio dynamic discovery to use native /api/v0/models metadata so VLM models advertise image input. (#2945)
@oh-my-pi/pi-coding-agent
Changed
- Optimized app startup time by lazily loading site-specific scraper modules
- Refactored internal archive handling into a unified src/utils/zip.ts module
- Centralized all fflate (ZIP) and Bun.Archive (tar/tar.gz) operations into zip.ts
- Optimized archive reading by using lazy, ranged central-directory access for ZIP files
- Updated internal image processing to no longer include metadata text for fetched images
- Optimized omp:// documentation indexing by compressing doc bodies into a lazily-inflated blob
- Changed Mermaid fenced-block ASCII rendering to use the first-party vendored renderer in @oh-my-pi/pi-utils (src/vendor/mermaid-ascii), dropping the beautiful-mermaid npm package, its transitive elkjs (~3.13MB), and the beautiful-mermaid bun patch; CJK/emoji width handling and the layout-direction override are preserved.
- Changed omp:// documentation embedding to a gzipped base64 index (docs-index.generated.txt, populated at build time and reset afterward) inflated on first read, instead of a ~1.6MB raw TypeScript map. The compiled binary / npm bundle drops ~0.9MB; the dev tree and source checkouts read docs/ from disk.
- Replaced the markit-ai package with a vendored in-house document engine (src/markit) for the document formats it converts: PDF (via mupdf), DOCX (via mammoth), and PPTX/XLSX/EPUB (via fast-xml-parser). Conversion output is preserved, including the PDF column/table-detection pipeline and HTML-table normalization; markit-ai's unused converters and their dependency tail are gone. Legacy .doc/.ppt/.xls/.rtf remain unsupported (a conversion error), as before.
- Centralized ZIP handling behind a single src/utils/zip.ts (fflate): the new document converters, the write tool's in-place archive editing, and the read tool's ranged archive reader now share one ZIP implementation instead of mixing jszip and fflate.
Fixed
- Fixed settings overlay crash when scrolling past the last row in list views
- Improved tool result formatting by correctly wrapping <out> blocks in dim-ink toggles
- Fixed auto-retry after transient model-stream socket closes to replay text/thinking-only partial assistant output, including turns where incomplete tool-call arguments were dropped; completed tool calls still block retry to avoid duplicating tool execution.
- Fixed omp update reporting EPERM: operation not permitted, unlink '<binary>.bak' on Windows when self-replacing a standalone binary, even though the new binary had already been installed. The backed-up old executable is still the running process image and cannot be unlinked until the process exits, so the post-verify backup cleanup is now best-effort, backups use a u
16.0.7 18 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Changed
- Switched Google OAuth callback hostname from localhost to 127.0.0.1 to prevent IPv6 loopback fallback delays and proxy routing interception.
Fixed
- Fixed OpenCode Go usage reporting to synthesize /usage limits from OMP-observed request costs for the 5h, weekly, and monthly provider caps. (#2942)
- Fixed MiniMax Anthropic-compatible requests to serialize adaptive thinking without an invalid Anthropic output_config.effort tier (#2928).
@oh-my-pi/pi-catalog
Fixed
- Fixed MiniMax Anthropic-compatible M2/M3 thinking metadata to expose the adaptive transport and keep M2 mandatory reasoning floored (#2928).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed /model in the TUI to open the model setup picker again, leaving /switch as the temporary session model switcher (#2933).
- Fixed OpenCode Go sessions recording per-request cost history so /usage can show local cap utilization. (#2942)
@oh-my-pi/pi-natives
Added
- Added Fortran support to the AST tooling, including file/alias resolution.
What's Changed
- fix(coding-agent): restore /model role picker by @roboomp in https://github.com/can1357/oh-my-pi/pull/2938
- fix(ai/kimi): sanitize client device headers to avoid Cloudflare 520 errors by @usr-bin-roygbiv in https://github.com/can1357/oh-my-pi/pull/2939
- fix(coding-agent): bypass stream-interrupted guard for Gemini malformed function calls by @usr-bin-roygbiv in https://github.com/can1357/oh-my-pi/pull/2940
- fix(providers): add OpenCode Go usage reporting by @roboomp in https://github.com/can1357/oh-my-pi/pull/2947
- perf(auth): switch Google OAuth callback hostname to 127.0.0.1 for fast loopback by @cagedbird043 in https://github.com/can1357/oh-my-pi/pull/2932
- fix(providers): map MiniMax Anthropic thinking by @roboomp in https://github.com/can1357/oh-my-pi/pull/2934
- feat(coding-agent): add Fortran support by @TobiasLinnITHE in https://github.com/can1357/oh-my-pi/pull/2964
New Contributors
- @TobiasLinnITHE made their first contribution in https://github.com/can1357/oh-my-pi/pull/2964
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.0.6...v16.0.7
16.0.6 18 de juny del 2026 — 64 bits Portable
16.0.5 17 de juny del 2026 — 64 bits Portable
16.0.4 17 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed tool argument coercion to parse double-encoded JSON strings, including quoted values like "300", when schema expects a number
- Fixed object-array coercion to parse JSON object and array strings into proper array arguments instead of wrapping raw strings
- Fixed handling of malformed JSON container strings for array schema fields so validation now surfaces a top-level expected array, received string error rather than nested element errors
- Fixed ChatGPT/Codex browser login missing connector OAuth scopes and rendering object-shaped token endpoint errors as [object Object]. (#2825)
- Fixed Zhipu/BigModel GLM-5.2 chat-completions requests so internal xhigh effort serializes as provider-native reasoning_effort: "max" and tool calls opt into tool_stream. (#2833)
- Fixed Google Gemini CLI and Antigravity tool calls with toolChoice: "auto" serializing an explicit toolConfig AUTO mode, which can cause Gemini-3 models to leak raw planning JSON instead of executing tools. (#2830)
@oh-my-pi/pi-catalog
Fixed
- Fixed GLM-5.2 catalog thinking metadata for Zhipu/BigModel so the top effort is exposed as xhigh and maps to provider-native max. (#2833)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed RPC/ACP startup forcing todo settings back to host defaults, so project-level todo.enabled, todo.reminders, and todo.eager opt-outs now suppress protocol-mode todo prompt injection; enabled todo reminders are now persisted to the JSONL transcript so the log matches the model-visible context (#2824).
- Fixed default prompts to instruct the agent to read applicable skill://<name> content before starting work, so discovered skills influence broad task requests like frontend generation (#2829).
- Fixed hashline visible-line validation for ACP editor reads so INS.POST anchors displayed by bridge-backed range and multi-range read output are merged into the session snapshot before edit validates them (#2773).
@oh-my-pi/pi-natives
Fixed
- Fixed summarizeCode BFS unfold aborting the entire pass when it hit an oversized, un-unfoldable leaf span (e.g. an HTML <style> raw-text block, an embedded blob, or a minified line) whose only unfold candidate is its whole body. The overflow check used to break the breadth-first loop, so any large leaf encountered before its siblings starved the rest of the tree — an HTML page summarized to <style> ... </style> plus <div class="page"> ... </div>, collapsing the document body into one dead .... An overflowing span is now skipped (left folded, its subtree unexplored) and the BFS keeps unfolding the remaining queued siblings, so structured siblings like the <body> DOM are revealed up to unfoldLimit while the oversized leaf stays folded.
What's Changed
- fix(hashline): record ACP read lines for edit provenance by @roboomp in https://github.com/can1357/oh-my-pi/pull/2823
- fix(rpc): honor todo setting opt-outs by @roboomp in https://github.com/can1357/oh-my-pi/pull/2828
- fix(ai): repair ChatGPT Codex token exchange by @roboomp in https://github.com/can1357/oh-my-pi/pull/2832
- fix(providers): omit google cli auto tool config by @roboomp in https://github.com/can1357/oh-my-pi/pull/2835
- fix(prompting): load applicable skills before work by @roboomp in https://github.com/can1357/oh-my-pi/pull/2837
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v16.0.3...v16.0.4
16.0.3 17 de juny del 2026 — 64 bits Portable
16.0.2 16 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added UMANS_WEBSEARCH_PROVIDER=native|exa support for routing Umans gateway-owned web search requests.
Fixed
- A single MCP tool whose input schema can't be emitted as a valid strict tool schema for the active provider no longer fails the whole turn with HTTP 400. convertTools (openai-responses) now validates each tool's emitted parameter schema for enum/const-vs-type contradictions that pass structural JSON-Schema validation but the provider rejects — e.g. a non-null enum on a type: "null" node, or an enum on an array node — and quarantines just the offending tool with a logger.warn naming the tool and schema path, keeping every other tool usable. Adds findStrictToolSchemaViolation to @oh-my-pi/pi-ai/utils/schema (#2652)
- Fixed OpenAI Responses-compatible streams from Ollama/local hosts dropping arguments for parallel tool calls whose deltas use fc_<call_id> item ids, which left earlier ast_grep calls with {} and failed validation. (#2715)
- Fixed dialect transcript rendering so literal thinking envelopes are unwrapped before adding the dialect's own thinking tags, preventing nested <thinking> output in advisor raw dumps (#2700).
- Fixed Anthropic-compatible Umans requests escaping client tool names and forwarding gateway web search headers so Kimi answers normally instead of returning raw gateway search results.
- Fixed Google Gemini tool calls with toolChoice: "auto" serializing an explicit toolConfig AUTO mode, which can cause Gemini-3 models to leak raw planning JSON instead of executing tools. (#2776)
- Fixed OpenAI-compatible Ollama completions that return empty finish_reason:length after filling num_ctx so they surface an actionable context-window error instead of an empty length stop. (#2774)
- Fixed Codex browser login issuing credentials for the opencode OAuth originator while OMP requests identify as pi, which could make the first authenticated Codex request return 401 (#2696).
@oh-my-pi/pi-catalog
Fixed
- Fixed Kimi output caps for Umans AI Coding Plan and Venice so discovery metadata cannot use context-sized token ceilings as request caps.
- Marked Umans Anthropic-compatible models as client-tool escaped so cached and bundled metadata do not expose web_search as a provider server tool.
@oh-my-pi/pi-coding-agent
Added
- Added the UMANS_WEBSEARCH_PROVIDER environment variable to CLI help for Umans gateway web search backend selection.
Changed
- The eager task (task.eager: always) and eager todo (todo.eager: preferred/always) hidden reminders now re-fire on the auto-continuation turn after a compaction (context-full / snapcompact / handoff / shake). Compaction summarizes away the first-message prelude, so the agent would otherwise silently lose the delegate-via-tasks / phased-todo guidance mid-work; the post-compaction todo nudge is reminder-only and never forces the todo tool onto the resumed turn.
Fixed
- Fixed edit-tool block operations on Emacs Lisp files: .el and .emacs paths now resolve top-level forms for SWAP.BLK, DEL.BLK, and INS.BLK.POST instead of reporting an unsupported-language block-resolution error.
- Fixed PDF reads leaking recoverable MuPDF WASM warnings into the terminal TUI by routing MuPDF output through the file logger before markit-ai loads it (#2766).
- Fixed /exit and /quit waiting one shutdown timeout per hanging extension by running session_shutdown handlers within a shared shutdown window (#2736).
- Fixed GitHub Copilot .github/instructions/*.instructions.md discovery by loading those files as rules that honor applyTo scoping, including always-apply ** files and rule://<name> access for glob-scoped entries (#2731).
- Fixed Windows bash-tool child processes defaulting interpreter pipe I/O to the ANSI codepage by adding UTF-8 encoding defaults when the inherited environment is unset (#2701).
- Fixed /advisor dump raw so Opus 4.5 thinking content that already includes literal <thinking> tags is not rendered with nested thinking tags (#2700).
- The plugin-extensions-d
16.0.1 15 de juny del 2026 — 64 bits Portable
16.0.0 15 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Breaking Changes
- Renamed owned tool-calling options from toolCallSyntax/exampleSyntax to dialect/exampleDialect.
- Changed compaction conversation serialization to use the target model's native dialect turn, thinking, tool-call, and tool-result envelopes when a dialect is selected.
- Renamed the owned dialect environment variable from PI_OWNED_TOOLS to PI_DIALECT.
Added
- Added onTurnEnd hook support (setOnTurnEnd/onTurnEnd) to run awaited per-turn bookkeeping with current messages before the next model request and skip callback execution for aborted or error turns
Changed
- Renamed toolCallSyntax option to dialect in AgentOptions and AgentLoopConfig
- Updated conversation serialization to use dialect's native transcript rendering when a dialect is selected
- Changed internal references from ToolCallSyntax type to Dialect type across agent loop and compaction modules
@oh-my-pi/pi-ai
Breaking Changes
- Renamed the public dialect entrypoint from @oh-my-pi/pi-ai/grammar to @oh-my-pi/pi-ai/dialect.
- Renamed grammar dialect identifiers from ToolCallSyntax to Dialect, renamed the Grammar interface to DialectDefinition, and renamed Grammar.syntax to DialectDefinition.dialect.
- Added DialectDefinition.renderThinking and DialectDefinition.renderTranscript so dialect implementations serialize complete native chat transcripts, not just tool call/result blocks.
Added
- Added renderTranscript method to dialect definitions for serializing complete native chat transcripts
- Added renderThinking method to dialect definitions for rendering thinking/reasoning blocks
- Added support for 11 dialect implementations: Anthropic, DeepSeek, Gemini, Gemma, GLM, Harmony, Hermes, Kimi, Pi-native, Qwen3, and XML
- Added createInbandScanner factory function to instantiate dialect-specific scanners
- Added getDialectDefinition function to retrieve dialect implementations by name
- Added renderToolCatalog and renderInbandToolPrompt functions for tool catalog rendering
- Added renderToolInventory function to generate human-readable per-tool documentation with examples
- Added renderToolExamples function to render tool usage examples in the model's native dialect
- Added encodeInbandToolHistory function to encode tool call history in dialect-specific format
- Added wrapInbandToolStream function to process streaming responses with in-band tool call parsing
- Added ThinkingInbandScanner for parsing thinking/reasoning blocks across dialects
- Added OwnedStream class for managing dialect-aware streaming with tool call events
- Added in-band thinking channels to every dialect that was missing one: gemini (a ```thinking fence mirroring ```tool_code), gemma (its native <|channel>thought…<channel|> reasoning channel), kimi (<think>…</think>), and pi (<thinking>…</thinking>). Each scanner now parses reasoning into thinking events instead of leaking chain-of-thought into the visible reply, and every dialect's renderThinking is a real channel that round-trips back through its scanner (no passthrough renderers).
Changed
- Moved public dialect entrypoint from @oh-my-pi/pi-ai/grammar to @oh-my-pi/pi-ai/dialect in package exports
- Updated internal imports in stream-markup-healing.ts to use new dialect module path
- Changed renderToolInventory to demote a tool description's own markdown headers by one level when it contains a top-level # header, so they nest under the wrapping # Tool: <name> heading instead of reading as sibling sections. Descriptions that already start at ## and headers inside fenced code blocks are left untouched.
Fixed
- Fixed Gemini, Gemma, Kimi, and Pi in-band scanners to respect parseThinking: false, leaving private reasoning markers in visible text when parsing is disabled
- Fixed thinking-channel parsing for streaming Gemini, Gemma, Kimi, and Pi outputs so split or partial <thinking> blocks no longer leak into visible replies
- Fixed in-band thinking finalization and Kimi stream-healing interactions so leaked <think> block
15.13.3 15 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added the interruptible tool field: when set, the agent loop may abort the tool mid-execution to deliver a queued steering message (honored only in immediate interrupt mode).
- Added support for gemini and gemma as valid owned tool syntax values in environment configuration
Fixed
- Fixed pruneToolOutputs blanking tiny tool results during overflow pruning: results below 50 tokens (MIN_PRUNE_TOKENS) are no longer replaced with the [Output truncated - N tokens] placeholder, which cost more tokens than the result itself and churned the prompt cache for zero savings.
@oh-my-pi/pi-ai
Added
- Added the gemini in-band tool-call syntax with Python-style tool_code blocks and default_api invocations
- Added the gemma token-delimited in-band tool-call syntax using <|tool_call> and <|tool_response> blocks
- Added gemini and gemma to owned stream tool-result token detection so their tool responses are recognized
- Fixed truncated Gemini and Gemma tool blocks from being emitted as plain text during streaming
- Added the Azure OpenAI provider definition (azure) to the registry; AZURE_OPENAI_API_KEY resolves as its env-var API key via the catalog provider table.
Changed
- Gemini tool-call examples now render without the default_api. namespace prefix, keeping <example> blocks concise. The live wire format still uses default_api. per the Gemini grammar.
Fixed
- Fixed duplicate tool call projections by deduplicating provider-native toolCall events against in-band tool_code calls and keeping only the first real channel
- Dropped nameless native toolCall events so they no longer appear as surfaced tool calls in owned-mode streams
- Fixed Gemini/Gemma in-band tool-call parsing around Python comments, raw/unicode string literals, and Gemma close-token text inside string values.
@oh-my-pi/pi-catalog
Added
- Added Azure OpenAI as a catalog provider (azure, default model gpt-5.5, env var AZURE_OPENAI_API_KEY), bundling the OpenAI-family models Azure serves over the Responses API (GPT-4/4.1/4o, GPT-5 family, o-series, Codex). Like Amazon Bedrock it is catalog-only — models ship in the bundle and become selectable once the env key is set, with the deployment base URL resolved at runtime from AZURE_OPENAI_BASE_URL/AZURE_OPENAI_RESOURCE_NAME.
- Added models.dev-backed bundled catalogs for providers that previously shipped no offline models: Hugging Face, Kilo, Moonshot, NanoGPT, Synthetic, Venice, Ollama Cloud, and the Xiaomi Token Plan regions (ams/cn/sgp). They still discover live when credentialed; the bundle is now a non-empty baseline.
Changed
- Updated stale provider default models to their latest bundled versions: OpenAI-family providers (azure, github-copilot, aimlapi) → GPT-5.5; Gemini providers (google, google-gemini-cli, google-vertex) → gemini-3.1-pro-preview; GLM providers (zai, zhipu-coding-plan) → glm-5.2, cerebras → zai-glm-4.7; Kimi providers (fireworks, opencode-go, moonshot) → kimi-k2.7-code, kimi-code → kimi-for-coding, together → moonshotai/Kimi-K2.7-Code; alibaba-coding-plan → qwen3.7-plus; and Claude-Sonnet defaults (cloudflare-ai-gateway, cursor, gitlab-duo, kilo, opencode-zen, vercel-ai-gateway) → Claude Opus 4.x.
- Restricted models.dev Azure discovery to OpenAI-family IDs (gpt-, o1, o3, o4, codex, chatgpt), excluding Foundry-hosted third parties (Claude/DeepSeek/Llama/Mistral/Phi) that Azure serves through non-Responses APIs.
- Detected the Azure OpenAI Responses compat surface (developer role, strict tool mode, strict tool-result pairing) by provider id as well as base URL, so bundled azure models whose deployment host is only known at runtime still get the right wire behavior.
- Renamed the Qwen3-ASR-Flash model label to Qwen3 ASR Flash
Fixed
- Fixed tool syntax selection for Gemini-family and Gemma model IDs by routing them to dedicated gemini and gemma formats instead of generic XML
- Fixed zhipu-coding-plan and together shipping no bundled models: their descriptors referenced non-existent model
15.13.2 15 de juny del 2026 — 64 bits Portable
15.13.1 15 de juny del 2026 — 64 bits Portable
15.13.0 14 de juny del 2026 — 64 bits Portable
15.12.4 13 de juny del 2026 — 64 bits Portable
15.12.3 12 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Restored the intended double-Esc default to /tree and fixed the Esc-opened selector path to reset the terminal display instead of preserving stale scrollback when the selector opens.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.12.2...v15.12.3
15.12.2 12 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Collab links now dot-join the room secret (<roomId>.<key>, host/r/<roomId>.<key>) instead of using a second #: RFC 3986 forbids a raw # inside a URL fragment, so macOS Foundation (behind terminal click-to-open) percent-encoded the browser deep link's second # to %23 and the web client rejected the session. /join, omp join, and the web client still accept legacy #-joined links and leniently decode %23-mangled ones
- Fixed brew install can1357/tap/omp failing on macOS with sandbox-exec … exited with 1: the generated Formula/omp.rb (rendered by scripts/ci-update-brew-formula.ts) now stamps using: :nounzip on every per-platform url so Homebrew leaves the bare Mach-O/ELF asset in the staging CWD (the previous Dir["omp-*"].first returned nil because UnpackStrategy::Uncompressed#extract_nestedly nested the file outside it), and wraps generate_completions_from_executable in with_env(HOME: buildpath) so the popened binary's ~/.omp lookup goes to the writable staging dir instead of the sandbox-denied real home (#2398)
@oh-my-pi/collab-web
Fixed
- Link parsing accepts the new dot-joined room secret (<roomId>.<key>, /r/<roomId>.<key>) and leniently decodes %23-mangled legacy deep links (macOS Foundation percent-encodes a second # when terminals open clicked links), which previously failed to connect
What's Changed
- fix(brew): unbreak the generated homebrew formula install by @roboomp in https://github.com/can1357/oh-my-pi/pull/2399
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.12.1...v15.12.2
15.12.0 12 de juny del 2026 — 64 bits Portable
15.11.8 12 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Breaking Changes
- Removed the Codex SSE stateful transport path, so SSE turns no longer send previous_response_id with delta input and now always send the full transcript
Changed
- Scoped x-codex-turn-state handling to within-turn continuations so only tool-loop follow-ups include the turn-state header and new user turns start without it
Removed
- Removed the statefulResponses option from OpenAICodexResponsesOptions, and SSE stateful mode is no longer controlled by the PI_CODEX_STATEFUL-style flag
Fixed
- Fixed the platform OpenAI Responses and Codex websocket stale-chain classifiers missing the "Unsupported parameter: previous_response_id" rejection phrasing (FastAPI-style detail body with no error.code), so a chained turn now falls back to a full-transcript replay instead of surfacing the 400
- Fixed the HTTP-400 raw-request dump for Codex SSE to record the body actually sent on the wire instead of the pre-transport request body, which made chained-request failures look like the rejected parameter was never sent
@oh-my-pi/pi-catalog
Fixed
- Fixed Antigravity gemini-3.1-pro --thinking high failing with Cloud Code Assist API error (400): Request contains an invalid argument. — the upstream gemini-3.1-pro-high deployment rejects every streamGenerateContent request on both CCA endpoints while discovery still advertises it. High effort now routes to gemini-pro-agent (the same "Gemini 3.1 Pro (High)" model, verified accepting the identical request body), and the model-cache fingerprint version was bumped (merge-v2 → merge-v3) so existing fresh caches refetch discovery and pick up the corrected routing immediately.
@oh-my-pi/pi-coding-agent
Added
- Updated collab link handling to accept compact roomId#key links and relay hosts without explicit scheme when joining or starting sessions
- Added /collab stop and /collab status options to control and inspect active shared sessions
- Added /collab, /join, and /leave for live session sharing: the host shares an end-to-end encrypted link (AES-256-GCM key only in the link fragment; the relay sees opaque bytes) and guests render the session natively in their own TUI — streaming text, tool cards, footer state, ctrl+o expansion, /dump — and can prompt or interrupt the host's agent. Guest prompts render with an author badge; session-mutating commands stay host-only. Defaults to the public relay.omp.sh relay (collab.relayUrl); a self-hostable Go relay lives in the pi-www repo as omp-collab-relay
- Added collab.relayUrl and collab.displayName settings plus a collab status-line segment showing the participant count (host) or guest role. Guests mirror the host's real model/thinking state into their replica agent and the host's subagent ecosystem end-to-end: the live subagent HUD, the Agent Hub table with live progress, hub chat/kill/revive (routed to the host), and on-demand subagent transcript viewing
- Added omp join <link> subcommand that launches the interactive TUI and immediately runs /join <link>
Changed
- Moved collab live-session wire contracts into @oh-my-pi/pi-wire and stopped broadcasting unsupported session events or entries to collab guests.
- The Ctrl+P role-cycle track and the plan-approval model slider now color segments by track position from the theme's own palette (accent/success/warning/error + markdown/syntax hues, deduplicated per theme since many themes alias them) instead of role-keyed colors with a gray fallback for custom roles
- Interactive boot no longer blocks first paint on MCP server discovery. For UI sessions, createAgentSession returns immediately and connects to configured MCP servers in the background; their tools and slash commands stream in through the existing live-refresh channel once connected. An explicitly requested MCP tool whose server has not finished connecting resolves to a deterministic "still connecting" placeholder instead of an "unknown tool" error, and each server's instructions join the system prompt once its background connecti
15.11.7 12 de juny del 2026 — 64 bits Portable
15.11.6 12 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Changed
- Renamed the saved Codex reset command from /reset-usage to /usage reset and added /usage show as an explicit alias for plain /usage.
Fixed
- Fixed the Subagents HUD order so progress updates no longer move active rows; batched subagents keep their task order.
- Fixed /settings dirty-value highlighting so changed values stay orange while the row has keyboard focus.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.11.5...v15.11.6
15.11.4 12 de juny del 2026 — 64 bits Portable
15.11.3 11 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed GitHub Copilot long-context model requests to use the upstream requestModelId when calling Anthropic, OpenAI Responses, and OpenAI Completions APIs
- Fixed GitHub Copilot model enablement to deduplicate catalog variants by upstream model ID when enabling all models
@oh-my-pi/pi-catalog
Added
- Added requestModelId on Model to represent the upstream model id used when a catalog entry is a local variant
- Added synthetic GitHub Copilot long-context model variants with -1m suffixes when tiered token pricing is advertised
Changed
- Changed GitHub Copilot discovery to request X-GitHub-Api-Version: 2026-06-01 from api.githubcopilot.com
- Changed GitHub Copilot discovery to cap base model contextWindow to the default token tier and keep long-context access as the separate -1m model entry
- Changed Copilot model mapping to omit non-chat /models entries and enable image input for models whose capabilities indicate vision support
Fixed
- Fixed long-context variant pricing to use billing.token_prices.long_context rates instead of default model pricing
- Fixed mapModel handling in OpenAI-compatible discovery so returning null now skips a model entry rather than falling back to defaults
- Fixed model ID precedence so a real upstream Copilot model id is kept when it conflicts with a synthesized -1m variant
@oh-my-pi/pi-coding-agent
Fixed
- Fixed stale Working… loader rows being committed to native scrollback above the live loader: the interactive status container now reports a live-region seam while it has mounted content (#2328 by @35844493).
- Fixed Mnemopi memory consolidation never running on session shutdown: MnemopiSessionState.dispose() now drains pending fact extractions and runs sleepAllSessions on every owned bank before closing handles (and AgentSession.dispose() awaits the result), matching the /memory enqueue slash command. mnemopiBackend.clear opts out via dispose({ consolidate: false }) so the destructive /memory clear path does not spend tokens consolidating memories that are wiped on the next line. consolidate() deliberately keeps no aliasOf short-circuit so /memory enqueue from a subagent still flushes and sleeps the parent's shared banks (the alias guard lives in dispose for lifecycle, not in consolidate for content). Without this, episodic_memory, gists, consolidation_log, graph_edges, and triples stayed empty for every deployment because the SHMR/beam pipeline only ran when a user typed /memory enqueue|rebuild (#2320).
- Fixed multi-path search collapsing distinct scopes into one walk rooted at their unrequested common ancestor: paths: [".", "~/.gitconfig"] scanned the entire home directory (typically until the 30s grep timeout), and explicit file entries inside walker-pruned directories (e.g. .git/config, gitignored files) silently never matched because they were folded into the directory walk's glob union. Multi-path scopes now fan out into per-target scans when the common ancestor is not itself a requested path, plain-file entries are read directly as their own targets, and matches from overlapping targets are deduplicated.
- Fixed lsp.formatOnWrite sending a hardcoded tabSize: 3, insertSpaces: true on every textDocument/formatting request, which silently re-indented 2-space YAML (and any LSP-formatted file) to 3-space on every write/edit through formatter-aware servers like yaml-language-server. Format options are now resolved per-file from .editorconfig (indent_size, indent_style, tab_width), falling back to the indent sniffed from the in-memory content the agent is about to write, then to a 2-space default. The duplicate DEFAULT_FORMAT_OPTIONS constant in lsp/index.ts and lsp/clients/lsp-linter-client.ts is replaced with a single shared resolveFormatOptions helper (#2329).
@oh-my-pi/pi-tui
Fixed
- Fixed the root compose letting a lower child's native-scrollback live seam overwrite a higher one: the topmost seam (and its commit-safe extension) now defines the commit boundary, so a stat
15.11.2 11 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- AgentTool.concurrency now also accepts a per-call resolver function (args) => "shared" | "exclusive", letting tools pick the scheduling mode from the call's arguments (a throwing resolver falls back to "exclusive")
Fixed
- Fixed whitespace-only error tool results so Anthropic requests no longer 400 with tool_result: content cannot be empty if is_error is true and wedge the session on every subsequent turn
@oh-my-pi/pi-ai
Fixed
- Fixed Anthropic encoding of error tool results with whitespace-only content so requests no longer 400 with tool_result: content cannot be empty if is_error is true
@oh-my-pi/pi-coding-agent
Added
- Added the Expert Elixir language server (expert, invoked as expert --stdio) to the built-in LSP server list, auto-detected for Mix projects (mix.exs/mix.lock). When both are installed, elixir-ls remains the primary navigation server (Expert is ordered after it).
- Added magicKeywords.enabled and per-keyword magicKeywords.ultrathink, magicKeywords.orchestrate, and magicKeywords.workflow settings to disable hidden magic-keyword notices and ultrathink auto-thinking escalation (#1796).
- Added external-editor support for Plan Review section annotations, preserving multiline feedback for Refine plan (#2305).
- Added plain-RPC slash command discovery with command source metadata and startup/update notifications (#2261).
Changed
- Bash tool calls in one assistant message now run in parallel instead of serializing the batch: non-pty bash is scheduled as a shared tool (pty: true stays exclusive), and overlapping calls on the same shell session key degrade to isolated one-shot shells so they cannot queue behind or abort each other
Fixed
- Fixed CDP attach target selection for Chromium/Edge endpoints that expose page targets through discovery before browser.pages() is populated, and improved ws:// app.cdp_url diagnostics (#2246).
- Fixed local memory consolidation on Responses-style models that reject user-only requests by sending a dedicated stage-two system prompt.
- Fixed extension, custom-tool, custom-command, and hook loaders injecting pi through bare @oh-my-pi/pi-coding-agent self-imports, which could pull a different cached package version into global installs during plugin load and trigger mixed-runtime stack overflows.
- Marked unsmoothed assistant streaming renders as transient so streamed code blocks can avoid synchronous syntax highlighting until the message finalizes.
- Routed LSP hover code rendering through the shared cached highlighter instead of calling the native highlighter directly on each render.
- Kept smooth assistant streaming renders transient until message_end so catch-up frames do not synchronously re-highlight still-growing code blocks.
- Fixed the job tool's poll header repeating the running count ("waiting on 19 of 19 19 running"): the title now reads "waiting on N jobs" (or "waiting on N of M jobs" when some settled) and the dim meta lists only settled categories (done/failed/cancelled)
- Fixed irc send await:true to a busy peer stranding the sender until timeout when async execution is disabled: the recipient (e.g. Main blocked in a synchronous task spawn awaiting the sender's own batch) can never reach a step boundary to reply, so it now generates an ephemeral side-channel auto-reply from its context (the pre-mailbox respondAsBackground behavior), records it as an irc:autoreply aside in its own history, and delivers it back over the bus with replyTo threading
- Fixed Windows-style backslash paths and globs being parsed as literal POSIX path segments by search/find tools (#2245).
@oh-my-pi/pi-tui
Fixed
- Fixed Ctrl+C/exit corrupting the parent shell on Windows: emergencyTerminalRestore() wrote \x1b[?1049l (leave alternate screen) unconditionally on every exit path, and conhost/Windows Terminal execute an unconditional cursor restore for it even when the alt buffer was never entered — with no prior save the cursor jumped to the viewport home, so the shell p
15.11.0 10 de juny del 2026 — 64 bits Portable
15.10.12 10 de juny del 2026 — 64 bits Portable
15.10.11 10 de juny del 2026 — 64 bits Portable
15.10.10 9 de juny del 2026 — 64 bits Portable
15.10.9 9 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added antigravityRankingStrategy and registered it as the default CredentialRankingStrategy for google-antigravity, so multi-account selection consumes the per-counter Antigravity usage reports (sorted ascending by remainingFraction in fetchAntigravityUsage) before falling back to round-robin — preventing the exhausted-counter credential from being chosen first when an unblocked sibling has headroom (#2187).
- Added Claude Fable 5 to the first-party Anthropic catalog, seeded directly via ANTHROPIC_CURATED_FALLBACK_MODELS rather than waiting on models.dev (1M context / 128k output, adaptive thinking, $10/$50 per MTok). The model parser recognizes the fable kind so effort tiers (low→max), adaptive thinking, and Opus-4.7-style sampling restrictions apply; token limits and pricing are pinned in applyAnthropicCatalogPolicy.
Fixed
- Fixed google-antigravity not rotating to another stored OAuth account when Cloud Code Assist returns 429 You have exhausted your capacity on this model. Your quota will reset after …. parseRateLimitReason matched the literal capacity before the quota will reset suffix and downgraded the failure to MODEL_CAPACITY_EXHAUSTED (45–75 s backoff), and isUsageLimitError returned false for the same message — so markUsageLimitReached was never invoked and the agent kept hammering the exhausted credential while the retry layer bailed on the multi-hour retry-after. Both paths now treat the Antigravity phrasing as QUOTA_EXHAUSTED / usage-limit, blocking the current credential until reset and letting the session pick an unblocked sibling (#2187).
- Fixed OpenRouter Anthropic chat-completions requests placing cache_control on empty assistant tool-call content. The cache marker now skips empty text and attaches to the most recent non-empty text part, avoiding HTTP 400 payloads with {type:"text", text:"", cache_control:...}.
- Fixed Fable-only Anthropic request shaping to cover Claude Mythos 5, and added Mythos 5 to the first-party Anthropic catalog seed. Adaptive display, sampling suppression, mid-conversation system messages, forced-tool-choice downgrade, and Bedrock adaptive metadata now handle both model families.
- Fixed adaptive-only Claude models (Opus 4.6+, Sonnet 4.6+, Fable/Mythos 5) returning HTTP 400 "thinking.type.disabled" is not supported for this model whenever thinking was turned off (utility calls and forced-tool turns route through the disable path). These models accept only thinking.type: "adaptive"; the request builder now omits the thinking field and pins the lowest adaptive effort instead of emitting type: "disabled".
- Widened the OpenAI-completions first-event watchdog floor from 120s to 300s for DeepSeek V4 reasoning models hosted on the official DeepSeek API. The reasoner emits no SSE bytes until its private chain-of-thought finishes, which routinely takes longer than the generic 100s first-event budget under load — every chat then aborted with OpenAI completions stream timed out while waiting for the first event and silently retried. Mirrors the existing GLM coding-plan widening (#2177).
@oh-my-pi/pi-coding-agent
Fixed
- Fixed streaming thinking (and other styled assistant content) vanishing from native scrollback once it scrolled past the viewport top during a foreground turn. The transcript's append-only commit detector compared raw row bytes, so a styled paragraph wrapping onto a new row (the span-closing SGR and width padding move while the visible cells stay identical) or a streamed token pushing the last word down a line flagged the block as permanently volatile — the commit boundary froze and every later row that crossed the viewport top was committed nowhere. Rows are now compared by visible content, a wrap-shrink of the in-flight bottom line counts as append-only, and a genuine one-off interior rewrite only suspends commits until the block re-earns append-only (30 clean frames), after which the pinned emitter backfills the stalled gap contiguously. Periodically
15.10.8 9 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added optional fetch overrides to SummaryOptions and compact/generateSummary so remote compaction can use custom HTTP clients
- Added optional fetch option to ProxyStreamOptions to control the HTTP request used by streamProxy
- Added optional fetch overrides to requestOpenAiRemoteCompaction and requestRemoteCompaction for injectable HTTP transport
- Added the upstream provider that served a request (AssistantMessage.upstreamProvider, e.g. OpenRouter's routed provider) as a pi.gen_ai.response.upstream_provider chat-span telemetry attribute, alongside the existing response id and time-to-first-chunk.
@oh-my-pi/pi-ai
Added
- Added optional fetch transport override (fetch?: FetchImpl) to Google, Ollama, and OpenAI-compatible model-manager options so dynamic model discovery and metadata lookups can use a caller-supplied HTTP client instead of only global fetch
- Added optional fetch on OAuth controller and API-key validation/login flows so token exchange, refresh, and device/PKCE login requests can be routed through a custom fetch implementation
- Added optional fetch support to usage polling context, allowing usage providers to execute usage checks using an injected HTTP client
- Added AssistantMessage.upstreamProvider, capturing the upstream provider an aggregator routed the request to (OpenRouter reports it via a top-level provider field on every chunk, e.g. "Anthropic"). Surfaced from the OpenAI-completions stream alongside responseId.
Fixed
- Fixed a degenerate OpenAI Codex stream (the model emits whitespace-only function_call_arguments.delta frames forever — commonly seen right after a todo tool call) terminating the turn with an error instead of recovering. The whitespace-loop circuit-breaker now (a) stops aborting the shared per-request AbortController — requestSignal is an AbortSignal.any over it, so aborting latched it and made every reopen on the reused requestSetup impossible — and (b) drops the half-built junk tool call and replays the request from scratch, bounded by CODEX_WHITESPACE_LOOP_RETRY_LIMIT (2). Sampling nondeterminism usually clears the loop on a fresh attempt; once the budget is exhausted the error is surfaced as before, but without the junk tool call polluting the message.
- Capped requested output tokens at 64k (OPENAI_MAX_OUTPUT_TOKENS, mirroring Anthropic's CLAUDE_CODE_MAX_OUTPUT_TOKENS) on OpenAI-family wires with a known upstream output cap — the openai-completions request builder (non-OpenRouter) and the shared responses sampling helper (openai-responses, azure-openai-responses). A model's catalog maxTokens often tracks its context window rather than the upstream's per-request output cap, so requesting the full ceiling 400'd (e.g. z-ai/glm-4.7 asking for 131072 output exceeded the upstream's 131072-token total context). Output is now min(requested, model.maxTokens, 64000).
- Stopped sending max_tokens/max_completion_tokens on OpenRouter (openrouter.ai) completions requests. OpenRouter filters out any upstream whose advertised output cap is below the requested max_tokens, so a value derived from the catalog (which reflects the highest-cap provider) silently excluded lower-cap upstreams — provider.order: ["cerebras"] for z-ai/glm-4.7 fell through to DeepInfra because Cerebras's ~40k output cap is below the request, while only: ["cerebras"] (no fallback target) bypassed the filter and worked. Omitting the field lets each upstream self-cap and keeps provider routing (only/order) honored. Kimi via OpenRouter stays exempt — it derives TPM rate limits from max_tokens.
@oh-my-pi/pi-coding-agent
Added
- Added an optional fetch option to CustomToolContext so custom tools can use a caller-provided HTTP implementation
- Added optional fetch overrides to ModelRegistry construction and MCP/web search/tool network calls, enabling callers to inject custom HTTP clients instead of relying on global fetch
- Added a bash.enabled setting to disable the model-facing bash tool while leaving
15.10.7 8 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Fixed
- Fixed first-party Anthropic requests returning HTTP 400 "Invalid signature in thinking block" after interrupting the model during its visible output. transformMessages stripped the signature from every thinking block of an aborted/error turn, including blocks that had already finished streaming — Anthropic delivers a block's signature at content_block_stop before the next block starts, so a thinking block followed by text/tool_use is fully signed. The valid signature was then replayed empty (signature: ""), which signature-enforcing Anthropic rejects, including when the provider is routed through an LLM gateway baseUrl. Only the single mid-stream block at the abort point is now treated as untrustworthy; completed thinking blocks keep their replayable signatures (#2144).
- Pinned a regression test against issue #2123: OAuth requests to adaptive-thinking Claude Opus models (4.6+) ship a context_management.edits[clear_thinking_20251015] block paired with the thinking field, but the eager-todo prelude (and other paths that force tool_choice to tool/any on the first user turn) route through disableThinkingIfToolChoiceForced, which would strip params.thinking while leaving the orphan context_management behind. The Anthropic API then rejected the request with 400 ... clear_thinking_20251015 strategy requires thinking to be enabled or adaptive. The fix that lands in [15.10.5] now drops both fields together; the new test locks the contract so the strategy can never outlive its enabling thinking payload again.
- Fixed Antigravity usage counters so exhausted Google/Gemini quota renders as 0% free while separate Anthropic/OpenAI-backed Antigravity model counters remain visible independently, without replaying stale pre-fix cached usage reports.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed MCP OAuth fallback rendering to show a short terminal hyperlink and keep the raw authorization URL on one unwrapped copy line (#2121).
- Fixed omp startup blocking 25–30 s on a single unresponsive MCP server when no cached tools were available for it. MCPManager.connectServers used to fall through to an unbounded Promise.allSettled over every still-pending server without a cached tool list, so one server stuck waiting on the per-request MCP timeout (OMP_MCP_TIMEOUT_MS, default 30 000 ms) gated the entire UI ready signal. Pending-without-cache servers are now left in flight: their tools surface via the existing background #onToolsChanged → refreshMCPTools path the moment the connect completes, and failures continue to log through the background catch handler (#2100).
What's Changed
- fix(ai): preserve completed thinking signatures when a turn is aborted mid-output by @roboomp in https://github.com/can1357/oh-my-pi/pull/2145
- fix(usage): separate Antigravity backend counters by @basedcorp99 in https://github.com/can1357/oh-my-pi/pull/2077
- fix(coding-agent): unblock omp startup when an MCP server stalls by @roboomp in https://github.com/can1357/oh-my-pi/pull/2140
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.10.6...v15.10.7
15.10.6 8 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added AIML API as an OpenAI-compatible provider preset with AIMLAPI_API_KEY discovery (#2105).
@oh-my-pi/pi-coding-agent
Added
- Added a /plan-review command that manually (re-)opens the plan-review overlay while plan mode is active. Since there is no fixed plan filename, it reviews the newest local://<slug>-plan.md the agent wrote — useful for pulling the review back up after dismissing it, or reviewing a plan the agent wrote without calling resolve.
Changed
- Reverted the task tool's result-header glyph from the ⇶ signature icon back to the quiet status.done bullet (•): white while a subagent is running, accent once it finishes. The reviewer verdict line and per-agent result lines use the same bullet.
Fixed
- Fixed /login API-key prompts (OpenCode Zen, Perplexity OTP, GitHub Enterprise URL, manual OAuth redirect URL, …) silently dropping pasted content on kitty/Linux/Wayland — and any other terminal supporting OSC 5522 enhanced paste. InputController enables kitty's enhanced clipboard protocol on TUI start and consumes the resulting OSC 5522 packets in an addInputListener that runs before focus dispatch, so the paste never reached the modal Input's bracketed-paste handler; the routing then stuffed the text into the main CustomEditor unconditionally, even when selector-controller had detached the editor and focused a temporary OAuth input. The pasted API key accumulated in the hidden editor and only resurfaced in the main prompt when the user dismissed the modal with Enter or Esc. The enhanced-paste callback now consults ui.getFocused() and routes the text to the focused component when it exposes a pasteText hook, falling back to the editor only when no modal target is in focus; image pastes refuse with a status message instead of stuffing a binary blob into the hidden editor. (#2127)
- Fixed an auto-compaction dead loop when compaction.strategy was shake and the configured threshold was low enough that a single shake pass could not bring the context below it (e.g. a 50K-token threshold on a session well above it). Each pass auto-continued, the next agent turn re-triggered the threshold check, and the second shake had nothing new to drop, so the session spun forever. The shake recovery path now estimates post-shake context and, when it is still above the threshold (or shake reclaimed nothing on overflow recovery), surfaces a one-shot warning and falls back to the summarization-driven context-full compaction so progress actually resumes (#2119).
- Fixed /skill: prompts so magic keywords and turn-budget directives in skill args inject the same hidden notices as normal user prompts, matching the editor highlight behavior (#2128).
- Fixed MCP OAuth fallback rendering to show a short terminal hyperlink and keep the raw authorization URL on one unwrapped copy line (#2121).
- Fixed the task tool rendering a success bullet and a success frame state for detail-less error results (e.g. an argument-validation failure that never executes): the header now shows the error glyph with an error border and error state, and surfaces the dispatched agent name.
- Fixed Agent Control Center new-agent creation so Windows Ctrl+Enter sequences submitted as a single LF generate the agent instead of inserting a newline (#2118).
- Fixed plan-mode subagents preserving read-only specialty tools such as report_finding while still stripping mutating tools (#1998).
- Removed unreachable standalone Exa tool-suite exports and stale tool-count barrel exposure while keeping the live Exa web_search provider helpers (#2093).
- Fixed omp commit split plans accepting hunk selectors that resolve to no parsed hunks, which crashed the apply step after the index reset and left the working tree fully unstaged (#2098).
@oh-my-pi/pi-tui
Added
- Added TUI.getFocused() accessor and Input.pasteText(text) method so callers consuming non-bracketed paste transports (e.g. kitty's OSC 5522 enhanced clipboard) can route a paste payload to the currently focus
15.10.5 8 de juny del 2026 — 64 bits Portable
15.10.4 8 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added anthropic-client-platform (desktop_app) and anthropic-client-version (1.11187.4) headers to the Anthropic request fingerprint for OAuth sessions
Changed
- Changed non-built-in tool names sent to Anthropic from proxy_ prefixing to _ prefixing (for example bash to _bash) while built-in tool names remain unchanged
- Updated the Anthropic OAuth stealth fingerprint to track Claude Code 2.1.165: claudeCodeVersion bumped to 2.1.165 (flows into both the cc_version billing header and the claude-cli/<version> user-agent), claudeCodeSystemInstruction changed to "You are a Claude agent, built on Anthropic's Claude Agent SDK.", and the billing-header cc_entrypoint changed from cli to local-agent.
- Clamped the Anthropic request max_tokens to Math.min(CLAUDE_CODE_MAX_OUTPUT_TOKENS, options.maxTokens || model.maxTokens) (64k) so OAuth requests match Claude Code's requested output cap instead of sending the model's full ceiling (e.g. 128k for Opus 4.8).
@oh-my-pi/pi-coding-agent
Added
- macOS release binaries are now signed with a Developer ID Application identity (hardened runtime + secure timestamp + JIT/library-validation entitlements) and notarized in CI when the APPLE_* signing secrets are configured; releases auto-fall back to ad-hoc signing until then. This makes the shipped binaries Gatekeeper-acceptable, unblocking an official Homebrew submission (#776). See docs/macos-signing-notarization.md.
- Added a Homebrew install path: brew install can1357/tap/omp. The can1357/homebrew-tap formula installs the prebuilt release binary, and a release_brew CI job regenerates it (version + per-asset sha256) from each published release via scripts/ci-update-brew-formula.ts (#776).
Changed
- Adjusted completion() model resolution so the default tier now prefers the session’s active model and falls back to the configured default role when needed
- Rewrote the session auto-title prompt (prompts/system/title-system.md) and the set_title tool description to ask for a concise, sentence-case title (3-7 words) that captures the session's topic/goal, with good/bad examples and explicit guidance to treat the first message as data (no following embedded links/instructions, no refusals, describe URL/reference asks). The local on-device title prompt (tiny-title-system.md) was aligned to the same 3-7 word, sentence-case convention. The deterministic greeting/low-signal filter and the none deferral sentinel are unchanged.
- Renamed the eval oneshot helper from llm() to completion() in both JavaScript and Python preludes, including status events, prompt docs, and runtime tests.
Fixed
- Fixed completion() to always send a non-empty default system prompt when system is omitted so providers that require instructions no longer reject requests
- Fixed structured completion() mode to return parsed JSON from plain text output when the model skips the forced respond tool call
- Fixed slow-tier completion() reasoning requests to avoid unsupported effort settings by only enabling reasoning on reasoning-capable models and capping effort to supported levels
- Fixed JS eval worker reset/dispose to close workers gracefully before forced termination, avoiding Bun 1.3.14 N-API teardown crashes with native modules such as canvas.
@oh-my-pi/pi-tui
Fixed
- Fixed Windows ConPTY session-resume painting the transcript with the last several rows truncated below the viewport until Alt+Tab forced a host repaint. After sessionReplace/historyRebuild/overlayRebuild paints that scroll-push content into native scrollback, the renderer now arms a 150 ms ConPTY settle window that coalesces spinner/blink-driven requestRender(false) calls into a single trailing render — Windows Terminal's viewport-follow logic no longer falls further behind the cursor on every tick of the post-paint storm. The arm also reclaims any render request queued during the in-flight composition (notably ImageBudget.endPass() calling requestRender() synchronously when a frame trips the live-gra
15.10.3 8 de juny del 2026 — 64 bits Portable
15.10.2 8 de juny del 2026 — 64 bits Portable
15.10.1 7 de juny del 2026 — 64 bits Portable
15.10.0 6 de juny del 2026 — 64 bits Portable
15.9.67 6 de juny del 2026 — 64 bits Portable
15.9.5 5 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Fixed
- Surfaced Anthropic stream failures whose message starts with Output blocked by conten as normal assistant error lifecycle events, so interactive clients render content-filter blocks instead of silently dropping the streaming bubble at agent_end.
@oh-my-pi/pi-coding-agent
Added
- Added a persistent error banner pinned above the editor when an assistant turn ends on a provider error (e.g. Anthropic's "Output blocked by content filtering policy"). The transcript Error: … line scrolls away as the conversation grows, so terminal turns that ended on a stream error could pass unnoticed; the banner stays in the fixed region above the input and is cleared when the next turn starts.
- Added bold, underlined, clickable [Image #N] placeholders in the draft editor and sent user-message bubbles, backed by extension-bearing blob-store sidecar files so terminal file:// links open in image viewers.
- Added the active model identifier (provider/id) to the system prompt's <workstation> block so the agent knows which model it is running as. Gated by the new includeModelInPrompt setting (default on); the base prompt is rebuilt on a mid-session model switch so the surfaced identifier stays current.
- Added OLLAMA_HOST support for implicit local Ollama discovery when OLLAMA_BASE_URL is unset, so OMP picks up the same host setting used by Ollama.
- Added OLLAMA_CONTEXT_LENGTH as a positive-integer context-window override for implicit local Ollama discovery, so users can correct OMP context budgeting without writing per-model overrides.
Changed
- Changed tools.discoveryMode to default to auto, which keeps discovery off for small tool sets and automatically switches to MCP-only tool discovery when more than 40 tools are registered.
Fixed
- Fixed user-message rendering to materialize image links from embedded image blocks when rebuilding chat output, so image placeholders remain clickable after replayed or restored messages
- Fixed queued/steering user messages carrying a pasted image rendering out of order — sometimes dropping the user bubble below the very tool output it was sent to steer. EventController.#handleMessageStart awaited async image-link materialization between the user message_start and addMessageToChat; since AgentSession.#emit dispatches TUI listeners fire-and-forget, that mid-handler yield let the next synchronously-handled events (assistant message_start, tool execution start/end) append their components first, scrambling transcript order and live-region block boundaries. The bubble is now appended synchronously, with clickable image links still materialized via the synchronous blob-store fallback.
- Fixed tool execution cards to finalize promptly when a turn is abandoned or completed so stale streaming previews and frozen spinner frames no longer keep transcript rows in the live region
- Fixed read and search TUI rendering to emit OSC 8 hyperlinks for HTTP URLs, local:// resources backed by files, and filesystem search targets, including line-specific links for search match rows.
- Fixed aborted streaming assistant messages staying frozen before their red "Operation aborted" label when status rows were appended underneath on ED3-risk terminals.
- Fixed omp / omp -c stacking a fresh welcome screen and transcript on top of the previous run's leftover terminal scrollback. The cold-launch transcript render was the only session-load path that did not pass clearTerminalHistory, so the TUI's scrollback-preserving initial paint left the prior run's welcome + conversation above the new one; the cold launch now clears native scrollback before painting, matching every in-process session switch.
- Fixed a long streamed assistant reply dropping its earlier lines on ED3-risk terminals (Ghostty/kitty/iTerm2) once it grew past the viewport — the head scrolled off the top and never reached scrollback, so the reply rendered as a ~viewport-tall circular buffer of only its latest lines. AssistantMessageComponent now repor
15.9.3 5 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Fixed
- Fixed @-mention auto-read injecting an unrelated, same-named file when a mention did not point at a real path — e.g. an npm scope like @scope/, a partial path, or a bare token. generateFileMentionMessages resolution previously fell back to prefix and repo-wide fuzzy matching (globbing the whole project on every such mention) and auto-read the single "best" guess. Resolution is now exact-only: a mention is auto-read only when it resolves to an existing file or directory; otherwise it is left as prose. The TUI @-selector already inserts the real, complete path before send, so post-send guessing was both unnecessary and the source of the wrong-file reads. Directories still resolve and are listed. Removes the per-mention **/* project scan.
@oh-my-pi/pi-tui
Fixed
- Fixed ED3-risk foreground streaming erasing the head of any block that alone overflows the viewport (a tall tool result drawn in one frame, or a multi-line assistant reply growing past the viewport as it streams). The live-region pin committed native scrollback only up to the sealed-prefix boundary (liveRegionStart), so rows of the live block that had physically scrolled above the viewport top were neither pushed into scrollback nor kept in the repainted viewport — they vanished. The commit boundary is now the viewport top: every row above the viewport enters scrollback (only the tail still visible in the viewport stays transient and deferred to the checkpoint).
- Fixed the same ED3-risk live-region pin duplicating already-committed scrollback rows when a foreground stream's live region collapsed mid-turn (a tool preview shrinking to its compact result, an assistant block re-wrapping shorter, a late tool completion). Because growth commits every row above the viewport top to native scrollback, a subsequent shrink moved the bottom-anchored viewport back across those committed rows and the repaint re-drew them into the viewport — so they appeared twice on scroll-up, and with no prompt-submit checkpoint to reconcile (autonomous multi-turn runs, or the session ending into the welcome screen) the duplicate was baked permanently into terminal history. The pinned repaint now separates commit geometry from repaint geometry: a collapse clamps the repaint to the committed sealed boundary (min(#scrollbackHighWater, liveRegionStart)) instead of re-exposing those rows, leaving native scrollback un-duplicated without emitting ED3 under a possibly-scrolled reader; stale mutable live-region saved lines still reconcile at the next checkpoint.
- Fixed hiding overlays during ED3-risk foreground streaming on unknown-viewport terminals leaving the overlay's transient rows in native scrollback. Overlay visibility reductions now bypass the streaming deferral path and rebuild once, so hidden dialog/notification sentinels are scrubbed immediately.
- Fixed ED3-risk / unknown-viewport terminals (including WSL fronted by Windows Terminal) keeping the foreground-stream eager-rebuild mode active after the stream had already settled. A later scrolled content shrink or resize-with-append could then bypass the anti-yank deferral and repaint from stale geometry, jumping the viewport or replaying the wrong rows. The eager opt-in now drops immediately when no teardown render is pending, and the one-frame post-checkpoint suffix-suppression path no longer overrides geometry reflow handling.
What's Changed
- fix(robomp): parameterize bot identity in system prompts by @roboomp in https://github.com/can1357/oh-my-pi/pull/1933
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.9.2...v15.9.3
15.9.1 4 de juny del 2026 — 64 bits Portable
15.9.0 4 de juny del 2026 — 64 bits Portable
15.8.3 3 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added getReadToolPath(context) to @oh-my-pi/pi-agent-core/compaction/tool-protection to extract a paired read tool call's path for embedders building read-targeted protection matchers
- Added getReadToolPath(context) to @oh-my-pi/pi-agent-core/compaction/tool-protection: the shared primitive that extracts a paired read tool call's path argument, so embedders can build their own read-targeted compaction protection matchers (e.g. plan-file reads) the same way isSkillReadToolResult does.
@oh-my-pi/pi-coding-agent
Fixed
- Fixed Jujutsu workspace detection failing in non-default workspaces created by jj workspace add, whose .jj/repo is a FILE pointing at the shared repo dir rather than a directory. Detection now matches jj's own criterion (.jj/repo present, file or dir) instead of requiring a .jj/repo/store directory, and jj.repo.resolve's storeDir follows the file indirection to the shared store.
Changed
- Changed the todo-write prompt to require initializing every item from a user-supplied multi-step plan as an individual todo task before execution
- Changed context compaction (prune/shake) to protect reads of the active plan file the same way it already protects skill:// reads, so the plan stays intact through automatic and manual compaction. Both the canonical local://PLAN.md alias and the session's current plan reference path (e.g. a titled local://<title>.md after approval) are kept, tolerating read selectors and local:/ scheme spelling.
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.8.2...v15.8.3
15.8.2 3 de juny del 2026 — 64 bits Portable
15.8.0 2 de juny del 2026 — 64 bits Portable
15.7.6 1 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Added ask option descriptions so agents can keep short labels and render explanatory text as separate muted rows in the selector.
- Added an extension API for rendering supplemental UI below visible assistant thinking blocks.
- Added default-on lsp.diagnosticsDeduplicate support so post-edit LSP diagnostics already shown for a file are suppressed within the session and only new or changed diagnostics are surfaced.
Fixed
- Fixed a single ESC press both dismissing the @/slash autocomplete popup and aborting the running agent operation. ESC now drains the popup first; only when no popup is visible does it route to the global interrupt handler (matching the standard TUI/IDE pattern). The shouldBypassAutocompleteOnEscape editor hook is removed — it had become the trigger for this bug (#1655).
- Fixed Claude Code slash command discovery to load subdirectory commands recursively while preserving basename commands (e.g. /apply) and adding namespace aliases (e.g. /opsx:apply) for tools that install colon-namespaced workflows (#1523).
- Fixed the eval tool's per-cell timeout killing cells that were not stalled. The timeout is now a plain wall-clock budget on the cell's own work that is paused only while a host-side agent()/parallel()/llm() bridge call is in flight — those calls pump a heartbeat that re-arms the watchdog, so a long fanout or a slow (e.g. reasoning-tier) completion runs to completion instead of being aborted mid-flight (a subagent's time-to-first-token, a long quiet nested tool, or an entire oneshot llm() request no longer trip it). Nothing else re-arms the budget: ordinary compute, print/stdout, log()/phase(), and non-agent tool calls all count against it, so a cell that is not delegating to an agent/llm is bounded by the regular wall-clock timeout (and the timeout message no longer says "of inactivity"). The heartbeat is a pure keepalive — never persisted or rendered.
@oh-my-pi/pi-tui
Fixed
- Fixed native Windows + Windows Terminal freezing the editor on the wrap keystroke, on /plan//resume/model-switch/status-line toggles, and on any other offscreen structural mutation until the next prompt submit. The 15.7.5 #1635 fix routed every viewport-saturating pure-append and structural mutation through deferredMutation (a literal no-op) whenever isNativeViewportAtBottom() returned undefined — which it always does under WT_SESSION because the kernel32 probe can't see WT host scrollback. The deferral was only ever meant for the confirmed-scrolled case; an unknown viewport now falls back to a non-destructive viewportRepaint instead, so the live UI keeps updating without emitting \x1b[3J and without yanking a possibly-scrolled reader. Confirmed-scrolled frames (probe returns false) still defer.
Fixed
- Removed the hard-coded 20-result cap on @-prefixed fuzzy file completion in CombinedAutocompleteProvider.#getFuzzyFileSuggestions. The dropdown now honors the existing maxResults: 100 ceiling already configured for fuzzyFind, so projects with many files sharing a common stem (e.g. @controller, @test) surface all relevant matches instead of being silently truncated. (#1652)
What's Changed
- fix(ai/anthropic): request current Claude Code OAuth scopes by @ephraimduncan in https://github.com/can1357/oh-my-pi/pull/1631
- Improve ask option rendering by @rimless-casualty in https://github.com/can1357/oh-my-pi/pull/1629
- feat(coding-agent): add assistant thinking renderers by @shoucandanghehe in https://github.com/can1357/oh-my-pi/pull/1577
- fix(agent): compact before oversized prompts by @roboomp in https://github.com/can1357/oh-my-pi/pull/1619
- feat(theme): make spinner frames overridable per custom theme by @roboomp in https://github.com/can1357/oh-my-pi/pull/1555
- fix(coding-agent): raised auto-discovered model maxTokens cap to 32K by @roboomp in https://github.com/can1357/oh-my-pi/pull/1529
- fix(coding-agent): resolve Claude subdirectory slash commands by @roboomp in https://github.com/can1357/oh
15.7.5 1 de juny del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Added
- Added Anthropic task budget support, forwarding taskBudget as output_config.task_budget with the required task-budgets-2026-03-13 beta header and accepting Anthropic gateway requests that send output_config.task_budget.
Fixed
- Fixed OpenAI-family first-event timeouts so PI_OPENAI_STREAM_IDLE_TIMEOUT_MS cannot be undercut by a lower generic PI_STREAM_FIRST_EVENT_TIMEOUT_MS while local OpenAI-compatible servers are still processing large prompts. PI_OPENAI_STREAM_FIRST_EVENT_TIMEOUT_MS is now available for an explicit OpenAI-specific first-event override. (#1603)
@oh-my-pi/pi-coding-agent
Fixed
- Fixed streaming assistant responses leaving duplicated tail rows in WSL/Windows Terminal scrollback by enabling eager native-scrollback rebuilds while assistant text is actively streaming (#1615).
Fixed
- Fixed the task tool mangling subagent prompts when a model double-JSON-encodes a string argument: context and each task's assignment/description are now repaired when they arrive uniformly double-escaped (literal \n, \", \uXXXX), so the subagent receives the intended prose and the call preview renders real newlines. The repair is guarded by a JSON-string round-trip and a double-encode signature, so legitimate backslashes/quotes (Windows paths, regexes, embedded quotes) are left untouched, and it is scoped to these natural-language fields only (never code-bearing tools).
Fixed
- Fixed pr:// PR views omitting formal review submissions and approvals when comments are enabled (#1600).
Fixed
- Fixed subagent yield-reminder loop logging benign user/compaction aborts as ERROR. The catch around session.prompt/waitForIdle in task/executor.ts now demotes ToolAbortError and signal-aborted exits to debug and keeps ERROR for genuine prompt failures only (#1623).
Fixed
- Fixed read local://<file> resolving to the wrong session's artifacts directory in multi-session ACP hosts (e.g. cmux). LocalProtocolHandler.resolve now honors context.localProtocolOptions supplied by the calling tool before falling back to the process-wide override or the first main-kind session in the global AgentRegistry; read, find, search, ast_grep, and ast_edit thread their session's options through so a local://PLAN.md lookup hits the calling session's local root instead of a sibling session's (#1608).
Fixed
- Fixed omp segfaulting on exit on Windows after the tiny title/memory model loaded onnxruntime-node (issue #1606). The tiny model now runs in a Bun subprocess instead of a Worker thread, so the NAPI finalizer that crashes during shutdown never executes in the agent's address space; the subprocess is SIGKILL'd on dispose to skip every native destructor on every platform.
Fixed
- Fixed unbounded MCP reconnect loop that could fork-bomb the host when a stdio MCP server completes the initialize/tools/list handshake and then exits. MCPManager now enforces a per-server crash circuit breaker (5 reconnects per 30 s window) on the automatic transport.onClose path; manual /mcp reconnect resets the window so users can recover after fixing the misconfiguration (#1592).
@oh-my-pi/pi-tui
Fixed
- Fixed native Windows + Windows Terminal scrollback being yanked to the top when a streaming response triggered a TUI full redraw. Under ConPTY the kernel32 GetConsoleScreenBufferInfo probe answers about the pseudo-console (always at the buffer tail) and not about WT's host scrollback, so isNativeViewportAtBottom() falsely returned true while the user was scrolled up and the shrink-across-viewport branch issued a destructive historyRebuild (\x1b[2J\x1b[H\x1b[3J). The probe now short-circuits to undefined whenever WT_SESSION is set, letting the existing deferred-rebuild path keep streaming-time mutations non-destructive and reconcile native history at the next prompt-submit checkpoint. (#1635)
What's Changed
- fix(lsp): shut down servers with exit notification by @roboomp in https://github.com/can1357/oh-my-pi/pull/1594
- fix(mcp): cap automatic reconnect bursts to prevent
15.7.3 31 de maig del 2026 — 64 bits Portable
15.7.2 31 de maig del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Added
- Added providers.autoThinkingModel setting so users can choose the auto thinking classifier backend (online smol or local tiny-memory model)
- Added an auto thinking level that classifies each real user turn and resolves to a concrete low-through-xhigh effort, with online smol classification by default and an opt-in local on-device classifier.
Changed
- Updated the interactive thinking selectors in model/model-role pickers and ACP thinking options to include auto as a selectable level
- Updated footer and status-line rendering to show auto while auto-thinking is being resolved and auto → <level> once it resolves
- Changed the local tiny-model device default to CPU on every platform; explicit providers.tinyModelDevice / PI_TINY_DEVICE values still opt into accelerated ONNX providers.
Fixed
- Prevented auto-thinking classification from running on non-user synthetic turns and non-reasoning models, keeping the session on its provisional concrete effort
- Added a bounded auto-thinking classification path that falls back to the provisional effort on failures/timeouts so prompts continue without interruption
- Bypassed auto classifier for ultrathink prompts and resolved directly to the highest supported auto effort
- Fixed the JavaScript eval kernel crashing the whole process with a segfault (SIGTRAP, getImportedModule on a null record) when imported code reached a local module whose relative-import graph contains a cycle — e.g. await import("…/edit/streaming.ts"), or any workspace path with cyclic re-exports. The LocalModuleLoader linked and evaluated each local module individually inside the recursive vm.SourceTextModule linker callback, which re-entered Bun's node:vm module linker mid-instantiation and detonated JSC on the first cycle. The loader now constructs the entire local module graph first and drives a single link() + evaluate() from the graph root, so cyclic graphs instantiate in one pass; external (node_modules) modules stay eagerly loaded since they carry no imports and cannot form a cycle.
- Fixed the streaming edit preview rendering a blank box for hashline edits whose payload sits on the trailing in-flight line (the common single-op replace/insert case). The preview path trimmed that still-typing line before diffing, so a single-payload op collapsed to a "No changes" result — shown as an empty box — for almost the entire stream. Hashline previews now feed the raw in-flight text through applyPartialTo, whose streaming-tolerant parser drops a payload-less trailing op and projects a partially-typed payload line as it grows, so the diff appears and fills in live. Transient errors from the actively-typed trailing section are also suppressed while streaming (regardless of section count) so a mid-typed op can't wipe an already-good preview frame; real errors still surface once args are complete.
@oh-my-pi/pi-mnemosyne
Fixed
- Fixed Windows startup crashes by keeping fastembed's older ONNX Runtime binding lazy until local embeddings are used.
- Fixed a segfault at startup from eagerly loading fastembed: importing the embeddings module pulled in fastembed, which eagerly loads the onnxruntime-node native addon. The import is now deferred until a local fastembed model is actually initialized, so API-model, disabled-embeddings, and test runtimes never load the native addon.
What's Changed
- fix: prevent Windows ORT preload crash by @segmentationf4u1t in https://github.com/can1357/oh-my-pi/pull/1561
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.7.1...v15.7.2
15.7.1 31 de maig del 2026 — 64 bits Portable
Notes de la versió
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.6.0...v15.7.1
15.6.0 30 de maig del 2026 — 64 bits Portable
15.5.15 30 de maig del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-agent-core
Added
- Added maxToolCallsPerTurn to AgentLoopConfig/AgentOptions, allowing callers to cut a streamed assistant turn after a completed tool-call batch and execute the runnable partial turn instead of waiting for the provider to yield.
Fixed
- Normalized maxToolCallsPerTurn to accept only positive integer limits, with non-finite or non-positive values treated as disabled
@oh-my-pi/pi-ai
Added
- Added PI_REQ_DEBUG=1 request/response recording for provider transports. Each request writes rr-session-N.json; each received response writes rr-session-N.res.log with response headers followed by raw body bytes.
Fixed
- Fixed OpenCode-Go dynamic model refresh downgrading qwen3.7-max from Anthropic Messages to OpenAI-compatible transport, which caused 401 Model qwen3.7-max is not supported for format oa-compat after /v1/models cache refreshes.
@oh-my-pi/pi-coding-agent
Changed
- Enabled the agent loop's tool-call batch cap for Anthropic Claude sessions, cutting oversized streamed tool-use bursts into runnable batches before continuing the conversation.
Removed
- Removed the calc tool (deterministic arithmetic evaluator) and its calc.enabled setting. The model can compute via eval instead.
Fixed
- Fixed Anthropic Claude tool-call batching to clear and reapply the Claude-specific batch cap whenever the session model changes
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.5.14...v15.5.15
15.5.12 29 de maig del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-ai
Removed
- Removed ANTML stream markup healing for antml:function_calls and antml:thinking envelopes, so Anthropic-compatible providers no longer parse those tags into toolCall/thinking events
Fixed
- Fixed GLM-5.x coding-plan OpenAI-compatible streams to use a longer default watchdog window, avoiding spurious OpenAI completions stream stalled while waiting for the next event errors during slow glm-5.1 thinking/output phases. (#1494)
- Fixed zhipu-coding-plan model discovery and credential validation to use the dedicated GLM Coding Plan endpoint (https://open.bigmodel.cn/api/coding/paas/v4) instead of the general BigModel endpoint, preventing requests from consuming ordinary account balance. (#1494)
- Fixed DeepSeek tool calls failing on NanoGPT (e.g. nanogpt/deepseek/deepseek-v4-pro with reasoning enabled) by routing tool-bearing DeepSeek requests through NanoGPT's :tools model route and adding nanogpt to the DSML leak allowlist so streamed <|DSML|tool_calls>...</|DSML|tool_calls> envelopes are healed into structured tool calls instead of being passed through as visible text. (#1488)
- Fixed DeepSeek tool calls failing on NanoGPT (e.g. nanogpt/deepseek/deepseek-v4-pro with reasoning enabled) by adding nanogpt to the DSML leak allowlist so streamed <|DSML|tool_calls>...</|DSML|tool_calls> envelopes are healed into structured tool calls instead of being passed through as visible text. The :tools model suffix is no longer appended on NanoGPT; that route triggered NanoGPT's server-side tool-call parser and 502'd with code: "malformed_tool_call" on complex tool schemas (todo_write) — the default route forwards delta.content (including DSML envelopes) which is healed client-side. (#1488)
- Fixed OpenAI-compatible streamed parallel tool calls losing indexed argument deltas by tracking active tool-call blocks by the provider's tool_calls[].index; this keeps parallel NanoGPT read calls from merging or dropping their path arguments. (#1488)
@oh-my-pi/pi-coding-agent
Added
- Added the omp-plugins discovery provider, which scans every extension package directory configured via extensions: (in ~/.omp/agent/settings.json or <cwd>/.omp/settings.json) or --extension/-e on the CLI for skills/, hooks/pre|post/, tools/, commands/, rules/, prompts/, and .mcp.json. Prior to this, only the extension's TypeScript factory module ran; every sibling capability the docs (https://omp.sh/docs/extension-authoring) advertised was silently ignored (#1496).
- Added the top-level omp install <target> subcommand documented at https://omp.sh/docs/extension-authoring. Local paths route to omp plugin link (so the directory is symlinked into the plugin set), and npm/marketplace specs route to omp plugin install. Before this, install was not a registered subcommand and the CLI runner silently forwarded install ./my-extension to launch as an initial LLM prompt (#1496).
Changed
- Changed the sticky Todos panel above the editor to advance as tasks close, instead of pinning to the first 5 tasks of the active phase. selectStickyTodoWindow now shows up to 5 open (pending / in_progress) tasks in original phase order and reports the count of remaining open tasks for the +N more hint, so every todo_write flip produces a visible row shift. Closed-phase tail falls back to the last 5 tasks (with the +N more line suppressed) until getActivePhase walks to the next phase.
- Linked the sticky Todos panel to the live SessionObserverRegistry so pending todos that have an in-flight subagent doing their work light up green with an animated spinner — the same theme.spinnerFrames ("status" preset) the task tool uses for its agent rows — instead of staying greyed out as if nothing is happening. A new exported todoMatchesAnyDescription(content, descriptions) does case- and whitespace-insensitive equality first with a 6-char minimum-overlap substring fallback in either direction, so "Sonnet #2: shallow bug scan" and a subagent description of "Sonnet #2" still link up. Comple
15.5.8 28 de maig del 2026 — 64 bits Portable
15.5.3 27 de maig del 2026 — 64 bits Portable
Notes de la versió
@oh-my-pi/pi-coding-agent
Breaking Changes
- Disallowed inline payload on hashline ↑, ↓, and : operations (including BOF/EOF inserts), requiring payload text to be supplied on standalone + continuation rows
Changed
- Warned when legacy inline LINE:TEXT lines are accepted as payload continuations only when inside a pending multi-line A-B: replacement
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.5.2...v15.5.3
15.3.2 25 de maig del 2026 — 64 bits Portable
Notes de la versió
What's Changed
- docs: correct keybinding config file reference by @roboomp in https://github.com/can1357/oh-my-pi/pull/1361
- fix(keys): named keys win over colliding ctrl+letter in legacy mode by @roboomp in https://github.com/can1357/oh-my-pi/pull/1362
- fix(coding-agent/tools): stop re-walking JTD-converted JSON Schema by @roboomp in https://github.com/can1357/oh-my-pi/pull/1363
- fix(lsp): support config-only marketplace servers by @roboomp in https://github.com/can1357/oh-my-pi/pull/1364
- fix(ai): repair orphan function_call_output in openai-responses replay by @roboomp in https://github.com/can1357/oh-my-pi/pull/1370
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.3.1...v15.3.2
15.3.1 25 de maig del 2026 — 64 bits Portable
Notes de la versió
What's Changed
- docs: add LSP configuration guide by @roboomp in https://github.com/can1357/oh-my-pi/pull/1357
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.3.0...v15.3.1
15.3.0 25 de maig del 2026 — 64 bits Portable
Notes de la versió
What's Changed
- fix(clipboard): route image reads through powershell.exe on WSL by @roboomp in https://github.com/can1357/oh-my-pi/pull/1281
- Append only mode for Deepseek Enjoyers by @H4vC in https://github.com/can1357/oh-my-pi/pull/1344
- Add DeepSeek API provider login by @watzon in https://github.com/can1357/oh-my-pi/pull/1181
- fix(xiaomi): use token-plan base URL for tp- API keys by @hanbinnoh in https://github.com/can1357/oh-my-pi/pull/1336
- fix(session): handle EPERM during session rewrite by @roboomp in https://github.com/can1357/oh-my-pi/pull/1339
- fix(loop-mode): block auto-submit while post-prompt background work is pending by @roboomp in https://github.com/can1357/oh-my-pi/pull/1297
- fix: preserve slash command text in editor history when plan/goal mode is already active by @yahyasaqban-lab in https://github.com/can1357/oh-my-pi/pull/1303
- feat: add OMP_NO_WEBP env var to exclude WebP from image resize by @cdanis in https://github.com/can1357/oh-my-pi/pull/1315
- feat: add autoloadSkills frontmatter field for agent definitions by @TommyC81 in https://github.com/can1357/oh-my-pi/pull/1335
- fix(coding-agent): preserve startup extension notifications by @roboomp in https://github.com/can1357/oh-my-pi/pull/1317
- fix(cli): allow /goal set to replace active goals by @roboomp in https://github.com/can1357/oh-my-pi/pull/1296
- fix(openai-responses): use canonical instructions field for non-developer-role endpoints by @roboomp in https://github.com/can1357/oh-my-pi/pull/1285
New Contributors
- @hanbinnoh made their first contribution in https://github.com/can1357/oh-my-pi/pull/1336
- @yahyasaqban-lab made their first contribution in https://github.com/can1357/oh-my-pi/pull/1303
- @cdanis made their first contribution in https://github.com/can1357/oh-my-pi/pull/1315
- @TommyC81 made their first contribution in https://github.com/can1357/oh-my-pi/pull/1335
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.2.4...v15.3.0
15.2.4 22 de maig del 2026 — 64 bits Portable
Notes de la versió
What's Changed
- fix(ai): prevent Codex OAuth 403 from port-fallback redirect_uri mismatch + device-code flow by @roboomp in #1278
Full Changelog: https://github.com/can1357/oh-my-pi/compare/v15.2.3...v15.2.4

El més popular a Assistents d’IA i LLM

Mostra-ho tot
  1. Claude CodeAssistents d’IA i LLM · Anthropic PBC
  2. Codex CLIAssistents d’IA i LLM · OpenAI, Inc.
  3. ChatGPT by lencxAssistents d’IA i LLM · lencx
  4. GPT4AllAssistents d’IA i LLM · Nomic
  5. JanAssistents d’IA i LLM · Jan
  6. oh-my-piAssistents d’IA i LLM · Can Bölük
  7. T3 CodeAssistents d’IA i LLM · T3 Tools
  8. WSL ManagerAssistents d’IA i LLM · Bostrot
  9. CC SwitchAssistents d’IA i LLM · farion1231
  10. LobeHubAssistents d’IA i LLM · LobeHub

Etiquetes

Comparteix