MiMoCode کے پرانے ورژن – تمام ورژنز

Summary
Interrupted or errored turns stay recoverable: the incomplete assistant message is kept as a candidate, and /recover continues the turn append-only without rewriting history.
Conversation titles can be generated through a dedicated API (POST /experimental/title, plus a genTitle helper) that falls back to a lite model when the preferred one is unavailable or returns invalid output.
TUI voice input is upgraded to a tool-based voice_input protocol: ASR applies caret/selection-aware insert or set, submit stays explicit through the send field, and the call shape is validated so models can drive it reliably.
Also in this release: compaction keeps a summary plus only the API rounds that arrive while that summary is generated, triggers at a flat 90% of the context window, and collapses the post-rebuild tail into an activity log; unused actor spawn/run resume arguments removed so models are pointed at send (already the supported path for waking and continuing a subagent); subagent actor use via exec limited to send; nested exec tool calls persisted and replayable in the TUI; same-session subagents inherit parent permission grants; per-session system prefix frozen in a snapshot for stable prompt-cache prefixes; optional auto_worktree notice (default off); skills updates for mate, pdf-official CJK fonts, and pptx-official image sourcing; plus provider/retry, MCP, compose-next, and build fixes.

中断或出错的回合仍可恢复：未完成的 assistant 消息会保留为候选，/recover 以 append-only 方式续跑，不改写历史。
会话标题可通过专用 API 生成（POST /experimental/title 与 genTitle helper）；首选模型不可用或返回无效结果时，会回退到 lite model。
TUI 语音输入升级为基于工具的 voice_input 协议：ASR 按光标/选区做 insert 或 set，发送始终由 send 字段显式触发；调用 schema 经过校验，模型可以稳定驱动。
本版本还包括：compaction 保留 summary 以及生成期间到达的 API 轮次，触发条件改为上下文窗口的固定 90%，重建后的 tail 收成活动日志；删除 actor spawn/run 中未实现的 resume 参数，把模型导向本来就能用的 send；子代理经 exec 使用 actor 时仅允许 send；嵌套 exec 调用会持久化且可在 TUI 回放；同会话子代理继承父级权限；会话 system prefix 冻结进快照以保持 prompt-cache 前缀稳定；可选的 auto_worktree 提示（默认关闭）；mate、pdf-official 中日文字体与 pptx-official 配图规则等 skill 更新；以及 provider/重试、MCP、compose-next 与构建相关修复。
Features
- tui: upgrade voice input to the tool-based voice_input protocol (caret/selection insert/set, explicit send, validated schema) by @yanyihan-xiaomi in #2308, #2309
- session: project compaction as a summary plus compression-time tail; fire at a flat 90% of the context window; add MIMOCODE_COMPACTION_MAX_CONTEXT env default by @MiMoHardFather in #2294, #2263, #2260
- session: support append-only interrupted turn recovery (/recover, TUI recover action); keep errored assistants as recovery candidates by @wqymi in #2205, #2279
- session: collapse rebuild tail into a compact activity log instead of replaying verbatim tool results by @wqymi in #2275
- session: freeze the per-session system prefix in a persistent snapshot; move skill catalogs from user messages to the system tail; preserve prompt cache prefixes by @MiMoHardFather in #2289, #2287, #2258
- sdk: expose reliable POST /experimental/title / genTitle with lite-model fallback and write guards by @wqymi in #2228
- actor: drop unimplemented spawn/run actor_id resume arguments; keep send as the path for waking and continuing a subagent by @yanyihan-xiaomi in #2262
- actor: limit subagent actor calls through exec to send only by @wqymi in #2217, #2223
- exec: persist replayable nested tool parts (versioned snapshots, TUI renderers, debounce) by @yanyihan-xiaomi in #2222
- permission: same-session subagents inherit parent grants (fall back to current session id for inherit parent) by @wqymi in #2273
- config: add auto_worktree toggle (default off); path-based once-per-session main-worktree mutation notice by @wqymi in #2227, #2270, #2292
- skill: add mate builtin skill for desktop pet creation by @oscar-cao7 in #2225
- skill: pdf-official CJK font protocol (per-OS TTF ladder, CID STSong-Light fallback, never Helvetica) by @oscar-cao7 in #2189
- skill: pptx-official image sourcing decision rules (draw / cite / search / generate; limited image_gen budget) by @oscar

What's Changed
- fix(runtime): use Node APIs for hashing and FTS queries in core by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2110
- chore: drop the dead dev branch and the beta release path by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2111
- fix(flag): read MIMOCODE_AUTO_APPROVE_DELETE lazily so embedders can toggle it at runtime by @JinyuXiang-Mimo in https://github.com/XiaomiMiMo/MiMo-Code/pull/2114
- feat(permission): make the delete-ask exemption instance-scoped, not process-global by @JinyuXiang-Mimo in https://github.com/XiaomiMiMo/MiMo-Code/pull/2115
- fix(provider): replay empty signed Anthropic thinking blocks by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2117
- fix(permission): let yolo auto-approve deletes by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2120
- feat(checkpoint): add disable switch by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2121
- fix(id): extend identifier timestamps to 64 bits before v1 wraps by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2122
- fix(checkpoint): hide disabled lifecycle prompt copy by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2124
- fix(checkpoint): hide disabled lifecycle prompt copy by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2130
- fix(session): stabilize MCP tool ordering by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2131
- fix(session): gate dynamic system prompt content behind flag by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2132
- fix(skill): add MIMOCODE_DISABLE_AGENTS_SKILLS flag by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2133
- fix(skill): dedupe skills catalog in model messages and cache global discovery by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2134
- perf(prompt): keep system environment cache-stable by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2119
- fix(agent): prevent subagents from spawning nested subagents by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2139
- fix(session): don't tell subagents to call the masked actor tool by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2140
- fix(provider): enforce Anthropic image dimension limits by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2146
- fix(tui): bash_delete prompt — stale glyph leak, dropped deletion lines, sizing by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2145
- fix(session): roll back the compaction boundary when no summary lands by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2148
- fix(codex): remove 372K context cap for GPT models by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2149
- feat(permission): decouple skip-all from permission ask timeout by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2151
- feat(server): OpenAI-compatible capability API on every session by @wqymi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2147
- fix: remove hasSideEffect for OpenAI Responses API compatibility by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2152
- feat(exec): expose all request-authorized MCP tools to exec without search gating by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2153
- feat(exec): add ALL_TOOLS catalog and transpile scripts with TypeScript by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2154
- feat: add MIMOCODE_CODEX_MODE flag and unify system prompts by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2157
- feat(runtime): isolate child process environments by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2159
- fix(gpt): use normal prompt for mimo-v2.5 models by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2161
- Update SDK types and OpenAPI spec by @yanyihan-xiao

What's Changed
- chore(sdk): sync generated types missed by earlier commits by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2078
- fix(release): widen npmmirror sync window, drop auto-retry, exit non-zero on failure by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2080
- feat(cli): add --yolo permission bypass alias by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2082
- docs: align contribution and security guidance by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2079
- feat(bash): exempt temp-scoped deletes from the forced-ask confirmation by @JinyuXiang-Mimo in https://github.com/XiaomiMiMo/MiMo-Code/pull/2099
- fix(session): retry GPT server overload errors by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2100
- fix(skill-search): blacklist reminder models by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/2103
- fix(auth): stop handing user credentials to child processes by @yuyang27 in https://github.com/XiaomiMiMo/MiMo-Code/pull/2101
- fix(compose): make Compose Next model-invocable by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/2102
- chore: bump version to 0.1.12 by @qiaozongming in https://github.com/XiaomiMiMo/MiMo-Code/pull/2105
New Contributors
- @yuyang27 made their first contribution in https://github.com/XiaomiMiMo/MiMo-Code/pull/2101
Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.11...v0.1.12

Summary
The headline of this release is memory.disable_write — a switch that stops MiMoCode from writing memory while leaving reads intact. With it on, no checkpoints, MEMORY.md, notes.md or task progress.md are written and dream/distill stop running automatically; nothing is ever deleted and the builtin memory search tool keeps working. One caveat we corrected in the same release: existing memory is not auto-loaded while writing is off — the dumps normally reach context through a checkpoint rebuild, which is short-circuited by design when no new checkpoint can be written, so the switch's wording no longer promises automatic availability.
Also notable: a patched @ai-sdk/openai-compatible fixes MCP tools with more than two parameters silently losing their arguments — the stream parser finalized a tool call as soon as the accumulated arguments first parsed as valid JSON, so providers that emit complete-JSON snapshots (e.g. OpenRouter) had all later fragments dropped. Plus a new memory-search builtin skill (querying the SQLite trajectory DB) and an MCP refactor that makes the single-client-per-process property structural instead of relying on layer memoisation.

本次发布的主角是 memory.disable_write —— 一个只停写、不停读的记忆开关。开启后不再写入 checkpoint、MEMORY.md、notes.md 和任务 progress.md,dream/distill 也不再自动运行；任何文件都不会被删除，内置 memory 检索工具照常可用。同一版本里我们同时修正了一处措辞：停写期间既有记忆不会自动加载——记忆内容平时是靠 checkpoint 重建进入上下文的，而停写时无法写新 checkpoint,重建会按设计短路并降级为压缩，因此开关文案不再承诺"自动可用"。
另一个重点：通过给 @ai-sdk/openai-compatible 打补丁，修复了参数多于两个的 MCP 工具静默丢参数的问题——流解析器一旦发现累积的 arguments 首次能解析为合法 JSON 就提前完结该工具调用，导致以完整 JSON 快照方式推流的 provider(如 OpenRouter)后续片段全部被丢弃。此外还新增 memory-search 内置技能(查询 SQLite 轨迹库),以及一次 MCP 重构：把"每进程单一 client"从依赖 layer memo 的巧合变成图上可见的结构性保证。
Features
- memory: add the memory.disable_write switch — stop memory writes, keep reads by @wqymi in #2040
- skill: add memory-search builtin skill — SQLite trajectory DB query guide by @JinyuXiang-Mimo in #2046, #2053
Bug Fixes
- mcp: patch @ai-sdk/openai-compatible to stop dropping multi-arg tool calls by @YOMXXX in #2054
- memory: stop promising memory loads itself while writing is off by @wqymi in #2071
- agent: align subagent prompts with runtime tool schemas by @MiMoHardFather in #2039
- session: scope diffs to requested message by @MiMoHardFather in #2055
- tui: hide runtime-spawned agent hosts from the Sessions list by @wqymi in #2035
- cli: normalize command output newlines by @yanyihan-xiaomi in #2041
Refactor
- mcp: share one process-wide client layer by @wqymi in #2044
Docs
- remove free MiMo Auto API mentions from READMEs by @MiMoHardFather in #2075
New Contributors
- @YOMXXX made their first contribution in https://github.com/XiaomiMiMo/MiMo-Code/pull/2054
Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.10...v0.1.11

Summary
The biggest theme this release is orchestrator and multi-session reliability. Orchestrator now routes to an existing session instead of spawning a new one (with a roster, tool-result affordances, and guards behind it), child liveness is derived from last activity rather than last completed step — so a child blocked mid-step is no longer mistaken for a dead one — and spawn became the default in the actor tool prompt. Worktrees and bash commits now inherit the repo's real git identity instead of inventing one.
Context and checkpoint handling got a round of hard fixes: per-model early-compaction budgets, honoring the configured context limit, no more duplicate context rebuilds, manual /rebuild now does an on-the-spot rebuild with writer-freshness and a waiting UI, and the checkpoint writer's blind failure counter was replaced by proper failure classification with bounded recovery.
Providers are more resilient: never send a message with empty content (the long-standing Bedrock/Anthropic 400), replay unsigned Anthropic reasoning, retry OpenAI stream server errors and internal request timeouts, DeepSeek reasoning-effort variants, and strict: false for function tools on OpenAI Responses. Also new: Vivid and Minimal visual modes in the TUI, exec with MCP dispatch + live sub-call trace, token-aware request-scoped MCP tool discovery, client-side MCP sampling with audio, and new skill bundles (Playwright CLI automation, Grok Build CLI).

本次发布的主线是 Orchestrator 与多会话可靠性。Orchestrator 现在会路由到已有会话而不是新建(配套 roster、工具结果提示与相应守卫);子会话的存活判定改为依据"最后活动时间"而非"最后完成的步骤"——卡在某一步中间的子会话不会再被误判为已死；actor 工具提示里 spawn 成为默认。worktree 和 bash 提交现在继承仓库真实的 git 身份，不再自行编造。
上下文与 checkpoint 做了一批硬修复：按模型的提前压缩预算、尊重配置的上下文上限、不再重复重建上下文、手动 /rebuild 改为即时重建(带写入新鲜度判断和等待 UI),并把 checkpoint writer 的"盲计数失败"换成了明确的失败分类 + 有界恢复。
Provider 健壮性同步提升：不再发送空内容消息(修掉长期存在的 Bedrock/Anthropic 400)、重放未签名的 Anthropic reasoning、重试 OpenAI 流式服务端错误与内部请求超时、DeepSeek 的 reasoning effort 变体、OpenAI Responses 的 function tool 显式 strict: false。另有新特性：TUI 的 Vivid / Minimal 视觉模式、exec 支持 MCP 分发与实时子调用轨迹、按 token 感知的请求级 MCP 工具发现、支持音频的客户端侧 MCP sampling,以及新技能包(Playwright CLI 自动化、Grok Build CLI)。
Features
- tui: add Vivid and Minimal visual modes by @yanyihan-xiaomi in #2031
- exec: restore MCP dispatch with request-scoped gating, structured results, and live sub-call trace by @JinyuXiang-Mimo in #1951
- mcp: add token-aware request-scoped tool discovery by @MiMoHardFather in #1923
- mcp: support client-side sampling with audio by @wqymi in #1972
- compaction: per-model early-compaction budget, and fix the Codex context clamp by @yanyihan-xiaomi in #1930
- provider: replay unsigned Anthropic reasoning by @MiMoHardFather in #1906
- session: render Claude environment dynamically by @MiMoHardFather in #1990
- skill: add Playwright CLI automation bundle by @MiMoHardFather in #1917
- skill: add Grok Build CLI guide + localize its metadata by @MiMoHardFather in #1911, #1927
- skill: separate model reachability from authorization by @yanyihan-xiaomi in #2026
- cli: add local install script by @MiMoHardFather in #1910
- config: disable auto-dream and auto-distill by default by @JinyuXiang-Mimo in #1919
Bug Fixes
- orchestrator: route to an existing session instead of creating one — roster, tool-result affordances, and the guards that back them by @wqymi in #1741
- actor: derive liveness from last activity, not last completed step — a child blocked mid-step was indistinguishable from a dead one by @wqymi in #1965
- actor: make spawn the default and run the exception in the tool prompt by @wqymi in #1942
- worktree,bash: propagate the repo's git identity into worktrees and bash commits — never invent one by @wqymi in #1825
- provider: never send a message with empty content (the Bedrock/Anthropic 400) — producer + backstop + inbox invariant by @wqymi in #1948
- provider: enable reasoning effort variants for DeepSeek models, scoped to the official API by @anandlo, @MiMoHardFather in #1895, #1913
- provider: send function tools to Open

Summary
The headline of this release is the upcoming sunset of the MiMo-V2.5 free trial. Free access to MiMo-V2.5 will end at 03:00 (PDT) on July 26. After that, MiMo Code enforces the sunset in the TUI (#1884).
To keep using the MiMo-V2.5 series after the trial, subscribe to the MiMo Token Plan on the MiMo open platform — 12% off your first subscription: https://platform.xiaomimimo.com/token-plan
(Bringing your own API Key is unaffected.)
This release also improves plan-mode continuation, slash command discovery, GPT subagent tool resolution, provider robustness, and aligns the python-toolchain ruff rules with defaults.

本次发布的核心是 MiMo-V2.5 限免即将结束。免费使用将于北京时间 7 月 26 日 18:00（UTC+8）结束，届时 MiMo Code 会在 TUI 内执行限免下线（#1884）。
限免结束后如需继续使用 MiMo-V2.5 系列，欢迎在 MiMo 开放平台订阅 MiMo Token Plan，首次订阅享 88 折：https://platform.xiaomimimo.com/token-plan
（自行配置 API Key 的使用不受影响。）
此外本版还改进了计划模式的续接、斜杠命令发现、GPT 子智能体工具解析、provider 健壮性，并将 python-toolchain 的 ruff 规则对齐默认集。
What's Changed
fix(plan): guide model to continue planning on non-Yes plan_exit/plan_enter answers by @yanyihan-xiaomi in #1883
fix(provider): bound OpenAI response header waits by @MiMoHardFather in #1877
feat(tui): enforce MiMo free API sunset by @MiMoHardFather in #1884
feat(tui): improve slash command discovery by @MiMoHardFather in #1886
fix(actor): resolve GPT subagent tools by catalog model by @MiMoHardFather in #1887
fix(workflow): make workflow tool opt-in by @MiMoHardFather in #1893
fix(skill): align python-toolchain ruff rule set with ruff defaults + UP/I by @yanyihan-xiaomi in #1900
chore: bump version to 0.1.9 by @qiaozongming in #1901
Full Changelog: v0.1.8...v0.1.9

What's Changed
- docs: update README with new workflow, expanded skills, and custom endpoint config by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1826
- docs: update mimocode-docs skill by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1823
- chore: bump workspace versions to 0.1.7 by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/1829
- fix(agent): confine dream/distill writes to memory and .mimocode by @JinyuXiang-Mimo in https://github.com/XiaomiMiMo/MiMo-Code/pull/1827
- docs(skill): pptx visual QA model selection — prefer current model, n… by @oscar-cao7 in https://github.com/XiaomiMiMo/MiMo-Code/pull/1828
- fix(session): never PRODUCE an empty-content user message (guard at createUserMessage + schema) by @wqymi in https://github.com/XiaomiMiMo/MiMo-Code/pull/1732
- refactor(skill-search): simplify reminder logic and add enable flag by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1845
- fix(project): preserve session time_updated during global re-parent by @oscar-cao7 in https://github.com/XiaomiMiMo/MiMo-Code/pull/1830
- feat(skills): per-OS CJK font guidance in office skills (docx/pptx/xlsx) by @clyfish in https://github.com/XiaomiMiMo/MiMo-Code/pull/1824
- feat(codex): add Codex login option and remove model filtering by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1852
- docs(mimocode): harden provider configuration guidance by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1856
- fix(session): contain checkpoint writer recovery by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/1859
- feat(skills): office skills pick up bundled runtimes via MIMO_* env vars; fix Windows soffice profile URI by @clyfish in https://github.com/XiaomiMiMo/MiMo-Code/pull/1862
- feat(tool): align GPT tool guidance and skill search by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1865
- feat(compose): add builtin /compose-next alongside deprecated Compose agent by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/1861
- chore(deps): remove unused sst & @actions/artifact — drop unlicensed transitive deps by @clyfish in https://github.com/XiaomiMiMo/MiMo-Code/pull/1869
- feat(model): support gpt model by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1864
- fix(prompt): clarify model identity and question channel by @MiMoHardFather in https://github.com/XiaomiMiMo/MiMo-Code/pull/1873
- docs: recommend /compose-next and sync missed v0.1.6/v0.1.7 features by @yanyihan-xiaomi in https://github.com/XiaomiMiMo/MiMo-Code/pull/1875
- chore: bump version to 0.1.8 by @qiaozongming in https://github.com/XiaomiMiMo/MiMo-Code/pull/1879
Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.7...v0.1.8

Summary
This release adds tool_script — programmatic tool orchestration in a QuickJS sandbox, letting the model chain multiple tool calls in one script instead of many round-trips, with a hardened serialization contract, sandbox jail, MCP dispatch, and opt-in gating.
It also introduces session handoff + a try-best detector (behind MIMOCODE_ENABLE_TRY_BEST_HANDOFF, default off) and a big skills push: new bundles (data-analytics, product-design, sales, learn-everything), a skill_search tool with BM25 matching, bundle autocomplete + i18n, and the mimocode bundle renamed to mimocode-docs. Other updates: xAI/Grok bump with OAuth login, interleaved-reasoning transform sync, several provider/session robustness fixes, and switching install docs to bun ci.

本次发布新增 tool_script:在 QuickJS 沙箱里以脚本方式编排工具调用，让模型在一段脚本里串联多个工具、减少多轮往返，并配有严格的序列化契约、沙箱隔离、MCP 分发和按需开关。
同时引入会话交接(handoff)+ try-best 检测器(由 MIMOCODE_ENABLE_TRY_BEST_HANDOFF 控制，默认关闭),以及一批技能(skill)增强：新增 data-analytics、product-design、sales、learn-everything 等 bundle,带 BM25 匹配的 skill_search 工具，bundle 自动补全与多语言，并把 mimocode bundle 更名为 mimocode-docs。其他更新：xAI/Grok 升级并支持 OAuth 登录、interleaved reasoning 的 transform 同步、若干 provider/session 健壮性修复，以及安装文档改用 bun ci。
Features
- tool: add tool_script — programmatic tool orchestration in a QuickJS sandbox by @JinyuXiang-Mimo in #1749
- skill: add skill_search tool with BM25 matching by @MiMoHardFather in #1799
- skill: add data-analytics, product-design, and sales skill bundles by @MiMoHardFather in #1792
- skill: add learn-everything bundle by @MiMoHardFather in #1805
- skill: skill bundle autocomplete + i18n; rename mimocode bundle to mimocode-docs by @MiMoHardFather in #1796, #1797
- session: session handoff and try-best detector (opt-in via MIMOCODE_ENABLE_TRY_BEST_HANDOFF, default off) by @MiMoHardFather in #1751, #1767
- provider(xai): bump @ai-sdk/xai to 3.0.102, sync Grok transform, add OAuth login by @wqymi in #1818
Bug Fixes
- tool_script: serialization contract, sandbox jail, MCP dispatch, and opt-in gating; discourage overuse and stop escaping return values by @JinyuXiang-Mimo, @yanyihan-xiaomi in #1784, #1777
- provider: guard non-array message content in part mapping (j.map is not a function) by @wqymi in #1728
- transform: sync interleaved handling — exclude openrouter + always echo reasoning field by @wqymi in #1819
- session: treat GPT reasoning-only steps as terminal by @MiMoHardFather in #1820
- session: narrow empty-step guard to only empty-args tool calls by @yanyihan-xiaomi in #1776
- session: session LLM and plugin hook fixes by @MiMoHardFather in #1815
- task: remove main-session task gate, keep subagent gate only by @yanyihan-xiaomi in #1759
- tui: clean up unsupported tips by @yanyihan-xiaomi in #1768
- skill: only expose top-level builtin entries; localize new bundles and hide internal workflows by @MiMoHardFather in #1795
- build: fresh-clone engine node build + pin ghostty-web to a commit by @clyfish in #1771
Docs
- document CJK/East-Asian font handling for office skills by @clyfish in #1758
- clarify pptx preview script paths and CLI-only scope by @yanyihan-xiaomi in #1789
- use bun ci for install instructions by @clyfish in #1775, #1778
Internal / CI
- default MIMOCODE_ENABLE_TRY_BEST_HANDOFF to off + align tests by @MiMoHardFather in #1767, #1770
- rename build:dev to build:local + add root shortcut by @yanyihan-xiaomi in #1821
- skip two cold-warmup-flaky workflow/tool tests by @yanyihan-xiaomi in #1779
New Contributors
- @clyfish made their first contribution in https://github.com/XiaomiMiMo/MiMo-Code/pull/1771
Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.6...v0.1.7

Summary
Slash skills are now enabled by default and support multi-skill orchestration — referencing 2+ /skill-name in a single message auto-loads their contents and injects an orchestration plan. /skip-permissions adds a runtime toggle for unattended runs — forced-ask operations (destructive bash, etc.) auto-reject after 60s with actionable feedback, while other non-deny permissions auto-allow.
Agent mode is now locked after first message: Build and Plan can still switch between each other, but entering Compose isolates the session to that mode. Many models tend to ignore tools that appear mid-conversation, so keeping the skill/tool set fixed from session start significantly improves tool-call reliability.
Custom provider compatibility: /modalities command lets users configure multimodal support (image, audio, video, pdf) for custom models directly in the TUI without manual config editing; system prompt is now sent as a single {role:'system'} message instead of two, fixing vLLM and other backends that expect exactly one system message; Moonshot/Kimi tool calling no longer 400s on discriminated-union schemas; oversized images are now compressed and area-averaged downscaled before sending, preventing permanent 400 failures.
Other highlights: orchestrator mode improvements (liveness tracking, fan-in aggregation, stall watchdog, structured notification cards); workflow meta.permissions manifest for up-front permission declaration; /rebuild command to manually rebuild context from checkpoint; file hooks hot-reload from any external editor without restart; builtin skills gated by CLI availability; broken custom tools no longer crash the session.

Slash skills 默认开启，支持多技能编排——一条消息里引用多个 /skill-name 会自动加载对应技能内容并生成编排计划。新增 /skip-permissions 命令，适合无人值守跑任务——危险操作 60s 无响应自动拒绝，其余非 deny 权限自动放行。
模式锁定：发送第一条消息后 agent 模式不可切换（Build/Plan 之间仍可互切，但 Compose 进入后隔离）。很多模型不喜欢中途出现的新工具、倾向于不调用动态插入的 skill，锁定模式让 skill/tool 集合从会话开始就固定，显著提升工具调用成功率。
Custom Provider 兼容性改进：新增 /modalities 命令，可以直接在 TUI 里为自定义模型设置多模态支持列表（图片、音频、视频、PDF），不用手动改配置文件；system prompt 合并为单条消息发送，修复 vLLM 等后端因双 system message 导致的指令重复问题；Moonshot/Kimi tool calling 不再因 discriminated-union schema 报 400；超大图片发送前自动压缩和面积均值缩放，不再因尺寸超限导致 400 永久卡死。
其他：orchestrator 模式改进（存活状态跟踪、fan-in 聚合、卡顿监控、结构化通知卡片）；workflow 支持 meta.permissions 清单用于前置权限声明；新增 /rebuild 命令手动从 checkpoint 重建上下文；文件 hooks 支持外部编辑器热重载；内置 skill 按 CLI 可用性自动门控；自定义 tool 文件有语法错误不再导致整个 session 崩溃。
Features
- tui: slash skills enabled by default with multi-skill orchestration — /skill-name triggers autocomplete anywhere in input, 2+ skills auto-load with orchestration prompt; compose: prefixed skills correctly filtered to compose agent only by @MiMoHardFather, @yanyihan-xiaomi in #1654, #1656
- tui: add /modalities command to configure per-model input capabilities (image/audio/video/pdf) with multi-select dialog, persistent config and live-reload by @JinyuXiang-Mimo in #1650, #1651
- permission: add /skip-permissions runtime toggle for unattended runs — auto-allows non-deny permissions; forced-ask permissions auto-reject after 60s with actionable CorrectedError feedback instead of hanging forever by @JinyuXiang-Mimo in #1645
- session: agent mode locks after first message — Build/Plan freely switch between each other, Compose isolated once entered; fixes models ignoring dynamically injected skills mid-session by @yanyihan-xiaomi in #1725
- orchestrator: liveness tracking (progressing/stalled), fan-in aggregation, stall watchdog notifications, reliable relay into idle peers, grant persistence, and structured notification cards in TUI by @wqymi in #1681, #1699, #1691, #1711
- workflow: meta.permissions manifest for up-front permission declaration + engine-level agent retry by @wqymi in #1686
- session: /rebuild command to manually rebuild context from the latest checkpoint; fixes large-session wedge when a checkpoint writer is in-flight by @wqymi in #1647
- provider: area-averaging image downscale (better than nearest-neighbor) + automatic compression of oversized 

Summary
This release ships a builtin skills system — 14 domain-specific skill bundles (document handling, academic writing, arxiv search, deep research, auto-research, html-to-video, skill authoring, and self-documentation) that auto-load when relevant files are attached or task context matches. Skills show localized descriptions in the TUI dialog and support auto-detection of office file attachments. New built-in research workflows: deep-research and fact-check. The self-extend skill is renamed to evolve with a rewritten SKILL.md framing total self-modification — tools, hooks, knowledge, workflows, and UI are all rewritable.
A new Orchestrator mode (opt-in, MIMOCODE_EXPERIMENTAL_ORCHESTRATOR) introduces a delegating coordinator agent that accomplishes goals by spawning and managing child sessions with git worktree isolation and cross-session permission routing.
The cron + loop scheduling system (opt-in, MIMOCODE_EXPERIMENTAL_CRON) enables recurring or self-paced prompts with 5-field cron expressions, durable persistence, and a TUI fire-time clock row.
Vision-aware image paste now spills images to temp files as @file references for non-vision models (instead of base64 that causes hallucination), with smart cost-aware vision_model auto-selection and an actor models [--vision] discovery command.
The Edit tool defaults to exact string matching — fuzzy matching is now opt-in via MIMOCODE_ENABLE_FUZZY_EDIT. Match failures include closest-match hints to help callers correct their old_string.
Other highlights: first-class Windows native installer (PowerShell one-liner alongside macOS/Linux with working upgrade/uninstall), automatic $schema URL migration from opencode.ai to mimo.xiaomi.com/mimocode/, gpt-5.5 support (encrypted reasoning round-trip fix), and a rewritten repetition detector that eliminates false-positive loop-kills on structured content.

本次发布引入 内置 Skills 体系 —— 14 个领域专用 skill bundle（文档处理、学术写作、arXiv 检索、深度调研、自动科研、HTML 转视频、skill 创作、自文档化），在附加相关文件或任务上下文匹配时自动加载，TUI 对话框展示本地化描述。新增研究型内置 workflow：deep-research 和 fact-check。self-extend 重命名为 evolve，SKILL.md 重写为全面自我修改接口——工具、钩子、知识、工作流、界面均可改写。
新增 Orchestrator 模式（opt-in, MIMOCODE_EXPERIMENTAL_ORCHESTRATOR）—— 委派式协调 agent，通过创建/管理子 session 完成目标，支持 git worktree 隔离和跨 session 权限路由。
Cron + Loop 调度系统（opt-in, MIMOCODE_EXPERIMENTAL_CRON）支持 5 字段 cron 表达式、持久化/会话级调度和自适应循环，TUI 新增触发时间行。
视觉感知图片粘贴 对非视觉模型将图片溢写为临时文件 @file 引用（避免 base64 导致幻觉），自动选择性价比最优的 vision_model，新增 actor models [--vision] 发现命令。
Edit 工具默认切换为精确字符串匹配 —— 模糊匹配改为 opt-in（MIMOCODE_ENABLE_FUZZY_EDIT），匹配失败时提供最近似候选提示。
其他重点：Windows 原生安装器转正（PowerShell 一行命令，upgrade/uninstall 全流程可用）、$schema URL 自动迁移（从 opencode.ai 到 mimo.xiaomi.com/mimocode/）、gpt-5.5 支持（加密 reasoning 往返修复）、重写重复检测器（消除对结构化内容的误杀）。
Features
- skill/builtin: ship 14 builtin skills — docx, pdf, pptx, xlsx document handling; arxiv paper search; research-paper-writing; html-to-video pipeline; skill-creator (replaces compose:new-skill); mimocode self-documentation; frontend-design; modern-python-toolchain; deep-research (parallel sub-agents → reflect → cited report, checkpoint-resumable); auto-research (literature surveys, experiment loops, paper reproduction, citation auditing, zero external deps). Add deep-research and fact-check (adversarial 3-juror verification) as built-in workflows. Office file attachments auto-inject relevant skill context; TUI skill dialog shows localized descriptions across 7 languages; slash visibility gated by MIMOCODE_ENABLE_SLASH_SKILLS (#1557, #1564, #1565, #1555, #1569, #1570, #1572, #1567, #1531, #1583, #1579, #1585, #1587)
- skill/builtin: rename self-extend to evolve — rewrite SKILL.md as proactive total self-modification interface (tools, hooks, knowledge, workflows, UI all rewritable); sync hook-api.md with 7 missing events (session/actor lifecycle); add ctx.ask() to tool-api.md (#1590)
- orchestrator: Orchestrator mode (opt-in, MIMOCODE_EXPERIMENTAL_ORCHESTRATOR) — a primary agent that delegates to child sessions via a session tool with create/switch/list/c

MiMoCode v0.1.4

MiMoCode v0.1.3

MiMoCode v0.1.2

MiMoCode v0.1.1

Initial open-source release of MiMo Code.