Aquesta és una versió anterior. La versió actual és la 0.113.1. Lody 0.113.1 · Data de publicació:
Fitxers d’instal·lació 0.110.0
| Arquitectura | Tipus | Àmbit | Mida del fitxer | Baixa | SHA256 |
|---|---|---|---|---|---|
| 64 bits | EXE (NSIS) | Usuari actual | 171,8 MB | updates.lody.ai | VirusTotal |
L’enllaç i l’SHA256 provenen del manifest comunitari del Gestor de paquets del Windows. Convé comparar el hash del fitxer baixat abans d’executar-lo.
Notes de la versió
Version 0.110.0
✨ New Features
- Image annotations: draw, add text, and leave notes in the image preview, then send the annotated image to an Agent as an attachment.
- Automatic conversation continuation: resume tasks when specified conditions are met, with per-conversation options to resume after quota recovery, archive after a PR is merged, and compact context when idle.
- Android emulator preview: a new preview and control panel with touch input, text pasting, rotation, screenshots, light and dark mode switching, and image quality options.
- Agent sign-in methods: see which sign-in methods an Agent supports before signing in.
- Pi update: upgraded built-in Pi to 1.4.0, adding support for Streamable HTTP MCP.
- Side-by-side iOS and Android previews: drag one device preview tab onto another to view both screens at once. Drag the divider to adjust their widths, or double-click it to restore the default split.
- Quote conversation text: add a selected message excerpt to the composer to ask a follow-up question with the original text as context.
- Stacked PRs: the PR panel shows the stack order recorded on GitHub, each PR’s status, and the current PR’s position in the stack. Linked PRs are also shown in the sidebar.
- More flexible scheduled tasks: prompts now support mentions, and repeat intervals can be set to a custom number of minutes or hours.
- Restore archived conversations directly: a new restore button in the composer brings back the conversation while preserving your draft.
- Cross-machine Agent steering: send follow-up instructions to a conversation running on another machine. If live steering is unavailable, the instructions are queued.
- Claude can now discover models offered by Anthropic-compatible gateways on a per-conversation basis.
- Agent queue management: Agents can inspect pending messages in a target conversation and withdraw or replace queued instructions they sent.
- More complete collaboration status: Agents can query a conversation’s active tools, last output time, and queue length, as well as machine disk information and remaining quota.
- Message references with attribution: Agents can reference the original message when forwarding instructions, including the author information recorded with it.
- Device pairing available by default: account-level device pairing no longer requires developer mode or a testing flag.
- Bulk conversation archiving: Agents can archive multiple conversations in a single MCP call.
- Two quota windows on the home page: view usage percentages and reset times for both short-term and weekly quotas.
- Independent conversation group names: rename sidebar conversation groups and individual tabs separately, without affecting each other.
- Clearer collaboration failures: MCP returns specific failure reasons, quota reset times, and other details. A successful retry now clears the previous failure state.
🐛 Fixes & Improvements
- The Claude quota-exhausted prompt now offers to enable automatic conversation resumption. Lody checks again and continues once quota recovers or the reset time is reached.
- Scheduled messages now show their full text, can be edited within the current conversation, and display send times in the viewer’s local time zone.
- Fixed scheduled message delivery getting stuck while a conversation waits for delegated tasks or when the message inbox fails to start.
- Improved recovery of background services, message queues, and MCP lists after brief connection failures, reducing stalled conversations.
- Fixed conversation settings, such as model and reasoning effort, being overwritten by earlier input or drafts.
- Improved Pi conversation forking and slash command loading, preventing command lists from leaking between conversations.
- Fixed image imports, list details, and startup error display for DeepSeek history.
- Added compatibility with Magpie models that do not declare their input capabilities.
- Improved scroll momentum in iOS conversations, button taps on the history page, sidebar resizing, and desktop page alignment.
- Improved the default font selection, preferring SF Pro Display when available.
- Fixed subscription information failing to load causing the entire workspace to fail, and improved home screen widget cache recovery.
- Upgraded built-in Claude Code to 2.1.316.
- Restored the previous conversation configuration and message-sending behavior, fixing message delivery failures with older background services. Remove the earlier entry about conversation settings no longer being overwritten by earlier input or drafts.
- Automatic continuation controls now reflect the target machine’s actual capabilities, avoiding unavailable actions with older background services.
- Moved Android preview quality options into the “More” menu to free up toolbar space.
- Fixed errors caused by loading organization information before authentication was ready.
- Improved multi-Agent collaboration: responses now include message delivery status and specific failure reasons, identify the source of Agent messages, and wait for the final reply after steering instead of returning only an intermediate result.
- MCP now accepts unambiguous short conversation IDs and supports reading history by role, time, and direction, reducing unnecessary context.
- Reduced memory usage when reading long conversation histories and running internal MCP processes, and improved history access during streaming output.
- Fixed the automatic continuation queue not waking promptly, and added clear connecting, offline, and unavailable states.
- Fixed history loading overwriting the state after exiting plan mode, and PR status getting out of sync between the information bar and sidebar.
- Multiple tabs for the same conversation now share the same simulator binding, reducing repeated device selection.
- Improved question card titles and Markdown rendering, and added compatibility with Droid question forms.
- Fixed annotated images exceeding attachment limits, PDF panel zoom fitting, and share card layouts in WebKit.
- PR images now load on demand. Fixed the sidebar tab scrollbar not being draggable, and text alignment in the message editor and queue.
- Scheduled tasks now defer execution when quota is exhausted and a reset time is known, avoiding repeated failures or a burst of runs when quota recovers.
- When a conversation finishes, only reasoning and tool activity are collapsed; all reply text remains visible. Expanded activity is shown in its original order.
- Used prompt suggestions are hidden after sending a message, preventing old suggestions from reappearing.
- Fixed the keyboard covering sign-in fields on phones, and the status bar covering back buttons on some pages.
- Fixed the folder browser failing to open when a project record has no path.
- Fixed CLI snapshot uploads potentially waiting indefinitely, and improved retries after initialization failures.
- Standardized font weights across the composer, send button, and attachment labels.
- Sidebar PR display: restored the original single-PR status icon. Remove the earlier entry about linked PRs in the sidebar; stacked PRs remain available in the PR panel.
- Home page cards now reorder based on task status. The new conversation composer remains usable if cards fail to load.
- Fixed pinned conversations losing their linked hierarchy, and sidebar icons shifting when a PR is shown as mergeable.
- Improved compatibility with Agent sign-in methods and authentication progress feedback, and added missing usage tracking for custom ACP Agents.
- Fixed healthy machines being incorrectly marked offline during brief sync failures.
- Optimized file indexing for large projects by skipping ignored directories and caches, reducing CLI stalls and files being incorrectly marked as deleted.
- Fixed Git diff calculation when no comparison base is recorded, using the remote’s default branch instead.
- Browser previews can now be opened directly on mobile without manually entering an address. Fixed a theme mismatch along the inside edge of the iOS simulator frame.
- Updated reply collapsing: progress updates explicitly marked by the Agent are now collapsed with other activity, while final replies remain visible. This replaces the earlier statement that all reply text remains visible.
Thanks to external contributors Innei, Astro-Han, delete-cloud, liyown, dreaifekks, ox01024, Dante-dan, and boris-w for their contributions to this update!