[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@intentface/latch-core@0.24.0

Minor Changes

• 97ad289: An input queue: what arrives while an answer is running is held, and the answer that ends starts the next turn with all of it.

  • handleChat({ ifBusy: "queue" }), or queue: true in the body of POST agents/:agent/:chatId: a message sent to a busy chat is held and answered 202 { queued: { id } }, where it was refused with chat_busy. The default is unchanged.
  • sendSignal on a busy chat, or one waiting on an approval or tool result, now holds the signal and returns { id, queued: true }; its turn starts when the chat is free. Before, it waited for the user's next message. wake: false is unchanged.
  • The run that answers held inputs opens first, then takes everything held in one step. An input is answered once however many workers go for it; while it is being answered it is not listed, not counted against the cap, and can't be taken back; one taken back in time is never answered; one replaced by a retry is answered as replaced; and if the turn can't start, what it took waits again.
  • One turn takes everything held, signals read with the messages. A chat holds at most inputQueue.max inputs (20 by default); past it a message is refused with QueueFullError (queue_full, 429).
  • runtime.listQueued, runtime.cancelQueued, and DELETE chats/:chatId/queue/:messageId.
  • runtime.pollChat and GET chats/:chatId/poll?cursor=…: whether an answer is running, what is held, and the history when it changed. For a page with no stream open.
  • waitUntil on the runtime: with it, the turn for held inputs runs after the response has ended, so the answer's stream closes on time and a held message's 202 returns at once. Without it the stream stays open until that turn is done.
  • runtime.cron() starts the turn for inputs held more than 15 seconds with nobody starting one (a process that died first), and reports them as held. Needs the optional storage method listChatsWithQueuedInputs.
  • Storage: six optional methods (enqueueInput, listQueuedInputs, removeQueuedInputs, takeQueuedInputs, returnQueuedInputs, finishQueuedInputs), covered by the conformance suite. The Drizzle adapters implement them on a new latch_queued_inputs table; run migrate. An adapter without them behaves as before. listRuns takes chatId.

• 1ef8051: runtime.sendSignal: tell a chat something from server code. The agent answers it at once, or reads it with the user's next message (wake: false); nothing already stored changes, so the prompt cache still hits.

  • A notice (the default) is the application speaking. It is sent as a system message after the user's next message on OpenAI models (direct or Azure) and on the Anthropic models that take one mid-conversation (direct, Vertex, or Bedrock through @ai-sdk/amazon-bedrock/anthropic), and as a <notice> block inside that message everywhere else. resolveNoticePlacement on the runtime overrides the choice per model.
  • A message is a named person other than the chat's user, always read as a person speaking.
  • Stored as an AppMessage with role: "system" and metadata.signal. Hidden from the user unless visibility: "user" is passed. A client can't create one.
  • By default it starts a turn with the signal as what the agent answers, when the chat is free, and resolves with the answer. With wake: false, or when the chat is busy, waiting on a person or over a limit, the signal is stored and waits for the next message.
  • toModelMessages takes notices and keepTrailingSignals; placeSignals, defaultNoticePlacement and isSignal are exported.

Patch Changes

• @intentface/latch-net@0.24.0

@intentface/latch-drizzle@0.24.0

Minor Changes

• 97ad289: An input queue: what arrives while an answer is running is held, and the answer that ends starts the next turn with all of it.

  • handleChat({ ifBusy: "queue" }), or queue: true in the body of POST agents/:agent/:chatId: a message sent to a busy chat is held and answered 202 { queued: { id } }, where it was refused with chat_busy. The default is unchanged.
  • sendSignal on a busy chat, or one waiting on an approval or tool result, now holds the signal and returns { id, queued: true }; its turn starts when the chat is free. Before, it waited for the user's next message. wake: false is unchanged.
  • The run that answers held inputs opens first, then takes everything held in one step. An input is answered once however many workers go for it; while it is being answered it is not listed, not counted against the cap, and can't be taken back; one taken back in time is never answered; one replaced by a retry is answered as replaced; and if the turn can't start, what it took waits again.
  • One turn takes everything held, signals read with the messages. A chat holds at most inputQueue.max inputs (20 by default); past it a message is refused with QueueFullError (queue_full, 429).
  • runtime.listQueued, runtime.cancelQueued, and DELETE chats/:chatId/queue/:messageId.
  • runtime.pollChat and GET chats/:chatId/poll?cursor=…: whether an answer is running, what is held, and the history when it changed. For a page with no stream open.
  • waitUntil on the runtime: with it, the turn for held inputs runs after the response has ended, so the answer's stream closes on time and a held message's 202 returns at once. Without it the stream stays open until that turn is done.
  • runtime.cron() starts the turn for inputs held more than 15 seconds with nobody starting one (a process that died first), and reports them as held. Needs the optional storage method listChatsWithQueuedInputs.
  • Storage: six optional methods (enqueueInput, listQueuedInputs, removeQueuedInputs, takeQueuedInputs, returnQueuedInputs, finishQueuedInputs), covered by the conformance suite. The Drizzle adapters implement them on a new latch_queued_inputs table; run migrate. An adapter without them behaves as before. listRuns takes chatId.

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0
  • @intentface/latch-agent-builder@0.24.0
  • @intentface/latch-memory@0.24.0

@intentface/latch-server@0.24.0

Minor Changes

• 97ad289: An input queue: what arrives while an answer is running is held, and the answer that ends starts the next turn with all of it.

  • handleChat({ ifBusy: "queue" }), or queue: true in the body of POST agents/:agent/:chatId: a message sent to a busy chat is held and answered 202 { queued: { id } }, where it was refused with chat_busy. The default is unchanged.
  • sendSignal on a busy chat, or one waiting on an approval or tool result, now holds the signal and returns { id, queued: true }; its turn starts when the chat is free. Before, it waited for the user's next message. wake: false is unchanged.
  • The run that answers held inputs opens first, then takes everything held in one step. An input is answered once however many workers go for it; while it is being answered it is not listed, not counted against the cap, and can't be taken back; one taken back in time is never answered; one replaced by a retry is answered as replaced; and if the turn can't start, what it took waits again.
  • One turn takes everything held, signals read with the messages. A chat holds at most inputQueue.max inputs (20 by default); past it a message is refused with QueueFullError (queue_full, 429).
  • runtime.listQueued, runtime.cancelQueued, and DELETE chats/:chatId/queue/:messageId.
  • runtime.pollChat and GET chats/:chatId/poll?cursor=…: whether an answer is running, what is held, and the history when it changed. For a page with no stream open.
  • waitUntil on the runtime: with it, the turn for held inputs runs after the response has ended, so the answer's stream closes on time and a held message's 202 returns at once. Without it the stream stays open until that turn is done.
  • runtime.cron() starts the turn for inputs held more than 15 seconds with nobody starting one (a process that died first), and reports them as held. Needs the optional storage method listChatsWithQueuedInputs.
  • Storage: six optional methods (enqueueInput, listQueuedInputs, removeQueuedInputs, takeQueuedInputs, returnQueuedInputs, finishQueuedInputs), covered by the conformance suite. The Drizzle adapters implement them on a new latch_queued_inputs table; run migrate. An adapter without them behaves as before. listRuns takes chatId.

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0
  • @intentface/latch-mcp@0.24.0

@intentface/latch-agent-builder@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0

@intentface/latch-channels@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0

@intentface/latch-mcp@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0
  • @intentface/latch-net@0.24.0

@intentface/latch-memory@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0

@intentface/latch-openapi@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0
  • @intentface/latch-net@0.24.0

@intentface/latch-otel@0.24.0

Patch Changes

• Updated dependencies [97ad289]
• Updated dependencies [1ef8051]
  • @intentface/latch-core@0.24.0

@intentface/latch-net@0.24.0

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
intentface-latch-docs	Ready	Preview	Nov 13, 2026 2:34pm UTC

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: b9e3038b-5fc4-4c7c-ada3-6e507f5b9fd4

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts