Skip to content

Version Packages - #2808

Merged
ben-reitz merged 3 commits into
mainfrom
changeset-release/main
Nov 20, 2026
Merged

ben-reitz merged 3 commits into
mainfrom
changeset-release/main

Conversation

@github-actions

@github-actions github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

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

agents@0.34.0

Minor Changes

  • #2832 c31b57e Thanks @ben-reitz! - Add onFetchError(request, error) to Lifecycle hosts and Agent, which chooses the response when a request or WebSocket upgrade fails. See Failed requests.

  • #2782 d6b4441 Thanks @mattzcarey! - MCP client: requests that need the user to authorize again throw MCPAuthorizationRequiredError with a link, and failed servers stay failed until the new reconnectMcpServer(id). See MCP client.

  • #2925 cf2ad50 Thanks @ben-reitz! - Support Pi 1.6 (~1.6.0; Pi 1.5 is no longer supported). harness.signal() delivers a signal to a parked Pi tool. PiSession.submit() and continue() take notify, so a parked tool can wake when another session's input settles, and submit() takes onAbort: "keep", so a queued input stays queued when the session is aborted. agents/models/pi-ai counts reasoning and cached tokens that a chat-completions vendor reports at the top level of usage, as pi-ai does, and resends a conversation in full when Anthropic can no longer continue its message thread, instead of failing the turn. See pi-durable 1.6.0's changelog and Pi harness.

  • #2926 6fef21d Thanks @ben-reitz! - PiHarness creates the root session on its first use instead of when Pi opens, and harness.root({ agent, init }), sessions.create({ agent, init }) and sessions.fork(from, { agent, init }) take Pi's init, so a session can start from an existing history. See Seed a session with an existing history.

Patch Changes

  • #2832 c31b57e Thanks @ben-reitz! - When onError returns instead of throwing, Agent and AIChatAgent now pass the original error on, instead of undefined (or an unawaited promise when onError is async).

  • #2861 f0f4ed9 Thanks @mattzcarey! - Fix chat recovery and resume races: stop now ends a running recovery, recovery yields to a newer question, and a stale resume ACK keeps the tab waiting for the active stream's replay.

  • #2860 7e2e11e Thanks @mattzcarey! - Fix chat turn races across tabs: clear stops pending replies, regenerate replaces a reply still streaming, a throwing onChatMessage sends an error frame, and repeated tool results no longer re-run the model.

  • #2835 5c33d89 Thanks @ben-reitz! - Send a burst of messages on a sub-agent connection in one call to the parent, so resuming a sub-agent chat stream shows the partial reply at once instead of replaying it frame by frame (#2820).

  • #2914 06a2715 Thanks @ben-reitz! - claudeCode() now defaults to claude-sonnet-5-5, since claude-sonnet-4-5 retires on November 30, 2026.

  • #2895 47ccc29 Thanks @ben-reitz! - DurableObjectOAuthClientProvider is now an OAuthClientProvider, and a Browser is now a BrowserSource, in apps that compile with exactOptionalPropertyTypes.

  • #2913 3d33563 Thanks @ben-reitz! - Export normalizeCdpSpec and RawCdpProtocol from agents/browser, so a custom BrowserSource can build its spec() from protocol JSON it already has. See Lower-level helpers.

  • #2899 1a450e7 Thanks @ben-reitz! - ai.image() and pi-ai's ai.images() now run the FLUX.2 image models (@cf/black-forest-labs/flux-2-klein-9b, flux-2-klein-4b, flux-2-dev), sending their multipart input and up to four reference images.

  • #2858 5071d3c Thanks @mattzcarey! - A schedule or queue item that is cancelled or replaced while its callback waits to retry no longer runs again with the old job. This also holds for sub-agent schedules and queue items, and when a second alarm invocation picks up the replacement.

  • #2829 88eb62c Thanks @github-actions! - Lifecycle.fetch no longer sends an error's stack to the client when startup or a request handler throws. HTTP requests get a generic Internal Server Error body and WebSocket upgrades close with 1011 without an error frame. The full error is still logged. See #2828.

  • #2827 e0f95e5 Thanks @mattzcarey! - Type LifecycleCapability.onStart() as returning MaybePromise<void> to match the runner, which awaits it. See Agents.

  • #2863 46daba4 Thanks @mattzcarey! - Fix MCP client races found by TLA+ model checking: a pending step-up link survives a token refresh, a refresh no longer overwrites a step-up grant, a removed server's tokens stay deleted, a sign-in survives closeConnection(), an old link no longer fails a server added again, overlapping closes and removals leave a server added again alone, addMcpServer(..., { id }) returns a link that works and renames the server in one transaction, configuring elicitation handlers while a server connects no longer fails it and is advertised, step-up rounds count for the scope set of the link that was completed, and a declined sign-in after a refused refresh is recorded as failed (see re-authorization during a session).

  • #2856 cff5dd1 Thanks @mattzcarey! - An MCP server removed and added again under the same id while it was still connecting now starts clean. The removed server's connect no longer marks the new one failed ("was removed while connecting"), and its retries no longer connect the new one a second time. With a custom createAuthProvider, removeMcpServer() no longer deletes the stored row of a server added again while it cleared credentials.

  • #2899 1a450e7 Thanks @ben-reitz! - Suggest @cf/deepseek-ai/deepseek-v4.1-flash-1017 instead of @cf/zai-org/glm-4.7-flash in the unknown-model error and doc comments.

  • #2781 366d5b5 Thanks @github-actions! - Fix Think recovering an interrupted continuation as a duplicate assistant message; recovery now rebuilds the message it continues. Think now requires agents >=0.34.0. See Chat Recovery.

  • #2807 1f3bd92 Thanks @ben-reitz! - The pi web_search and web_fetch tools give code mode scripts their results as objects.

  • #2899 1a450e7 Thanks @ben-reitz! - Move off the Workers AI models retired on November 20, 2026: drop MeloTTS and whisper-tiny-en handling, send @cf/google/embeddinggemma-2 its { input } shape, and suggest @cf/deepseek-ai/deepseek-v4.1-flash-1017 in examples and messages. See Planned model deprecations.

  • #2859 f41f395 Thanks @mattzcarey! - A Task run cancelled while its attempt is still being claimed no longer starts its handler.

  • #2867 0f2727a Thanks @mattzcarey! - Stop running client messages on a WebSocket the host has closed. A connection rejected from onConnect with connection.close() could still run one @callable method, state update, or onMessage call. The close happens before the client has the upgrade response, so anything the client sent as soon as it connected arrived after the close, and the platform still delivered it. Connection.close() now records the close in the socket attachment, so the record survives hibernation, and the WebSockets capability drops every later message on both the hibernating and Cap'n Web wires. Messages the client sent before closing the connection itself still run.

@cloudflare/think@0.24.0

Minor Changes

  • #2819 9b54e51 Thanks @ben-reitz! - Think's @cloudflare/computer peer is now ^0.10.0, so upgrade Computer to 0.10 alongside Think. Its JavaScript backend is now CodemodeBackend from @cloudflare/computer/backends/codemode, and host modules such as ws:tools are now codemode:tools. See Running code with exec.

Patch Changes

  • #2781 366d5b5 Thanks @github-actions! - Fix Think recovering an interrupted continuation as a duplicate assistant message; recovery now rebuilds the message it continues. Think now requires agents >=0.34.0. See Chat Recovery.

  • #2899 1a450e7 Thanks @ben-reitz! - Move off the Workers AI models retired on November 20, 2026: drop MeloTTS and whisper-tiny-en handling, send @cf/google/embeddinggemma-2 its { input } shape, and suggest @cf/deepseek-ai/deepseek-v4.1-flash-1017 in examples and messages. See Planned model deprecations.

  • #2912 81726f7 Thanks @github-actions! - ExtensionManager now derives the Dynamic Worker id from a hash of the extension's code and configuration (plus the agent and context labels for extensions with a host binding, via the new workerIdScope option) instead of Date.now(), so restarting a Think Durable Object no longer creates (and bills) a new Dynamic Worker per extension. See #2911.

@cloudflare/ai-chat@0.12.3

Patch Changes

  • #2832 c31b57e Thanks @ben-reitz! - When onError returns instead of throwing, Agent and AIChatAgent now pass the original error on, instead of undefined (or an unawaited promise when onError is async).

  • #2818 a2f1216 Thanks @ben-reitz! - Cancelling a chat request while it waits in the queue now stops it from running, instead of the cancel being dropped. See #2701.

  • #2861 f0f4ed9 Thanks @mattzcarey! - Fix chat recovery and resume races: stop now ends a running recovery, recovery yields to a newer question, and a stale resume ACK keeps the tab waiting for the active stream's replay.

  • #2860 7e2e11e Thanks @mattzcarey! - Fix chat turn races across tabs: clear stops pending replies, regenerate replaces a reply still streaming, a throwing onChatMessage sends an error frame, and repeated tool results no longer re-run the model.

@cloudflare/agent-think@0.0.21

Patch Changes

@devin-ai-integration devin-ai-integration Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note

Newer findings are available below. Devin Review posted a newer report on this PR, in addition to the findings presented here.

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

Devin Review

@devin-ai-integration devin-ai-integration Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

Devin Review

@github-actions
github-actions Bot force-pushed the changeset-release/main branch 4 times, most recently from 6685c2d to cd137b2 Compare November 17, 2026 12:26
@github-actions
github-actions Bot force-pushed the changeset-release/main branch 4 times, most recently from 0cfd728 to ff4e52f Compare November 19, 2026 09:53
@github-actions
github-actions Bot force-pushed the changeset-release/main branch from ff4e52f to b50a425 Compare November 20, 2026 15:31
Custom notes for agents 0.34.0, Think 0.24.0 and ai-chat 0.12.3, covering Pi 1.6 (#2925) and seeding a Pi session with an existing history (#2926).
@pkg-pr-new

pkg-pr-new Bot commented Nov 20, 2026 •

Copy link
Copy Markdown

Open in StackBlitz

agents

npm i https://pkg.pr.new/agents@2808

@cloudflare/ai-chat

npm i https://pkg.pr.new/@cloudflare/ai-chat@2808

@cloudflare/codemode

npm i https://pkg.pr.new/@cloudflare/codemode@2808

hono-agents

npm i https://pkg.pr.new/hono-agents@2808

@cloudflare/shell

npm i https://pkg.pr.new/@cloudflare/shell@2808

@cloudflare/think

npm i https://pkg.pr.new/@cloudflare/think@2808

@cloudflare/voice

npm i https://pkg.pr.new/@cloudflare/voice@2808

@cloudflare/worker-bundler

npm i https://pkg.pr.new/@cloudflare/worker-bundler@2808

commit: 383f822

@agent-think

agent-think Bot commented Nov 20, 2026 •

Copy link
Copy Markdown
Contributor

✅ agents import sizes: no significant changes (6fef21d8 → 383f8228, workflow run)

@ben-reitz
ben-reitz merged commit 8d8acd0 into main Nov 20, 2026
25 checks passed
@ben-reitz
ben-reitz deleted the changeset-release/main branch November 20, 2026 15:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant