Conversation
Introduces a dedicated `chat/ai` subpath as the home for every Vercel AI SDK helper that ships with Chat SDK. Importing from this subpath keeps the optional `ai` and `zod` peer dependencies out of bundles that don't use them. What's new: - `createChatTools` exposes Chat SDK operations as ready-to-use AI SDK tools so an agent can read, post, react, edit, delete, and manage thread subscriptions across every adapter the supplied `Chat` instance has registered. Write operations require user approval by default and can be toggled globally or per-tool via `requireApproval`. Three presets (`reader`, `messenger`, `moderator`) scope the toolset, and individual tools can also be cherry-picked. - `toAiMessages` (and the supporting `Ai*` / `ToAiMessagesOptions` types) now ship from `chat/ai` alongside the tools. The previous `chat` re-exports continue to work, but are flagged `@deprecated` with an editor hint pointing to the new home — migrating is a one-line import change. Coverage on the new module is comprehensive: the `createChatTools` orchestrator hits 100% statements / 94.7% branches, every tool factory's `execute()` is exercised end-to-end, and the existing `toAiMessages` test suite continues to cover role mapping, attachment handling, links, transforms, and unsupported-attachment fallbacks.
Restructures the AI-related documentation into a dedicated section slotted between the Usage and Adapters groups in the sidebar: - /docs/ai — Overview of the chat/ai subpath - /docs/ai/ai-sdk-tools — createChatTools guide (was /docs/ai-tools) - /docs/ai/to-ai-messages — toAiMessages reference (was /docs/api/to-ai-messages) - /docs/ai/types — Reference for every type exported from chat/ai The section uses the same `"...folder"` flattening trick as API Reference, so the four pages render as flat siblings under the AI header instead of inside an accordion. The legacy `/docs/api/to-ai-messages` entry is removed from the API meta and the API overview now points at the new section. All cross-links in the streaming, conversation history, posting messages, and message reference pages are updated to the new URLs.
…command Wires the example app up to the new chat/ai surface so the new tools can be exercised end-to-end: - "Run Agent Demo" button on the welcome card runs a fixed, deterministic prompt that reacts to the trigger message, fetches recent messages, and posts a short markdown summary using the `messenger` preset. - `/agent <prompt>` slash command lets you invoke the same toolset with arbitrary natural-language instructions. The reply streams in via `ToolLoopAgent.stream(...).fullStream`, with an explicit "Agent thinking..." placeholder posted first so users get immediate feedback in channel contexts where Slack's typing-status API is a no-op (it requires a thread context). Approval gating is disabled in both flows for ease of testing — production bots should keep the default `requireApproval: true`. Also adds a `chat/ai` path mapping so the workspace TypeScript resolution lines up with the source-file mapping the example already uses for `chat`, and migrates the existing `toAiMessages` import to its new home.
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![vercel[bot]](https://avatars.githubusercontent.com/in/8329?s=60&v=4){kind=small}
`subscribeThread` and `unsubscribeThread` discarded the `ToolOptions` they received from `createChatTools`, so `needsApproval` was never set on the underlying AI SDK tool. Both write tools therefore ran without the approval gate the rest of the toolset enforces. Wire the option through and tighten the approval tests so every write tool is asserted under the default, global-off, and per-tool-override paths — covering both subscription tools.
HugoRCD
reviewed
May 12, 2026
HugoRCD
reviewed
May 12, 2026
- Swap the bash `pnpm add ai zod` snippets for the project's
`<PackageInstall>` component so the install instructions cover every
package manager and match the rest of the docs.
- Tighten the `Overview` example: drop the manual async-iterator buffer
and use the documented `thread.adapter.fetchMessages(thread.id, { limit })`
one-liner — same behavior, far less code, mirrors the canonical
`toAiMessages` example.
- Add an inline comment next to `requireApproval: false` in
`ai-sdk-tools.mdx` explaining the trade-off, so newcomers don't
silently disable the human-in-the-loop guard.
Standardize every model id in `bot.tsx` — the welcome-card responder plus the two `chat/ai` demos added in this PR (agent button and `/agent` slash command) — on `openai/gpt-5.4` so the example app runs on a single, current AI Gateway model.
`createChatTools` previously instantiated all 18 AI SDK tools — each with its own zod input schema and approval lookup — and only then filtered the result down to the requested preset. For an agent wired to `preset: "reader"` that meant building (and immediately discarding) every write tool, plus computing `requireApproval` for each. Convert the table to a record of factories and only invoke the ones the active preset allows. Behavior is unchanged — same return shape, same overrides semantics, same per-tool `needsApproval` defaults — but the unused tools and their schemas are no longer constructed at all.
HugoRCD
approved these changes
May 12, 2026
# Conflicts: # packages/adapter-web/package.json # pnpm-lock.yaml
dancer
approved these changes
May 15, 2026
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
This was referenced May 29, 2026
patrick-chinchill
added a commit
to Chinchill-AI/chat-sdk-python
that referenced
this pull request
May 30, 2026
…th (#116) * refactor(ai): convert chat_sdk.ai module to package for chat/ai subpath (vercel/chat#492) Mirror upstream PR vercel/chat#492 which split ai.ts into ai/messages.ts to make room for ai/tools.ts. Convert the chat_sdk.ai module into a package: - src/chat_sdk/ai/messages.py: to_ai_messages + supporting types, moved verbatim from the former ai.py (docstring path updated only). - src/chat_sdk/ai/__init__.py: re-export shim exposing every symbol the old module exported, so from chat_sdk.ai import to_ai_messages keeps working. - tests/test_ai.py renamed to tests/test_ai_messages.py (logic unchanged). No public import path changed. PR 1 of 3 (module move only); the tool factory (create_chat_tools) lands in PR 2. Design issue #109, tracking #98. https://claude.ai/code/session_01FyMxQn2BEAzmwKS1GZczKj * fix(scripts): update AI test-fidelity mapping after module move (review) * fix(ai): unbreak strict fidelity check + re-export TEXT_MIME_PREFIXES Two CI / review follow-ups on PR #116: 1. Fidelity-mapping fix went too far in the previous commit. The script's parity pin is `chat@4.26.0` (see `lint.yml:63`), where the upstream AI test file is still the flat `packages/chat/src/ai.test.ts` — the move to `ai/messages.test.ts` happened in 4.29.0. Pointing the mapping at the 4.29-era path made `verify_test_fidelity.py --strict` fail with "upstream checkout missing — mapped TS file not found". Revert just the TS half of the mapping to `ai.test.ts`; keep the Python half at `tests/test_ai_messages.py` (the file this PR actually renamed). When the parity pin moves to 4.29.0 the TS path can advance too. Locally verified against a fresh `chat@4.26.0` clone: TOTAL: 564/564 matched (100%), 0 missing All TS tests have Python equivalents. 2. The package shim was missing `TEXT_MIME_PREFIXES` (flagged by Codex P2). The PR description promises to "re-export everything the former `chat_sdk.ai` module exposed" — add the constant to both the import block and `__all__` so the contract holds. https://claude.ai/code/session_01FyMxQn2BEAzmwKS1GZczKj --------- Co-authored-by: Claude <noreply@anthropic.com>
patrick-chinchill
added a commit
to Chinchill-AI/chat-sdk-python
that referenced
this pull request
May 30, 2026
* feat(ai): add create_chat_tools tool factory (PR 2 of 3) (vercel/chat#492) Port of `packages/chat/src/ai/tools.ts` (plus the supporting `tools/{channels,messages,reactions,threads,users}.ts` and `types.ts`) introduced by vercel/chat#492. Builds on top of #116, which moved `chat_sdk.ai` from a single module to a package so `tools.py` has a home. `create_chat_tools(chat, preset=, require_approval=, overrides=)` returns a mapping of tool-name -> `ChatTool` dataclass holding a description, JSON-Schema-shaped `input_schema`, an async `execute` callable, and the `needs_approval` flag. Three presets (`reader` / `messenger` / `moderator`) and per-tool overrides match the upstream API surface verbatim. Individual factories (`post_message`, `add_reaction`, ...) are also exported so callers can cherry-pick. PR 3 will wire these tools into the existing handler paths; this PR adds the surface only. Validation: uv run ruff check src/ tests/ -> clean uv run ruff format --check src/ tests/ -> clean uv run python scripts/audit_test_quality.py -> 0 hard failures uv run pytest tests/ -q -> 4081 passed, 3 skipped Refs #98, #109. * fix(ai/tools): wrap ChatNotImplementedError + tighten preset check (gemini review) Four small follow-ups to PR 2 of the chat-ai port from Gemini's review: - Import `ChatNotImplementedError` alongside `ChatError`. - Wrap `fetch_channel_messages` invocation in `try/except ChatNotImplementedError` and re-raise as `ChatError` (preserves cause). The early `getattr(adapter, "fetch_channel_messages", None) is None` branch only catches missing attributes; `BaseAdapter` exposes the attribute as a stub that itself raises `ChatNotImplementedError`, so without this wrap the tool surfaces a different exception type than callers expect. Matches the `Chat.get_user` pattern. - Same wrap for `list_threads`. - `_resolve_preset_tools` now distinguishes a single `str` preset from any iterable of presets via `isinstance(preset, str)` rather than `isinstance(preset, list)`, so tuples/sets/Sequences work correctly. Two new regression tests (`test_fetch_channel_messages_wraps_not_implemented`, `test_list_threads_wraps_not_implemented`) drive an `AsyncMock` whose `side_effect` is `ChatNotImplementedError`, assert the surfaced exception is `ChatError` with the expected message, and pin the cause chain (`__cause__ is ChatNotImplementedError`). https://claude.ai/code/session_01FyMxQn2BEAzmwKS1GZczKj * fix(ai/tools): isolate per-tool input schemas to prevent mutation bleed (review) * fix(ai/tools): pass list_threads options as keyword (codex P2) `MockAdapter.list_threads` is declared as `list_threads(self, channel_id, **kwargs)`, so the tool's positional `list_method(channel_id, ListThreadsOptions(...))` raised `TypeError` for any consumer wiring `create_chat_tools` into MockAdapter (the SDK's own mock). Production adapters accept the kwarg form just as readily as the positional form, so the change is universally safe. The existing `test_list_threads_projects_summaries` used `AsyncMock(return_value=...)` which masked the TypeError by replacing the real `MockAdapter.list_threads` entirely. New `test_list_threads_uses_keyword_options` exercises the real MockAdapter and asserts the default empty result — fails on the old positional call with `TypeError`, passes after. https://claude.ai/code/session_01FyMxQn2BEAzmwKS1GZczKj * style(ai/tools): apply is-not-None idiom + hoist core import (audit nits) --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Introduces a dedicated
chat/aisubpath as the home for every Vercel AI SDK helper that ships with Chat SDK. Importing from this subpath keeps the optionalaiandzodpeer dependencies out of bundles that don't use them.What's new
createChatTools— exposes Chat SDK operations as ready-to-use AI SDK tools so an agent can read, post, react, edit, delete, and manage thread subscriptions across every adapter the suppliedChatinstance has registered.requireApproval: true); toggle globally or per-tool.reader,messenger,moderator— scope the toolset.import { postMessage, addReaction } from "chat/ai").toAiMessages(and theAi*/ToAiMessagesOptionstypes) now live alongside the tools atchat/ai. The previouschatre-exports continue to work, but are flagged@deprecatedwith an editor hint pointing to the new home — migration is a one-line import change./docs/aisection between Usage and Adapters in the sidebar:/docs/ai— Overview/docs/ai/ai-sdk-tools—createChatToolsguide/docs/ai/to-ai-messages—toAiMessagesreference/docs/ai/types— Reference for every type exported fromchat/aiexamples/nextjs-chatnow demos the new surface via a "Run Agent Demo" button on the welcome card and a free-form/agent <prompt>slash command (streaming, with a placeholder so users get immediate feedback in channel contexts where Slack's typing-status API is a no-op).Future plans
createChatToolscurrently exposes the cross-adapter Chat SDK surface only. A natural follow-up is to also support platform-specific tools — e.g. expose Slack-onlypin/unpin, Discord-only thread archiving, GitHub-only issue commenting, etc., so users can further extend what their agent can do without dropping back to raw adapter calls. The shape would likely be additional opt-in factories underchat/ai(or per-adapter subpaths like@chat-adapter/slack/ai) that return tools layered on top of the platform-specific adapter clients, while keeping the cross-platformcreateChatToolsAPI as the lowest common denominator.Coverage
createChatToolsorchestrator: 100% statements / 94.7% branches.execute()is exercised end-to-end (29 tests inindex.test.ts).toAiMessageskeeps its existing 35-test suite covering role mapping, attachment handling, links, transforms, and unsupported-attachment fallbacks.