vercel__chat/packages/chat/CHANGELOG.md

chat

4.33.0

Minor Changes

• ef2542c: add X (Twitter) adapter: reply to public mentions, send and receive direct messages, post and edit from the bot account, and like posts, using the X API v2 with OAuth 2.0 and managed token refresh

Patch Changes

• 3abdc69: docs(adapters): add Cloudflare Agents as a vendor-official state adapter (agents/chat-sdk) to the catalog and docs listing. It is hidden from the create-chat-sdk CLI (Worker/Durable Objects runtime), and the interactive state picker now filters out CLI-incompatible state adapters.
• 0b63791: Raise the default message dedupe TTL from 5 to 10 minutes so it outlives the longest platform redelivery window. Slack's Events API retries up to ~5 minutes after the original delivery — exactly at the old TTL boundary, where a retried event could miss the expired dedupe entry from its first processing and be handled twice. Configurable behavior is unchanged (dedupeTtlMs still overrides).
• 0c761f1: docs(adapters): add Dial as a vendor-official adapter (@getdial/chat-sdk-adapter) to the catalog, docs listing, and CLI scaffold spec
• 24a04d5: docs(adapters): add Photon as a vendor-official adapter (@photon-ai/chat-adapter-imessage) to the catalog, docs listing, and CLI scaffold spec
• 076fe5d: preserve skipped mention routing for debounce and message patterns

4.32.0

Minor Changes

• 2e47351: Add autoCompletePrevious option to Plan.addTask(). Defaults to true so sequential workflows keep auto-completing existing in-progress tasks; pass false to keep multiple tasks in progress for parallel execution.

Patch Changes

• eccc6b9: detect bot mentions in skipped queue and burst messages
• 438f551: Fix lightweight thread handles so streaming posts without an incoming message context no longer crash.
• d034b8b: docs(adapters): add Linq as a vendor-official adapter (@linqapp/chat-sdk-adapter) to the catalog, docs listing, and CLI scaffold spec
• 06af3e1: docs(adapters): add Novu as a vendor-official adapter (@novu/chat-sdk-adapter) to the catalog, docs listing, and CLI scaffold spec
• efa9610: Sync bundled KB resources from Edge Config: add four new guides (Vercel Connect, the Slack Vercel Connect bot, AI Gateway + AI SDK, and the daily digest bot), refresh existing guide bodies, and regenerate resources/templates.json. The sync-resources script now fetches and validates all guides before writing (so a failed fetch leaves the tree untouched), validates the source config shape, rejects duplicate slugs, retries transient fetches, and mirrors SKILL.md to all four committed copies.

4.31.0

Minor Changes

• 778ae69: Add a zero-dependency chat/adapters catalog subpath for official and vendor-official adapter metadata, environment specs, and setup helpers.
• 171657a: Adding support for stable IDs to link button action handlers

4.30.0

Minor Changes

• 5461ea9: Add native Telegram private chat draft streaming with fallback streaming elsewhere.

4.29.0

Minor Changes

• ac8a207: Add chat/ai subpath as the home for AI utilities, including createChatTools for the Vercel AI SDK and toAiMessages for converting chat history into AI SDK prompts.

  createChatTools exposes Chat SDK operations as ready-to-use AI SDK tools so an agent can read messages, post replies, send DMs, 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 tools can also be cherry-picked from the same subpath.

  toAiMessages (and the AiMessage / AiMessagePart / ToAiMessagesOptions types) now ship from chat/ai alongside the tools — keeping the optional ai and zod peer dependencies out of bundles that don't use them. The previous chat re-exports continue to work, but are marked @deprecated so editors surface a hint pointing at chat/ai; existing code keeps compiling, and migrating is a single import-path change.

• b75eedb: add burst concurrency strategy

Patch Changes

• e60bc8c: chore: set supported Node versions in engines

4.28.1

Patch Changes

• 0cc3d06: fix StreamEvent type from step-finish to finish-step

4.28.0

Minor Changes

• eb5f94a: add message.subject for resolving parent issues and PRs from webhook messages, expose typed platform client via adapter.client

• c1cd9b5: Add callbackUrl to Button and Modal. When a button is clicked or a modal is submitted, the SDK POSTs the action payload to callbackUrl in addition to firing any registered onAction / onModalSubmit handler. This pairs naturally with webhook-based workflow engines for awaitable button/modal flows.

  Supported platforms: Slack, Teams, Google Chat, WhatsApp, Telegram, and Discord.

• 46d183b: Rename messageHistory → threadHistory (with backwards compatibility).

  The per-thread history cache was previously named messageHistory, which collides conceptually with the new cross-platform per-user Transcripts API. Renamed to threadHistory to make the distinction clear.

  Renamed:

  • ChatConfig.messageHistory → ChatConfig.threadHistory
  • Adapter.persistMessageHistory → Adapter.persistThreadHistory
  • MessageHistoryCache → ThreadHistoryCache
  • MessageHistoryConfig → ThreadHistoryConfig
  • File message-history.ts → thread-history.ts

  Backwards compatibility:

  • The old ChatConfig.messageHistory field is still read; threadHistory takes precedence when both are set.
  • The old Adapter.persistMessageHistory flag is still read; either flag being true enables persistence.
  • MessageHistoryCache and MessageHistoryConfig are re-exported as deprecated aliases of the new names.
  • The state-adapter storage key prefix (msg-history:) is unchanged — renaming it would silently orphan existing data.

  The @chat-adapter/telegram and @chat-adapter/whatsapp adapters now use persistThreadHistory. Custom adapters built against persistMessageHistory continue to work unchanged.

• 46d183b: Add Transcripts API for cross-platform per-user message persistence.

  bot.transcripts (when configured via ChatConfig.transcripts + ChatConfig.identity) provides append / list / count / delete keyed by a stable cross-platform user key. Backed by the existing StateAdapter.appendToList primitive, so every built-in state adapter (memory, redis, ioredis, pg) supports it with no changes.

  • IdentityResolver runs once per inbound message during dispatch; the result is cached on the Message instance as message.userKey.
  • Distinct from the existing per-thread threadHistory config (which backfills thread context for adapters that lack server-side history).
  • delete wipes every stored entry under a user key. Single-entry and time-range deletes are not part of this API — the underlying appendToList primitive can't support them safely under concurrent writes.

• 3490a8c: Add @chat-adapter/web — a new platform adapter that lets a chat-sdk bot serve a browser chat UI alongside Slack/Teams/Discord, without writing any client-side glue.

  The adapter speaks the AI SDK UI message stream protocol, so @ai-sdk/react's useChat and the ai-elements component library work out of the box. The same bot.onDirectMessage(...) handler fires for both web and other platforms — including stream-based replies via thread.post(stream).

  Two subpath exports:

  • @chat-adapter/web — server-side createWebAdapter({ userName, getUser }) that produces an Adapter for the Chat constructor.
  • @chat-adapter/web/react — thin client wrapper exposing useChat() preconfigured with DefaultChatTransport. Re-exports UIMessage and UseChatHelpers types.

  // server
  const bot = new Chat({
    userName: "mybot",
    adapters: {
      web: createWebAdapter({
        userName: "mybot",
        getUser: (req) => ({ id: getUserIdFromCookie(req) }),
      }),
    },
    state: createMemoryState(),
  });
  export const POST = bot.webhooks.web;
  

  // client
  import { useChat } from "@chat-adapter/web/react";
  const { messages, sendMessage, status } = useChat();
  

  v1 covers text + markdown, native streaming, DM-style routing (isDM: true), persisted message history (persistMessageHistory: true by default — required for channel.messages since web has no platform history API), and abort propagation via request.signal. Out of scope for v1: cards/JSX rendering, reactions, modals, file uploads, edit/delete, and multi-tab proactive push.

Patch Changes

• 9824d33: Security fixes for HIGH-severity findings:

  • adapter-slack: Replace timing-unsafe !== with crypto.timingSafeEqual when validating the x-slack-socket-token header on forwarded socket-mode events.
  • adapter-github: In multi-tenant App mode, eagerly auto-detect the bot user ID on the first installation client / first webhook so isMe checks work and self-reply loops are prevented. Falls back to apps.getAuthenticated + users.getByUsername when users.getAuthenticated is unavailable for installation tokens.
  • adapter-linear: Add optional encryptionKey config (or LINEAR_ENCRYPTION_KEY env var) that AES-256-GCM-encrypts accessToken and refreshToken at rest in the state store. Tolerates plaintext records for zero-downtime rollout.
  • adapter-gchat: Fail-closed by default — the constructor now throws ValidationError if neither googleChatProjectNumber nor pubsubAudience is configured. To accept unverified webhooks (development only), set the new disableSignatureVerification: true flag (or GOOGLE_CHAT_DISABLE_SIGNATURE_VERIFICATION=true). Mirrors the Slack adapter's signing-secret requirement.
  • adapter-shared: New decodeKey / encryptToken / decryptToken / isEncryptedTokenData utilities (AES-256-GCM, hex or base64 32-byte keys), shared by Slack and Linear.

4.27.0

Minor Changes

• 1e7c551: restore attachment fetchData after queue/debounce serialization

• b0ab804: Bundle guide markdown and a templates manifest with the package so AI agents can discover Chat SDK resources without fetching at runtime. Guides are written to resources/guides/*.md and templates to resources/templates.json by the root-level pnpm sync-resources script.

• b9a1961: Switch Telegram adapter's outbound parse_mode from legacy Markdown to MarkdownV2, and replace the standard-markdown passthrough renderer with a proper AST → MarkdownV2 renderer. Standard markdown (**bold**) and legacy Markdown (*bold*) use different syntaxes and have no shared escape rules, so any message containing ., !, (, ), -, _ in regular text — which is virtually every LLM-generated message — was being rejected with can't parse entities. The new renderer walks the mdast tree and emits MarkdownV2 with context-aware escaping (normal text vs. code blocks vs. link URLs), uniformly applies MarkdownV2 parse_mode to every format-converter output (including AST messages, which previously shipped without parse_mode and rendered asterisks literally), and escapes card fallback text.

  Also fix silent message truncation that the MarkdownV2 migration widened from a rare bug into a reliable 400. The previous truncator sliced messages at 4096/1024 chars and appended literal ..., but in MarkdownV2 . is a reserved character that must be escaped, the slice can leave an orphan trailing \, and it can cut through a paired entity (*bold*, `code`) leaving it unclosed — all of which cause can't parse entities. The two truncate methods are unified into truncateForTelegram(text, limit, parseMode), which appends an escaped \.\.\. for MarkdownV2 and walks back past unbalanced entity delimiters or orphan backslashes before appending. Plain-text messages keep literal ....

  Internal typing hardening: renderMarkdownV2 is now typed exhaustively on mdast's Nodes union with a never assertion, so new mdast node types fail the build rather than silently falling through. Introduce TelegramParseMode = "MarkdownV2" | "plain" replacing the previous string | undefined at call sites, with toBotApiParseMode mapping to the Bot API wire format at the boundary. The chat package gains a re-export of mdast's Nodes union so adapters can build exhaustively typed renderers without importing mdast directly.

• a520797: Add chat.getUser() method and UserInfo type for cross-platform user lookups. Implement getUser on Slack, Discord, Google Chat, GitHub, Linear, and Telegram adapters.

• 70281dc: add initialOption and option_groups support for ExternalSelect

• 9093292: add streaming options to thread.post() with platform-specific namespacing

• 7e90d9c: Add Socket Mode support for environments behind firewalls that can't expose public HTTP endpoints, and add { action: "clear" } modal response to close the entire modal view stack

• bca4792: Allow task_update streaming chunks to include optional details text for Slack task cards

• 37dbb4a: Add thread.getParticipants() to get unique human participants in a thread

• 608d5f0: Add chat.thread(threadId) method to create Thread handles outside of webhook contexts

• a179b29: Implement external_select block kit for Slack

• a8f2aab: Allow plan.updateTask() to target a specific task by ID via { id: taskId } instead of always updating the last in_progress task

Patch Changes

• 8a0c7b3: Fix Slack structured streaming when thread.post(stream) is called from a handler created by an interactive (block_actions) payload. The team ID is now resolved from team.id in addition to team_id / team.
• d630e6c: fix(chat): honor concurrency.maxConcurrent in the concurrent strategy. The cap was documented but never applied, so handlers dispatched unbounded. Also warns when maxConcurrent is paired with a non-concurrent strategy (previously ignored silently) and throws on maxConcurrent < 1 to prevent a deadlock.

4.26.0

Minor Changes

• 2235c16: export standalone reviver for workflow-safe deserialization without adapter dependencies

Patch Changes

• ddb084b: guard fallback streaming against empty post and edit calls

4.25.0

Minor Changes

• 2700ce8: Allow Slack native streaming to send markdown tables without wrapping them in code fences, while preserving the previous append-only table fallback for other consumers.

4.24.0

Minor Changes

• 4f5d200: Add Teams dialog (task module) support with actionType: "modal" on buttons and onOpenModal webhook hook

Patch Changes

• 8d89274: fix: disable source maps in published packages
• 27b34e1: Use adapter to parse channel id for thread.channelId

4.23.0

Minor Changes

• 4166e09: Add channelVisibility enum to distinguish private, workspace, external, and unknown channel scopes. Implements getChannelVisibility() on the Adapter interface and Slack adapter, replacing the previous isExternalChannel boolean.

4.22.0

Minor Changes

• f2d8957: Implement new concurrency strategies for overlapping messages

4.21.0

Minor Changes

• e45a67f: Add optional disconnect() hook to the Adapter interface, called during chat.shutdown() for resource cleanup

Patch Changes

• 13ba1c7: Fix fromFullStream() step separator detection for AI SDK v5+: rename step-finish event check to finish-step
• 95fd8ce: Add missing toJSON() method declarations to Thread and Channel interfaces to match their implementations.

4.20.2

4.20.1

Patch Changes

• e206371: new toAiMessages API for history-to-AI-SDK transformation. And introduces LinkPreview object on Message
• 8d88b8c: Fix GitHub adapter streaming: accumulate text and post once instead of using the fallbackStream post+edit loop, which caused 422 errors from empty body edits during TTFT. Also log fallbackStream edit failures via the Logger instead of silently swallowing them.

4.20.0

4.19.0

Minor Changes

• eb49b2a: Add forceReleaseLock to StateAdapter and onLockConflict config option for interrupt/steerability of long-running handlers
• 5b41f08: Add thread.schedule() and ScheduledMessage type for scheduling messages to be sent at a future time. Slack adapter implements scheduling via chat.scheduleMessage API with cancel() support.
• c4b0e69: Tighten Adapter & StateAdapter interfaces: make channelIdFromThreadId required, make EphemeralMessage generic over TRawMessage, add satisfies Adapter to mock adapter, migrate remaining adapters to shared error types

4.18.0

Patch Changes

• a3cfc1a: AI SDK6 compat fixes and support for native slack tables

4.17.0

Patch Changes

• cc65dc3: fix: non-atomic message deduplication causes app_mention events to be silently dropped

4.16.1

Patch Changes

• 130e780: Fix openModal crash when action has no thread context (e.g. home tab buttons)
• ff954f9: Better streaming for AI SDK
• f27c89b: Improve StreamChunk type safety with discriminated union and fix url_verification security bypass

4.16.0

Minor Changes

• 02e7ef6: Implements table markdown rendering, and fully streaming markdown rendering including for Slack which has native streaming. Overhauls adapters to have better fallback-render behavior
• f0c7050: add onMemberJoinedChannel on slack adapter

Patch Changes

• 9522b04: Add disabled prop to Button() for Google Chat and Discord
• f5a75c9: Add fallbackStreamingPlaceholderText config option to suppress the initial "..." placeholder in post+edit fallback streaming
• 73de82d: Add remend for streaming markdown healing

4.15.0

Minor Changes

• 0f85031: Added configurable dedupeTtlMs option to ChatConfig

  Change default dedupe ttl to 5min

• 5b3090a: Add CardLink element

4.14.0

Minor Changes

• 90dc325: Add typing indicators for Slack adapter using Slack assistants API

4.13.4

Patch Changes

• 716ce2a: Automatically create new ConsoleLogger

4.13.3

Patch Changes

• ce33270: Currently deserialized threads cannot post streams to Slack due to lost metadata

4.13.2

Patch Changes

• 7d00feb: Add docs to package

4.13.1

4.13.0

Minor Changes

• f371c0d: feat(slack): full Slack Assistants API support

  • Route assistant_thread_started and assistant_thread_context_changed events
  • Add onAssistantThreadStarted and onAssistantContextChanged handler registration
  • Add setSuggestedPrompts, setAssistantStatus, setAssistantTitle methods on Slack adapter
  • Extend stream() to accept stopBlocks for Block Kit on stream finalization
  • Bump @slack/web-api to ^7.11.0 for chatStream support
  • Export all new types

4.12.0

Minor Changes

• 8c50252: Adding support for slash commands.

4.11.0

Minor Changes

• 417374b: Adding inline Select components and Radio buttons to cards

4.10.1

Patch Changes

• c99b183: Added support for creating modals from ephemeral messages.

4.10.0

Minor Changes

• c7d51cb: Added support for passing arbitrary metadata through the modal lifecycle via a new privateMetadata field.

4.9.1

4.9.0

4.8.0

Minor Changes

• cca9867: GitHub + Linear integrations

4.7.2

4.7.1

Patch Changes

• 160f1f7: Fetch relatedMessage separately from the event thread.

4.7.0

Minor Changes

• a13f43e: Add relatedThread and relatedMessage to modal events.

4.6.0

Minor Changes

• 68e3f74: Add component

4.5.0

Minor Changes

• efa6b36: add postEphemeral() for ephemeral messages

4.4.1

Patch Changes

• 1882732: Serde from chat singleton
• b5826c2: Adding private metadata field to onModalClose events.
• 9e8f9e7: Serde support

4.4.0

Minor Changes

• 8ca6371: Add support for modals, modal events, text inputs and selectors.

4.3.0

Minor Changes

• 498eb04: Discord support

Patch Changes

• d80ea3f: Refactor

4.2.0

Minor Changes

• 0b5197a: Fixed and tested fetchMessages and allMessages

4.1.0

Minor Changes

• 9b95317: Native streaming support

4.0.2

4.0.1

Patch Changes

• b27ea10: READMEs