Files
vercel__chat/packages/state-pg
github-actions[bot] e3c136b6dc chore(release): version packages (#710)
This PR was opened by the [Changesets
release](https://github.com/changesets/action) 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
## @chat-adapter/discord@4.35.0

### Minor Changes

- 26c0522: Add an opt-in channel allowlist for treating non-bot Discord
messages as directed to the bot without requiring a mention. Configure
via `respondToChannelIds` or the `DISCORD_RESPOND_TO_CHANNEL_IDS` env
var (comma-separated).

### Patch Changes

- b605cf6: Preserve Discord's angle-bracket syntax for suppressing link
previews when rendering markdown.
- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/slack@4.35.0

### Minor Changes

- bb7cd12: Expose sender email addresses on normalized incoming Slack
message authors. `message.author.email` is populated from the same
cached `users.info` lookup used for display names and requires the
`users:read.email` scope; without it the field stays undefined.
- 907450d: Enterprise Grid fixes:

- `handleOAuthCallback` now handles org-wide installs
(`is_enterprise_install`): Slack returns `team: null` for these, and the
installation is now keyed by the enterprise ID — the same key webhook
token resolution looks up — instead of failing. The result includes
`enterpriseId` and `isEnterpriseInstall`, and `SlackInstallation`
records both.
- Socket mode now resolves per-installation tokens for events, slash
commands, and interactive payloads in multi-workspace deployments
(matching the HTTP webhook path), and no longer drops `enterprise_id` /
`is_enterprise_install` / `is_ext_shared_channel` from event payloads.
- The user profile cache and display-name mention reverse index are now
scoped by installation in multi-workspace deployments, so profiles
fetched with one workspace's token no longer bleed into another and
mentions can no longer resolve to a same-named user from a different
workspace. Existing cache entries repopulate on first lookup
(single-workspace keys are unchanged). `withBotToken` accepts an
optional `{ installationId }` so proactive/cron posts outside webhook
handling scope these caches too.
- API calls made while handling an event from an org-wide install now
pass the event's `team_id` explicitly, as Slack requires for
workspace-scoped methods (`conversations.list`, `usergroups.*`, …) on
org tokens. When an event carries a `context_team_id` (shared channels
hosted on an "away" workspace), channel-addressed calls echo it back as
`client_context_team_id`.
- Retried event deliveries (`x-slack-retry-num`, socket `retry_num`) are
dropped when the original delivery was already dispatched, using an
`event_id` marker in the state adapter (24-hour TTL). Events whose first
delivery never arrived are still recovered via the retry.
- Bare `@W…` mentions in outgoing messages are now recognized as raw
Enterprise Grid user IDs (previously only `@U…` was), so they render as
real mentions instead of being treated as display names.
- Event token resolution now prefers the envelope's `authorizations[0]`
— Slack's documented location for the event's installation identity —
over the top-level `team_id`/`enterprise_id`, which can name a different
workspace for Slack Connect shared-channel events. Top-level fields
remain as a fallback.

### Patch Changes

- 80def3a: Add optional `isSystem` field to the normalized message
`Author` type to distinguish platform-generated messages from humans and
bots. The Slack adapter now sets `isSystem: true` for messages authored
by Slack's reserved `USLACK` user (e.g. "@user archived the channel"
notifications in DMs), so consumers no longer need to hard-code
Slack-specific user IDs.
- 92530dd: Return a replyable Slack thread ID from `channel.post()` by
using the posted top-level message's timestamp as the thread root.
- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/teams@4.35.0

### Minor Changes

- 5eb8b84: Add support for adding and removing reactions from Microsoft
Teams messages.
- 46681f5: Expose Microsoft Graph email addresses on normalized incoming
Teams message authors. Resolved user profiles are cached in the state
adapter (1 hour, failed lookups 5 minutes) so the lookup doesn't add a
Graph call per message.
- 160140e: Add native Microsoft Teams targeted message support via
`thread.postEphemeral()` and `channel.postEphemeral()`.
- e06b4b6: Add a `token` config option to `TeamsAdapterConfig` for
supplying a custom token factory, forwarded to the Teams SDK's
`AppOptions.token`. This lets bots authenticate on runtimes that can't
reach Azure IMDS (so `federated` managed identity isn't reachable) but
can still mint access tokens through an external mechanism, without
needing a static client secret.

### Patch Changes

- 3895ab3: Fall back to Microsoft Graph's user principal name when a
Teams user has no mail address.
- 93a58af: Show explicitly configured progress as a native Teams DM
status while preserving native streaming.
- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/telegram@4.35.0

### Minor Changes

- 54eea71: Add an optional Telegram user allowlist via `allowedUserIds`
or the comma-separated `TELEGRAM_ALLOWED_USER_IDS` environment variable.

### Patch Changes

- 0701679: Cache the compiled bot-mention regex in `isBotMentioned`
instead of recompiling it per message, and make the protected `sleep`
helper accept an optional `AbortSignal` so `stopPolling()` interrupts
the polling backoff delay immediately.
- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## chat@4.35.0

### Minor Changes

- 4cb7e5d: Add a `chat/workflow` subpath with `requestApproval()`:
durable human-in-the-loop approvals built on Workflow SDK. It posts an
approval card with Approve/Deny buttons, suspends the workflow until a
user decides (or an optional timeout elapses), validates approvers,
finalizes the card with the outcome, and returns `{ approved, timedOut,
user }`. Also exports the `buildApprovalCard` and `buildResolvedCard`
builders. Requires the new optional `workflow` peer dependency.
- 46681f5: Expose Microsoft Graph email addresses on normalized incoming
Teams message authors. Resolved user profiles are cached in the state
adapter (1 hour, failed lookups 5 minutes) so the lookup doesn't add a
Graph call per message.

### Patch Changes

- 80def3a: Add optional `isSystem` field to the normalized message
`Author` type to distinguish platform-generated messages from humans and
bots. The Slack adapter now sets `isSystem: true` for messages authored
by Slack's reserved `USLACK` user (e.g. "@user archived the channel"
notifications in DMs), so consumers no longer need to hard-code
Slack-specific user IDs.
- 93a58af: Show explicitly configured progress as a native Teams DM
status while preserving native streaming.
- 25f3099: `toAiMessages` no longer drops messages that have no text. A
message with an empty text body is now kept when it has links or
attachments the converter can include: images and text files (`text/*`,
JSON, XML, YAML, etc.) with a working `fetchData()`. Messages whose only
attachments are unsupported (video, audio, other file types, or
attachments without `fetchData()`) are still skipped, and
`onUnsupportedAttachment` now fires for video/audio attachments on these
previously filtered messages.

Note: multipart `content` no longer always starts with a text part. When
a kept message had no text, its `content` array contains only attachment
parts.
## @chat-adapter/gchat@4.35.0

### Patch Changes

- 270b1c2: fix(gchat): accept `endpointUrl` as a direct-webhook verifier
and verify each token type correctly

When a Google Chat app's connection setting **Authentication audience**
is set
to **HTTP endpoint URL** — Google's recommended option for HTTP-hosted
apps
not behind Cloud Run IAM, and the only mode available for Workspace
Add-on
Chat apps — incoming tokens are Google OIDC ID tokens whose `aud` is the
endpoint URL rather than the GCP project number. Previously the adapter
only
verified against `googleChatProjectNumber`, so URL-audience tokens
always
failed with 401 Unauthorized. The adapter now accepts `endpointUrl` as a
direct-webhook verifier (including in the constructor's fail-closed
check),
validating the OIDC token's audience plus the Google Chat issuer email
claims
  (`chat@system.gserviceaccount.com`, or the
`service-{projectNumber}@gcp-sa-gsuiteaddons.iam.gserviceaccount.com`
service
identity for Workspace Add-on Chat apps) with `email_verified: true` — a
public endpoint URL audience alone is not sufficient to forge a request.

  Project-number-audience tokens are now verified per Google's reference
  implementation: they are JWTs self-signed by
`chat@system.gserviceaccount.com`, so the adapter checks them against
that
  service account's X.509 certificates with issuer
`chat@system.gserviceaccount.com` (previously it used `verifyIdToken`,
which
only accepts Google OIDC issuers and certs and therefore rejected every
real
project-number token). When both verifiers are configured, either token
type
  is accepted.

  The adapter still infers an endpoint URL from incoming requests for
button-click action routing only — that inferred value is never used as
a
JWT verification audience, and inference now only happens after a
request
has passed verification (or verification was explicitly disabled),
because
  `request.url` derives from the attacker-controllable `Host` header in
  serverless runtimes.

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/github@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/linear@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/messenger@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/shared@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
## @chat-adapter/twilio@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/web@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/whatsapp@4.35.0

### Patch Changes

- 09b72e9: fix whatsapp card media duplication

- Prevent card titles and body content from appearing twice when sending
cards with files on WhatsApp.
- Avoid adding the full card fallback text as an image caption when an
interactive message follows.
- Keep interactive WhatsApp messages responsible for rendering card
titles, bodies, and actions.

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/x@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
  - @chat-adapter/shared@4.35.0
## @chat-adapter/state-ioredis@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
## @chat-adapter/state-memory@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
## @chat-adapter/state-pg@4.35.0

### Patch Changes

- d88789c: Fix `setIfNotExists()` so it can claim a cache key whose
existing row has expired. Previously the query used `ON CONFLICT DO
NOTHING`, so an expired row in `chat_state_cache` still blocked
acquisition until opportunistic cleanup deleted it — diverging from the
memory and Redis adapters, which treat expired entries as absent. Keys
stored without a TTL remain permanent and are never overwritten.
- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
## @chat-adapter/state-redis@4.35.0

### Patch Changes

- Updated dependencies [80def3a]
- Updated dependencies [4cb7e5d]
- Updated dependencies [46681f5]
- Updated dependencies [93a58af]
- Updated dependencies [25f3099]
  - chat@4.35.0
## @chat-adapter/tests@4.35.0

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
2026-07-24 22:21:13 +01:00
..

@chat-adapter/state-pg

npm package: @chat-adapter/state-pg

Agent Stack MIT License

Production PostgreSQL state adapter for Chat SDK built with pg (node-postgres). Use this when PostgreSQL is your primary datastore and you want state persistence without a separate Redis dependency.

Documentation: chat-sdk.dev/adapters/official/postgres · Guides: vercel.com/kb/chat-sdk

Installation

pnpm add @chat-adapter/state-pg

Scaffold with the CLI

To scaffold a new Slack bot that uses PostgreSQL for state:

npx create-chat-sdk@latest my-bot --adapter slack postgres

Visit the adapters directory to see other available official and vendor-official adapters.

Usage

createPostgresState() auto-detects POSTGRES_URL (or DATABASE_URL) so you can call it with no arguments:

import { Chat } from "chat";
import { createPostgresState } from "@chat-adapter/state-pg";

const bot = new Chat({
  userName: "mybot",
  adapters: { /* ... */ },
  state: createPostgresState(),
});

To provide a URL explicitly:

const state = createPostgresState({
  url: "postgres://postgres:postgres@localhost:5432/chat",
});

Using an existing client

import pg from "pg";

const client = new pg.Pool({ connectionString: process.env.POSTGRES_URL! });
const state = createPostgresState({ client });

Configuration

Option Required Description
url No* Postgres connection URL
client No Existing pg.Pool instance
keyPrefix No Prefix for all state rows (default: "chat-sdk")
logger No Logger instance (defaults to ConsoleLogger("info").child("postgres"))

*Either url, POSTGRES_URL/DATABASE_URL, or client is required.

Environment variables

POSTGRES_URL=postgres://postgres:postgres@localhost:5432/chat

Data model

The adapter creates these tables automatically on connect():

chat_state_subscriptions
chat_state_locks
chat_state_cache
chat_state_lists
chat_state_queues

All rows are namespaced by key_prefix.

Features

Feature Supported
Persistence Yes
Multi-instance Yes
Subscriptions Yes
Distributed locking Yes
Key-value caching Yes (with TTL)
Automatic table creation Yes
Key prefix namespacing Yes

Locking considerations

The Redis state adapters use atomic SET NX PX for lock acquisition, which is a single atomic operation. The PostgreSQL adapter uses INSERT ... ON CONFLICT DO UPDATE WHERE expires_at <= now(), which relies on Postgres row-level locking. This is safe for most workloads but under extreme contention (many processes competing for the same lock simultaneously) may behave slightly differently than Redis. For high-contention distributed locking, prefer the Redis adapter.

Expired row cleanup

Unlike Redis (which handles TTL expiry natively), PostgreSQL does not automatically delete expired rows. The adapter performs opportunistic cleanup — expired locks are overwritten on the next acquireLock() call, expired cache entries are deleted on the next get() call for that key, and expired queue entries for a given thread are purged on the next enqueue() or dequeue() call. Expired list entries are filtered out on read but never deleted by the adapter.

For high-throughput deployments, you may want to run a periodic cleanup job:

DELETE FROM chat_state_locks WHERE expires_at <= now();
DELETE FROM chat_state_cache WHERE expires_at <= now();
DELETE FROM chat_state_lists WHERE expires_at <= now();
DELETE FROM chat_state_queues WHERE expires_at <= now();

AI Coding Agents

If you use an AI coding agent such as OpenAI Codex, Claude Code, or Cursor, install the Chat SDK skill so it knows the SDK APIs, adapter patterns, and project conventions before writing code.

npx skills add vercel/chat

The skill references bundled documentation in node_modules/chat/docs, plus adapter guides and starter templates in the published package.

You can also install the Vercel Plugin for a broader agent toolkit — it includes the Chat SDK skill alongside specialist agents, agent slash commands, and more:

npx plugins add vercel/vercel-plugin

The plugin is optional; the skill alone is enough to build with Chat SDK.

For agent-readable documentation, see chat-sdk.dev/llms.txt (page index) or chat-sdk.dev/llms-full.txt (full text).

License

MIT