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 ## @langchain/langgraph-checkpoint-mongodb@1.4.0 ### Minor Changes - [#1928](https://github.com/langchain-ai/langgraphjs/pull/1928) [`3d7fcea`](https://github.com/langchain-ai/langgraphjs/commit/3d7fcea7d7ea7f1203d24be9df607d5a8e8717bc) Thanks [@Mihailoff](https://github.com/Mihailoff)! - Add TTL support for automatic checkpoint expiration - Add optional `ttl` parameter to MongoDBSaver (value in seconds) - Add `setup()` method to create TTL indexes on collections - Add `upserted_at` timestamp to documents when TTL is enabled - Each write refreshes TTL (expires after inactivity, not creation) ### Patch Changes - [#2556](https://github.com/langchain-ai/langgraphjs/pull/2556) [`bee3c91`](https://github.com/langchain-ai/langgraphjs/commit/bee3c91d0adc315ebde0622d8c4b1fff041c1bfd) Thanks [@mohamedkhaled4053](https://github.com/mohamedkhaled4053)! - Fix `MongoDBSaver.putWrites` throwing `MongoServerError: Invalid BulkOperation, Batch cannot be empty` when called with an empty `writes` array. This is reached by human-in-the-loop / `interrupt()` flows, where a task can complete producing zero channel writes and LangGraph calls `putWrites(config, [], taskId)`. `putWrites` now no-ops on empty writes, matching the behavior of the postgres and sqlite savers (which iterate and naturally skip empty batches). - [#2550](https://github.com/langchain-ai/langgraphjs/pull/2550) [`2b8cc2f`](https://github.com/langchain-ai/langgraphjs/commit/2b8cc2f3fd5c9d3c33b56e013292daf5d936428e) Thanks [@lazydiv](https://github.com/lazydiv)! - feat(checkpoint-mongodb): add setup() to create required indexes ## @langchain/langgraph-api@1.4.0 ### Minor Changes - [#2559](https://github.com/langchain-ai/langgraphjs/pull/2559) [`48cbdd2`](https://github.com/langchain-ai/langgraphjs/commit/48cbdd23fdf29277530f6aa05c397c9902e81206) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langgraph-cli): add `deploy` command for LangSmith Deployment Port the Python CLI's `langgraph deploy` workflow to `@langchain/langgraph-cli`, including local and remote build paths, deployment lifecycle subcommands (`list`, `revisions list`, `delete`, `logs`), and host-backend client utilities with tests. ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langgraph-api): preserve non-empty response_metadata on protocol-v2 state messages The protocol-v2 state normalizer stripped `response_metadata` from messages, dropping data that HITL flows rely on — an interrupt's card is carried on `AIMessage.response_metadata` (e.g. `{ cards: ... }`). Non-empty `response_metadata` is now retained so the card reaches the client. - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`48cbdd2`](https://github.com/langchain-ai/langgraphjs/commit/48cbdd23fdf29277530f6aa05c397c9902e81206)]: - @langchain/langgraph-ui@1.4.0 ## @langchain/langgraph-cli@1.4.0 ### Minor Changes - [#2559](https://github.com/langchain-ai/langgraphjs/pull/2559) [`48cbdd2`](https://github.com/langchain-ai/langgraphjs/commit/48cbdd23fdf29277530f6aa05c397c9902e81206) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langgraph-cli): add `deploy` command for LangSmith Deployment Port the Python CLI's `langgraph deploy` workflow to `@langchain/langgraph-cli`, including local and remote build paths, deployment lifecycle subcommands (`list`, `revisions list`, `delete`, `logs`), and host-backend client utilities with tests. ### Patch Changes - Updated dependencies \[[`48cbdd2`](https://github.com/langchain-ai/langgraphjs/commit/48cbdd23fdf29277530f6aa05c397c9902e81206), [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e), [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-api@1.4.0 ## @langchain/langgraph-ui@1.4.0 ### Minor Changes - [#2559](https://github.com/langchain-ai/langgraphjs/pull/2559) [`48cbdd2`](https://github.com/langchain-ai/langgraphjs/commit/48cbdd23fdf29277530f6aa05c397c9902e81206) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langgraph-cli): add `deploy` command for LangSmith Deployment Port the Python CLI's `langgraph deploy` workflow to `@langchain/langgraph-cli`, including local and remote build paths, deployment lifecycle subcommands (`list`, `revisions list`, `delete`, `logs`), and host-backend client utilities with tests. ## @langchain/langgraph@1.4.5 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 ## @langchain/langgraph-sdk@1.9.24 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. ## @langchain/angular@1.0.25 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 ## @langchain/react@1.0.25 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 ## @langchain/svelte@1.0.25 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 ## @langchain/vue@1.0.25 ### Patch Changes - [#2557](https://github.com/langchain-ai/langgraphjs/pull/2557) [`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(sdk): apply state update and goto alongside interrupt resume `respond(decision, { update, goto })` now maps to LangGraph's `Command(resume, update, goto)`, so a human-in-the-loop UI can commit a state update (e.g. push the interrupt card into state) in the **same superstep** as the resume — one checkpoint, no separate `updateState` write, no flicker. `@langchain/langgraph-api` forwards `update`/`goto` through `input.respond`, and `@langchain/core` message instances in `update` are serialized to dicts before transport, exactly like `submit()`. Bumps `@langchain/protocol` to `^0.0.18` for the `Goto` type. `respond`/`respondAll` also apply `update` **optimistically** (mirroring `submit()`): the pushed messages paint immediately, with stable ids minted so the resumed run's echo reconciles them in place. Without this the interrupt is cleared the instant `respond()` dispatches while the pushed card only reappears a server round-trip later — so the card would flicker in that gap. The optimistic state settles on the resumed run's terminal (pending → sent, or rolled back on a failure before any echo). User-initiated optimistic writes (`submit()` / `respond()` / `respondAll()`) now commit to the store **synchronously**, in the same tick as the triggering event, instead of being coalesced onto the next macrotask. This lets a framework render the pushed message in the **same commit** as any local UI state the caller flips alongside it (e.g. a HITL form swapping its inputs for the resolved card), so the card no longer blinks out for the one-macrotask window before the flush lands. High-frequency streaming writes keep their macrotask coalescing. - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 ## @example/ai-elements@0.1.41 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/react@1.0.25 - @langchain/langgraph@1.4.5 ## @examples/assistant-ui-claude@0.1.41 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/react@1.0.25 - @langchain/langgraph@1.4.5 ## @examples/ui-angular@0.0.51 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 - @langchain/angular@1.0.25 - @langchain/langgraph@1.4.5 ## @examples/ui-multimodal@0.0.27 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/react@1.0.25 - @langchain/langgraph@1.4.5 ## @examples/ui-react@0.0.27 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph-sdk@1.9.24 - @langchain/react@1.0.25 - @langchain/langgraph@1.4.5 ## langgraph@1.0.45 ### Patch Changes - Updated dependencies \[[`b1e856d`](https://github.com/langchain-ai/langgraphjs/commit/b1e856d987ac16148dc0872d1fecf70e659ef28e)]: - @langchain/langgraph@1.4.5 Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
@langchain/svelte
Svelte 5 SDK for Deep Agents, LangChain and LangGraph.
useStream binds a LangGraph agent into a Svelte 5 component. Reactive fields are exposed as getters on a stable handle (stream.messages, stream.isLoading, …) so templates and $derived expressions track updates automatically — no stores, no $ prefix, no destructuring.
Installation
npm install @langchain/svelte @langchain/core
Peer dependencies: svelte ^5.0.0, @langchain/core ^1.0.1
Quick start
<script lang="ts">
import { useStream } from "@langchain/svelte";
const stream = useStream({
assistantId: "agent",
apiUrl: "http://localhost:2024",
});
</script>
{#each stream.messages as msg (msg.id)}
<div>{msg.content}</div>
{/each}
<button
disabled={stream.isLoading}
onclick={() =>
stream.submit({ messages: [{ type: "human", content: "Hello!" }] })}
>
Send
</button>
Note: Access fields through the live
streamhandle. Destructuring (const { messages } = stream) freezes the values at that moment — usestream.messagesin templates instead.
Highlights
- v2-native streaming protocol. Session-based transport with automatic re-attach on remount; no more
reconnectOnMount/joinStreamdance. - Always-on root projections.
values,messages,toolCalls, andinterruptsare reactive at the root with zero extra subscription cost. - Selector composables for scoped data. Per-subagent / per-subgraph messages, tool calls, and media stream only when a component actually mounts the matching composable, and release on unmount.
- Discriminated option bag. The hosted Agent Server path and the custom-adapter path are two arms of a single typed union — mixing them is a compile-time error.
- Reactive
threadId. PassthreadId: () => activeto drive in-place thread swaps without remounting. - Agent-brand type inference.
useStream<typeof agent>()unwraps state, tool calls, and subagent state maps from an agent brand. - Multimodal media streams. Built-in assembly for audio, images, video, and files — plus opinionated playback helpers.
- Headless tools. Register local tool implementations that auto-resolve server-emitted tool-call interrupts without a round-trip through the UI.
Documentation
In-depth guides live in docs/:
useStream— options, return shape, reactivethreadId- Selector composables (
useMessages,useToolCalls,useValues, …) - Interrupts,
respond(),stop(),hydrationPromise - Submission queue
- Stream context (
provideStream/getStream) - Headless tools
- Custom transport (
AgentServerAdapter,HttpAgentServerAdapter) - Media (images, audio, video, files) & playback helpers
- Type safety
Migrating from v0
@langchain/svelte v1 targets the v2 streaming protocol. The useStream import stays the same, but the option bag, return shape, and how you subscribe to scoped data all change. Most chat apps migrate in well under an hour — the full guide with line-by-line diffs lives in docs/v1-migration.md.
Playground
For full end-to-end examples, see the LangChain UI Playground.
License
MIT