# langchain

## 1.5.14

### Patch Changes

- [#11712](https://github.com/langchain-ai/langchainjs/pull/11712) [`34edee1`](https://github.com/langchain-ai/langchainjs/commit/34edee1153f0fa40ba6997c8ff7bd45834ca8ce8) Thanks [@blove](https://github.com/blove)! - `createAgent` no longer ends the run when a `returnDirect` tool fails. A tool error, including arguments that fail the tool's schema, now goes back to the model so it can correct the call and retry; the run still ends on the first successful `returnDirect` result. Error `ToolMessage`s produced by the tool node now carry `status: "error"`, and a `ToolMessage` returned from a custom `handleToolErrors` function defaults to that status when it sets none.

## 1.5.13

### Patch Changes

- [#11698](https://github.com/langchain-ai/langchainjs/pull/11698) [`1f66c83`](https://github.com/langchain-ai/langchainjs/commit/1f66c838c344a65581cce6fc22699cd6fc7194df) Thanks [@thushanth-bengre-langchain](https://github.com/thushanth-bengre-langchain)! - `createAgent` now uses the OpenAI Responses API by default when `model` is an `openai:` string. To keep using Chat Completions, pass a model instance, e.g. `new ChatOpenAI({ model: "gpt-5.5", useResponsesApi: false })`.

## 1.5.12

### Patch Changes

- [#11682](https://github.com/langchain-ai/langchainjs/pull/11682) [`a8c336f`](https://github.com/langchain-ai/langchainjs/commit/a8c336f3c75b76e9d94236cb4a05f2b2f414cac6) Thanks [@casparb](https://github.com/casparb)! - Preserve ReactAgent callbacks when its compiled graph is extracted, without duplicate calls through the agent.

## 1.5.11

### Patch Changes

- [#11436](https://github.com/langchain-ai/langchainjs/pull/11436) [`702be70`](https://github.com/langchain-ai/langchainjs/commit/702be70bcaadbfe9a47a3fcccbb75bdea17cae04) Thanks [@chiliec](https://github.com/chiliec)! - fix(agents): ignore thought/reasoning text blocks when parsing structured output in `ProviderStrategy`

## 1.5.11-rc.0

### Patch Changes

- [#11568](https://github.com/langchain-ai/langchainjs/pull/11568) [`fbc3206`](https://github.com/langchain-ai/langchainjs/commit/fbc32067aee6068dd7f96aff1889deba76d5c838) Thanks [@hntrl](https://github.com/hntrl)! - Add opt-in middleware `tracePolicy` support for transforming hook span input and output payloads, plus re-exports of LangGraph's `TracePolicy` and `omitPayload`.

## 1.5.10

### Patch Changes

- [#11412](https://github.com/langchain-ai/langchainjs/pull/11412) [`7df258c`](https://github.com/langchain-ai/langchainjs/commit/7df258c0af8362fede14d42bb982597a56f41b78) Thanks [@hntrl](https://github.com/hntrl)! - chore(langgraph): update langgraph deps to track serialization fix

## 1.5.9

### Patch Changes

- [#11369](https://github.com/langchain-ai/langchainjs/pull/11369) [`d6ad973`](https://github.com/langchain-ai/langchainjs/commit/d6ad9735640a6c729f81ef79361acc5e83c526f1) Thanks [@hntrl](https://github.com/hntrl)! - fix(langchain): use unified endpoint for gateway

- [#11342](https://github.com/langchain-ai/langchainjs/pull/11342) [`3b0e4c4`](https://github.com/langchain-ai/langchainjs/commit/3b0e4c48a31811031a460c4d95519a7c1163dc41) Thanks [@thushanth-bengre-langchain](https://github.com/thushanth-bengre-langchain)! - feat(core): mark errors as retryable or not, and stop retrying the ones that aren't

  Retry middleware retried every failure up to `maxRetries`, including deterministic ones like a bad API key or an unknown model. Retries also nest, so a single such failure could cost dozens of API calls.

  `@langchain/core/errors` adds `stampRetryable(error, retryable)` and `getRetryable(error)`. Marking an error leaves its class and shape untouched, so a provider SDK error can be classified without breaking `instanceof`. `getRetryable` returns `undefined` for errors nobody classified, and both are exported so tool authors can mark their own failures.

  `modelRetryMiddleware` and `toolRetryMiddleware` now respect the mark by default, and retries stop as soon as one is found rather than each layer spending its own budget. Aborted calls, context overflow, and oversized payloads are marked non-retryable out of the box.
  Models accept a per-call `maxRetries` so a surrounding retry loop can take over.

  **Behavior change:** errors marked non-retryable now fail on the first attempt. Unclassified errors — including any from third-party integrations or custom tools — retry exactly as before. Pass `retryOn: () => true` to restore the old default. A custom `onFailedAttempt` replaces the built-in handler and opts out of marking.

## 1.5.8

### Patch Changes

- [#11366](https://github.com/langchain-ai/langchainjs/pull/11366) [`c068bbf`](https://github.com/langchain-ai/langchainjs/commit/c068bbf8c113132bf16ac7a8add44e486a147b41) Thanks [@hntrl](https://github.com/hntrl)! - fix(core,langchain): patch and release core, update peer dependencies

## 1.5.7

### Patch Changes

- [#11363](https://github.com/langchain-ai/langchainjs/pull/11363) [`2d1f744`](https://github.com/langchain-ai/langchainjs/commit/2d1f7449c95a554d429535abae0115b4f578bbac) Thanks [@hntrl](https://github.com/hntrl)! - feat(langchain): add langsmith gateway to initChatModel- [#11362](https://github.com/langchain-ai/langchainjs/issues/11362)

## 1.5.6

### Patch Changes

- [#11331](https://github.com/langchain-ai/langchainjs/pull/11331) [`18765b0`](https://github.com/langchain-ai/langchainjs/commit/18765b002c3819bc4dc42123a293cab27a35ce4f) Thanks [@thushanth-bengre-langchain](https://github.com/thushanth-bengre-langchain)! - fix(langchain): exclude middleware-internal model calls from the message projection

  Bookkeeping model calls made by `summarizationMiddleware` and `toolEmulatorMiddleware` no longer appear in `run.messages` or `stream({ streamMode: "messages" })`, and the summary `summarizationMiddleware` writes back to state is no longer projected as a new message. These calls remain observable via `streamEvents({ version: "v2" })`, identified by `lc_source`.

- [#11344](https://github.com/langchain-ai/langchainjs/pull/11344) [`f08e0c6`](https://github.com/langchain-ai/langchainjs/commit/f08e0c6d50156accf95a36469a3e107a5598a3a0) Thanks [@hntrl](https://github.com/hntrl)! - fix: apply [Symbol.hasInstance] method to all comparable properties using .isInstance()

  We have some internal schemas that rely on `z.instanceof()`. This uses a strict `instanceof` check which can conflict if there are multiple versions of core installed. This overrides the [Symbol.hasInstance](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Symbol/hasInstance) method to use the same logic as `.isInstance()` to compare objects at runtime.

## 1.5.5

### Patch Changes

- [#11277](https://github.com/langchain-ai/langchainjs/pull/11277) [`0bd12f0`](https://github.com/langchain-ai/langchainjs/commit/0bd12f0bbb3c3f71e016f2ac26fd63a0f6ad0556) Thanks [@vladislav-nechakhin](https://github.com/vladislav-nechakhin)! - fix(langchain): resolve LocalFileStore keys with path.relative

## 1.5.4

### Patch Changes

- [#11201](https://github.com/langchain-ai/langchainjs/pull/11201) [`a8cd4d9`](https://github.com/langchain-ai/langchainjs/commit/a8cd4d9ae35ae8ae05e011115a351d976560ade9) Thanks [@hntrl](https://github.com/hntrl)! - feat(langchain): add middleware for handling tool execution errors

## 1.5.3

### Patch Changes

- [#11167](https://github.com/langchain-ai/langchainjs/pull/11167) [`4c7a06e`](https://github.com/langchain-ai/langchainjs/commit/4c7a06ec3e4728bb036a28f678dd8ec2521e1c82) Thanks [@colifran](https://github.com/colifran)! - fix(langchain): malformed tool input schemas are unrecoverable

## 1.5.2

### Patch Changes

- [#11100](https://github.com/langchain-ai/langchainjs/pull/11100) [`3205b35`](https://github.com/langchain-ai/langchainjs/commit/3205b35ac83037a2fff2998f16a66b5126b306f8) Thanks [@colifran](https://github.com/colifran)! - fix(langchain, openai): decouple strict tools from strict structured output response

## 1.5.1

### Patch Changes

- [#11087](https://github.com/langchain-ai/langchainjs/pull/11087) [`534b43a`](https://github.com/langchain-ai/langchainjs/commit/534b43a0719c89e3fb4ab7052d3f39797660aa43) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): keep tool call streams pending across any tool interrupt

  A raw `interrupt()` raised from inside a tool surfaced as a `tool-error` and
  rejected the call's un-awaited `output` promise, producing an unhandled
  rejection that crashed HITL runs. The tool-call stream transformer now treats
  any serialized graph interrupt as control flow (the call stays pending and
  resumes), not just `humanInTheLoopMiddleware` interrupts.

## 1.5.0

### Minor Changes

- [#11062](https://github.com/langchain-ai/langchainjs/pull/11062) [`3ebd10a`](https://github.com/langchain-ai/langchainjs/commit/3ebd10ae381ef7c9e12759f08d2f2b745bdfae62) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): surface tool-dispatched subagents on `run.subagents`

  Add a native subagent stream transformer to `createAgent` so v3 runs expose
  named nested agents (`createAgent({ name })` invoked from tools) as typed
  `SubagentRunStream` handles with `name`, `cause`, scoped `messages` /
  `toolCalls`, and `output`. Refactors agent stream transformers into
  `agents/transformers/` and exports only the public stream types from the
  package entry.

## 1.4.6

### Patch Changes

- [#11012](https://github.com/langchain-ai/langchainjs/pull/11012) [`4aeaa76`](https://github.com/langchain-ai/langchainjs/commit/4aeaa767a48dcc5451371f14adf4a06f064d23b9) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): add `when` predicate to human-in-the-loop middleware

  Add an optional `when` callback on `InterruptOnConfig` so callers can
  dynamically skip interrupts for specific tool calls. The predicate receives
  a `ToolCallRequest` (batch `afterModel` context) and returns whether to
  interrupt or auto-approve, matching Python `HumanInTheLoopMiddleware`.

- [#11080](https://github.com/langchain-ai/langchainjs/pull/11080) [`e51478a`](https://github.com/langchain-ai/langchainjs/commit/e51478a4436d69a21f75f1c9e0bb84f3abb6d6f5) Thanks [@aolsenjazz](https://github.com/aolsenjazz)! - feat(aws): bedrock prompt caching middleware

  Adds bedrockPromptCachingMiddleware. The interface largely matches what was previous implemented with anthropicPromptCachingMiddleware, making a best-effort attempt at utilizing prompt caching features for supported models.

## 1.4.5

### Patch Changes

- [#11026](https://github.com/langchain-ai/langchainjs/pull/11026) [`cbf274c`](https://github.com/langchain-ai/langchainjs/commit/cbf274c21d25f3c131aaceaa37c35a390d3ea9c1) Thanks [@aolsenjazz](https://github.com/aolsenjazz)! - feat(langchain): add `providerToolSearchMiddleware`

  Adds `providerToolSearchMiddleware` - provider-side tool search for agents. `providerToolSearchMiddleware` enables API consumers to opt tools into tool deferral + discovery by providing tool instances/names to the middleware's `searchableTools` arg. `searchableTools` are marked as `defer_loading` in subsequent model requests, consumed by OpenAI and Anthopic to power their tool search systems.

## 1.4.4

### Patch Changes

- [#10945](https://github.com/langchain-ai/langchainjs/pull/10945) [`bb30838`](https://github.com/langchain-ai/langchainjs/commit/bb30838d884000d168dbbb26ddf7cfb9fa63b437) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): propagate ReactAgent withConfig defaults to inner graph

  Apply static defaults from `#defaultConfig` onto the compiled pregel so
  `recursionLimit`, metadata, and other LangGraph config survives LangGraph API
  loading, which unwraps ReactAgent to `.graph` before execution.

- [#10939](https://github.com/langchain-ai/langchainjs/pull/10939) [`58f4c1f`](https://github.com/langchain-ai/langchainjs/commit/58f4c1fee51424c538402a695b91e9720c718320) Thanks [@oritwoen](https://github.com/oritwoen)! - fix(langchain): support agent invoke messages when strictNullChecks is disabled

## 1.4.3

### Patch Changes

- [#10936](https://github.com/langchain-ai/langchainjs/pull/10936) [`1a4ac84`](https://github.com/langchain-ai/langchainjs/commit/1a4ac8451748cfc0d014ff1bac8c641dad4d467b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): register stream transformers on middleware

  `createMiddleware` accepts `streamTransformers` factories that are merged with
  `createAgent({ streamTransformers })` at compile time. Types flow through
  `CombineStreamTransformers` so `run.extensions` is inferred from both sources.

## 1.4.2

### Patch Changes

- [#10900](https://github.com/langchain-ai/langchainjs/pull/10900) [`6bbddca`](https://github.com/langchain-ai/langchainjs/commit/6bbddca1f8f7f280583fd839bb40c00bf19a0d51) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): unwrap tool message outputs in agent streams

- [#10706](https://github.com/langchain-ai/langchainjs/pull/10706) [`4ecb660`](https://github.com/langchain-ai/langchainjs/commit/4ecb6606feae3156a07de67e39e2027f857c476e) Thanks [@JadenKim-dev](https://github.com/JadenKim-dev)! - fix(langchain): set name on todoListMiddleware ToolMessages

## 1.4.1

### Patch Changes

- [#10879](https://github.com/langchain-ai/langchainjs/pull/10879) [`eb480cb`](https://github.com/langchain-ai/langchainjs/commit/eb480cb6df8e0fa792826155bfa00a6db4536444) Thanks [@vignesh-gep](https://github.com/vignesh-gep)! - fix(langchain/createAgent): throw on terminal `providerStrategy` parse failure instead of silently resolving with `structuredResponse: undefined`

  When `createAgent` was configured with `responseFormat` resolving to a `providerStrategy` (either passed explicitly or auto-promoted from a bare Zod / JSON schema for models whose profile reports `structuredOutput: true`), and the model produced a terminal response (no `tool_calls`) whose text could not be JSON-parsed or did not satisfy the schema, the agent silently exited with no `structuredResponse`, surfacing later as `TypeError: Cannot read properties of undefined`. The agent now throws a `StructuredOutputParsingError` in that case while still allowing the agent loop to continue when tool calls are present. Closes [#10878](https://github.com/langchain-ai/langchainjs/issues/10878).

- [#10872](https://github.com/langchain-ai/langchainjs/pull/10872) [`a640079`](https://github.com/langchain-ai/langchainjs/commit/a64007997a4940f51bba3c1c83dae89d1ccfb692) Thanks [@hntrl](https://github.com/hntrl)! - chore(deps): remove redundant @types/uuid declarations

  Remove `@types/uuid` from package manifests that rely on `@langchain/core/utils/uuid` or do not require uuid type stubs directly, and refresh the lockfile entries accordingly.

- [#10160](https://github.com/langchain-ai/langchainjs/pull/10160) [`bba900c`](https://github.com/langchain-ai/langchainjs/commit/bba900c7c8781c7efec856d5d3e539a93f14e797) Thanks [@JadenKim-dev](https://github.com/JadenKim-dev)! - fix(langchain): prevent llmToolSelectorMiddleware from leaking into message stream

## 1.4.0

### Minor Changes

- [#10815](https://github.com/langchain-ai/langchainjs/pull/10815) [`499ce78`](https://github.com/langchain-ai/langchainjs/commit/499ce78a65cddd85b04a38d40f44ac8b65bd085d) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): adopt new streaming primitives

## 1.3.5

### Patch Changes

- [#10776](https://github.com/langchain-ai/langchainjs/pull/10776) [`20a9abe`](https://github.com/langchain-ai/langchainjs/commit/20a9abea23ffacf4ae8dc9a7aeec217143bbdeb6) Thanks [@hntrl](https://github.com/hntrl)! - fix(deps): remediate uuid vulnerability by removing direct uuid usage

- Updated dependencies [[`20a9abe`](https://github.com/langchain-ai/langchainjs/commit/20a9abea23ffacf4ae8dc9a7aeec217143bbdeb6)]:
  - @langchain/core@1.1.42

## 1.3.4

### Patch Changes

- [#10713](https://github.com/langchain-ai/langchainjs/pull/10713) [`49ac9e7`](https://github.com/langchain-ai/langchainjs/commit/49ac9e7404e5a9269b9ac047711ee96dd928b231) Thanks [@jacoblee93](https://github.com/jacoblee93)! - feat(langchain): Adds ls_agent_type to create agent runs as tracing metadata

- Updated dependencies [[`589f29c`](https://github.com/langchain-ai/langchainjs/commit/589f29ce844eb252c2d5e6b0f8d26de37763a0d7), [`2e9e696`](https://github.com/langchain-ai/langchainjs/commit/2e9e6969e248a53ede0659a41d0ac8dbaf291ab4)]:
  - @langchain/core@1.1.41

## 1.3.3

### Patch Changes

- [#9386](https://github.com/langchain-ai/langchainjs/pull/9386) [`71e53f1`](https://github.com/langchain-ai/langchainjs/commit/71e53f1c07bc60e2c2304b10f0edd3c85c62b192) Thanks [@Josh-Engle](https://github.com/Josh-Engle)! - Prevent local file corruption when using `LocalFileStore`

- Updated dependencies [[`d3e0809`](https://github.com/langchain-ai/langchainjs/commit/d3e080995bb267bf3797067ab53c96bc2a6c8e3f)]:
  - @langchain/core@1.1.40

## 1.3.2

### Patch Changes

- [#10693](https://github.com/langchain-ai/langchainjs/pull/10693) [`5a6e0ab`](https://github.com/langchain-ai/langchainjs/commit/5a6e0ab6617587f3aed19d07bf3ed91994dcdac8) Thanks [@hntrl](https://github.com/hntrl)! - fix(agents): derive middleware hook state from invocation state

  Prevents middleware state from leaking across threads by deriving middleware hook input state from the current invocation state instead of cross-node cached state.

## 1.3.1

### Patch Changes

- [#10673](https://github.com/langchain-ai/langchainjs/pull/10673) [`f069365`](https://github.com/langchain-ai/langchainjs/commit/f0693657b7e5bdbf888a46d04e68431e446c7bc0) Thanks [@hntrl](https://github.com/hntrl)! - bump langgraph dependency to 1.2.8

## 1.3.0

### Minor Changes

- [#10430](https://github.com/langchain-ai/langchainjs/pull/10430) [`d3d0922`](https://github.com/langchain-ai/langchainjs/commit/d3d0922c24afcd3006fb94dcadd3ebe08fbf2383) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): support for browser tools

### Patch Changes

- [#10591](https://github.com/langchain-ai/langchainjs/pull/10591) [`d7a98cd`](https://github.com/langchain-ai/langchainjs/commit/d7a98cda1a5d9bf9b93b503fc54374f1aaf1a37e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat: add @langchain/perplexity standalone provider package

- [#10594](https://github.com/langchain-ai/langchainjs/pull/10594) [`884c2d3`](https://github.com/langchain-ai/langchainjs/commit/884c2d3d1b2c49225d73ddec2235ad174db36f86) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(fireworks): extract standalone provider package

- [#10593](https://github.com/langchain-ai/langchainjs/pull/10593) [`0fb6fa4`](https://github.com/langchain-ai/langchainjs/commit/0fb6fa40dcd3a09a4fb91f36c9f2ca869552961e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(together-ai): migrate Together AI into provider package

- [#10654](https://github.com/langchain-ai/langchainjs/pull/10654) [`3bd85c1`](https://github.com/langchain-ai/langchainjs/commit/3bd85c17869e95d5d1a67c5fccc0c4cab2646616) Thanks [@jabiinfante](https://github.com/jabiinfante)! - feat(langchain): add ChatGoogle support to initChatModel

- Updated dependencies [[`d3d0922`](https://github.com/langchain-ai/langchainjs/commit/d3d0922c24afcd3006fb94dcadd3ebe08fbf2383)]:
  - @langchain/core@1.1.39

## 1.2.39

### Patch Changes

- [#10543](https://github.com/langchain-ai/langchainjs/pull/10543) [`7ed93b8`](https://github.com/langchain-ai/langchainjs/commit/7ed93b8b8a6e9eb0da3e103d74087c692fee2773) Thanks [@pawel-twardziak](https://github.com/pawel-twardziak)! - fix(langchain): allow dynamic tools in wrapModelCall with wrapToolCall

- [#10554](https://github.com/langchain-ai/langchainjs/pull/10554) [`11a295f`](https://github.com/langchain-ai/langchainjs/commit/11a295fdadec3809f40c10492e3fd474e832c468) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): add support for dynamic structured output

- [#10555](https://github.com/langchain-ai/langchainjs/pull/10555) [`f548053`](https://github.com/langchain-ai/langchainjs/commit/f54805305787fa383c3ce1e287daafdb5464a98b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): bump langgraph dep

- [#10552](https://github.com/langchain-ai/langchainjs/pull/10552) [`589ab9b`](https://github.com/langchain-ai/langchainjs/commit/589ab9be391a5d6c104f34877fc1b3e2a32fa449) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): accept cross-version runnable models in createAgent

- Updated dependencies [[`589ab9b`](https://github.com/langchain-ai/langchainjs/commit/589ab9be391a5d6c104f34877fc1b3e2a32fa449)]:
  - @langchain/core@1.1.38

## 1.2.38

### Patch Changes

- [#10527](https://github.com/langchain-ai/langchainjs/pull/10527) [`3408008`](https://github.com/langchain-ai/langchainjs/commit/3408008fe59eb8ee72f733349ae85afe5c23eaca) Thanks [@pawel-twardziak](https://github.com/pawel-twardziak)! - fix(langchain): export createAgent and prebuilt middleware from browser entry point

- [#10545](https://github.com/langchain-ai/langchainjs/pull/10545) [`68e0a19`](https://github.com/langchain-ai/langchainjs/commit/68e0a19238be592514b6c01243c41aefaa6a7668) Thanks [@JadenKim-dev](https://github.com/JadenKim-dev)! - fix(langchain): revert zod import in utils.ts to fix v3/v4 interop

- Updated dependencies [[`6933769`](https://github.com/langchain-ai/langchainjs/commit/6933769836fe3cec835588e5f8db9883200865f6), [`50d5f32`](https://github.com/langchain-ai/langchainjs/commit/50d5f32fd30cabebf058b1c13255c1daadde6107), [`5552999`](https://github.com/langchain-ai/langchainjs/commit/555299917c90322e25d7671bad2e20c9b104bad6), [`8331833`](https://github.com/langchain-ai/langchainjs/commit/8331833c93ba907063c9fe28e9f935ed5dfec11c)]:
  - @langchain/core@1.1.37

## 1.2.37

### Patch Changes

- [#10475](https://github.com/langchain-ai/langchainjs/pull/10475) [`3d35eb1`](https://github.com/langchain-ai/langchainjs/commit/3d35eb112b46b36aea80cebe0147e315d03a1d8e) Thanks [@hntrl](https://github.com/hntrl)! - fix(langchain): add "aws" alias to MODEL_PROVIDER_CONFIG so hub/node auto-detects ChatBedrockConverse from Python-serialized prompts

- [#10258](https://github.com/langchain-ai/langchainjs/pull/10258) [`ae4122f`](https://github.com/langchain-ai/langchainjs/commit/ae4122f9523904423504e5a92c134c625977261d) Thanks [@irfiacre](https://github.com/irfiacre)! - Align Zod Importation For 'libs/langchain/src/agents/nodes/utils.ts'

- Updated dependencies [[`bbbfea1`](https://github.com/langchain-ai/langchainjs/commit/bbbfea185c0777ae06df2b24a1a84f941d499c2a)]:
  - @langchain/core@1.1.36

## 1.2.36

### Patch Changes

- [#10489](https://github.com/langchain-ai/langchainjs/pull/10489) [`21094f3`](https://github.com/langchain-ai/langchainjs/commit/21094f31d99e30a9c8433072523cb08b8539c1ab) Thanks [@maahir30](https://github.com/maahir30)! - support structured output (providerStrategy) for Google Gemini models in createAgent

- [#10433](https://github.com/langchain-ai/langchainjs/pull/10433) [`7af0b65`](https://github.com/langchain-ai/langchainjs/commit/7af0b65d5ab9a173b528d6a821d269a79fbabdc6) Thanks [@tanushree-sharma](https://github.com/tanushree-sharma)! - feat: Add LangSmith integration metadata to createAgent and initChatModel

- Updated dependencies [[`5dc11b5`](https://github.com/langchain-ai/langchainjs/commit/5dc11b55cccfe35e4dad910a33e904cf49b3088a), [`7af0b65`](https://github.com/langchain-ai/langchainjs/commit/7af0b65d5ab9a173b528d6a821d269a79fbabdc6)]:
  - @langchain/core@1.1.35

## 1.2.35

### Patch Changes

- [#10466](https://github.com/langchain-ai/langchainjs/pull/10466) [`4b1a874`](https://github.com/langchain-ai/langchainjs/commit/4b1a874a93d8475a98421e157fe7a1e65ec94076) Thanks [@hntrl](https://github.com/hntrl)! - fix: renamed getSubgraphAsync -> getSubgraphsAsync

- Updated dependencies [[`bfb7944`](https://github.com/langchain-ai/langchainjs/commit/bfb7944a105470eee98fe4a0eef91e586600e1de)]:
  - @langchain/core@1.1.34

## 1.2.34

### Patch Changes

- [#10443](https://github.com/langchain-ai/langchainjs/pull/10443) [`ff6822e`](https://github.com/langchain-ai/langchainjs/commit/ff6822e43c38c4328d3f2c0ef1cf67998741339a) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): respect version:"v1" in afterModel router's pending tool call path

- [#10446](https://github.com/langchain-ai/langchainjs/pull/10446) [`888224c`](https://github.com/langchain-ai/langchainjs/commit/888224c64e4c95935836a5089f8b28a8c90da5e6) Thanks [@hntrl](https://github.com/hntrl)! - fix(agents): propagate store and configurable to ToolNode middleware runtime

- [#10444](https://github.com/langchain-ai/langchainjs/pull/10444) [`82d56cb`](https://github.com/langchain-ai/langchainjs/commit/82d56cbdfb59b48b3f8d98a294e4d4720e12c733) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain/agents): dispatch tool calls via Send in afterModel router for version:"v2"

  **Breaking change for `version: "v2"` + `afterModel` middleware users.**

  Previously, when `afterModel` middleware was present, `createAgent` always routed all tool calls from an `AIMessage` to a single `ToolNode` invocation — regardless of the `version` option. This meant `version: "v2"` silently behaved like `version: "v1"` (parallel via `Promise.all` in one node) whenever `afterModel` middleware was used.

  `#createAfterModelRouter` now correctly respects `#toolBehaviorVersion`:
  - `version: "v1"` — routes the full `AIMessage` to a single `ToolNode` invocation; all tool calls run concurrently via `Promise.all` (unchanged behaviour).
  - `version: "v2"` — dispatches each tool call as a separate `Send` task, matching the behaviour of `#createModelRouter` when no `afterModel` middleware is present, and matching Python LangGraph's `post_model_hook_router`.

  **Migration:** If you use `version: "v2"` (the default) together with `afterModel` middleware and rely on the previous single-node parallel execution, switch to `version: "v1"` to preserve that behaviour. See the `version` JSDoc on `CreateAgentParams` for guidance on which option to choose.

## 1.2.33

### Patch Changes

- [#9862](https://github.com/langchain-ai/langchainjs/pull/9862) [`5b250b1`](https://github.com/langchain-ai/langchainjs/commit/5b250b1a273df0ab4126b22d140dc2420130421f) Thanks [@stellarrover](https://github.com/stellarrover)! - fix(langchain/agents): Fix ReactAgent routing with returnDirect + beforeModel middleware

  This fixes a routing error when an agent has both tools with returnDirect: true and middleware with beforeModel hooks (e.g., summarizationMiddleware). Before this fix, non-returnDirect tools would fail with "Branch condition returned unknown or null destination".

- Updated dependencies [[`6db417b`](https://github.com/langchain-ai/langchainjs/commit/6db417b03ecb5e2ace413389d982294e0ac88433), [`d69dfcc`](https://github.com/langchain-ai/langchainjs/commit/d69dfcca97503cf1c0b7e70ccf5fb7d507c60982)]:
  - @langchain/core@1.1.33

## 1.2.32

### Patch Changes

- [#10405](https://github.com/langchain-ai/langchainjs/pull/10405) [`afbf5f1`](https://github.com/langchain-ai/langchainjs/commit/afbf5f1b1a9c6014f994829ab92bbfcaecb8a30b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): add browser export

## 1.2.31

### Patch Changes

- [#10265](https://github.com/langchain-ai/langchainjs/pull/10265) [`2647b48`](https://github.com/langchain-ai/langchainjs/commit/2647b4841bc4ae28dc6c8b245532df8620611a74) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): export todo schema

- [#10167](https://github.com/langchain-ai/langchainjs/pull/10167) [`ca826f6`](https://github.com/langchain-ai/langchainjs/commit/ca826f6fecae6087bf0dee7781ee80b587396ec1) Thanks [@colifran](https://github.com/colifran)! - feat: implement type inference for tool streams

- [#10290](https://github.com/langchain-ai/langchainjs/pull/10290) [`a596d3f`](https://github.com/langchain-ai/langchainjs/commit/a596d3f7395c0ab27357aa0cd30bafb2d5d967c1) Thanks [@colifran](https://github.com/colifran)! - fix(langchain): serializable schemas aren't exposed on create agent response format

- Updated dependencies [[`26488b5`](https://github.com/langchain-ai/langchainjs/commit/26488b596f01b7b7fe2f1d97d07164e52365ade5), [`ca826f6`](https://github.com/langchain-ai/langchainjs/commit/ca826f6fecae6087bf0dee7781ee80b587396ec1), [`a602c42`](https://github.com/langchain-ai/langchainjs/commit/a602c42db75d7e7e01cab38b12e0b65b9c0cce95), [`db7d017`](https://github.com/langchain-ai/langchainjs/commit/db7d017f7ce13cb937147aabcbfa3847d80bde9d)]:
  - @langchain/core@1.1.32

## 1.2.30

### Patch Changes

- [#10262](https://github.com/langchain-ai/langchainjs/pull/10262) [`b0175a5`](https://github.com/langchain-ai/langchainjs/commit/b0175a5d3b68e8fba44a85bc23879bd06def2f52) Thanks [@maahir30](https://github.com/maahir30)! - fix: Move fakeModel from utils/testing to testing namespace
  move to updated namespace

- [#10185](https://github.com/langchain-ai/langchainjs/pull/10185) [`414f6ed`](https://github.com/langchain-ai/langchainjs/commit/414f6ed402ac6f1c0fd6cce4bed64fa3708eea3d) Thanks [@maahir30](https://github.com/maahir30)! - feat: add custom Vitest matchers for LangChain message and tool call assertions

  Adds a new `@langchain/core/testing/matchers` export containing custom Vitest matchers (`toBeHumanMessage`, `toBeAIMessage`, `toBeSystemMessage`, `toBeToolMessage`, `toHaveToolCalls`, `toHaveToolCallCount`, `toContainToolCall`, `toHaveToolMessages`, `toHaveBeenInterrupted`, `toHaveStructuredResponse`) that external users can register via `expect.extend(langchainMatchers)` in their Vitest setup files. Re-exported from `langchain` for convenience.

- Updated dependencies [[`7373b4c`](https://github.com/langchain-ai/langchainjs/commit/7373b4cd6a78bee105a952a11838c573fd1aafae), [`b0175a5`](https://github.com/langchain-ai/langchainjs/commit/b0175a5d3b68e8fba44a85bc23879bd06def2f52), [`414f6ed`](https://github.com/langchain-ai/langchainjs/commit/414f6ed402ac6f1c0fd6cce4bed64fa3708eea3d)]:
  - @langchain/core@1.1.31

## 1.2.29

### Patch Changes

- [#10245](https://github.com/langchain-ai/langchainjs/pull/10245) [`48dfa1d`](https://github.com/langchain-ai/langchainjs/commit/48dfa1dd08c0c2c801356a8fd0d17f08c85f967e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(agents): propagate structured output retry Command through wrapModelCall middleware

- [#10243](https://github.com/langchain-ai/langchainjs/pull/10243) [`96c630d`](https://github.com/langchain-ai/langchainjs/commit/96c630dfd009f2546d5bc36f5067ff868bb4067f) Thanks [@hntrl](https://github.com/hntrl)! - fix: add explicit `: symbol` type annotations to Symbol.for() declarations for cross-version compatibility

  TypeScript infers `unique symbol` type when Symbol.for() is used without an explicit type annotation, causing type incompatibility when multiple versions of the same package are present in a dependency tree. By adding explicit `: symbol` annotations, all declarations now use the general symbol type, making them compatible across versions while maintaining identical runtime behavior.

  Changes:
  - Added `: symbol` to `MESSAGE_SYMBOL` in messages/base.ts
  - Added `: symbol` to `MIDDLEWARE_BRAND` in agents/middleware/types.ts (also changed from Symbol() to Symbol.for() for cross-realm compatibility)

- [#10252](https://github.com/langchain-ai/langchainjs/pull/10252) [`0bf01a2`](https://github.com/langchain-ai/langchainjs/commit/0bf01a29bca124f76b08ee52a7795b2eafbc5272) Thanks [@colifran](https://github.com/colifran)! - feat: implement standard schema support for structured output

- Updated dependencies [[`96c630d`](https://github.com/langchain-ai/langchainjs/commit/96c630dfd009f2546d5bc36f5067ff868bb4067f), [`a8b9ccc`](https://github.com/langchain-ai/langchainjs/commit/a8b9ccca5a85984a5a30008acd09f9991e591638), [`a1f22bb`](https://github.com/langchain-ai/langchainjs/commit/a1f22bba907731a18dca23c31cec5333444a3f55)]:
  - @langchain/core@1.1.30

## 1.2.28

### Patch Changes

- [#10169](https://github.com/langchain-ai/langchainjs/pull/10169) [`3b1fd54`](https://github.com/langchain-ai/langchainjs/commit/3b1fd5458a4aa29c398122829f383f21b5ac39da) Thanks [@hntrl](https://github.com/hntrl)! - fix(core, langchain): bump uuid dependency from ^10.0.0 to ^11.0.0 to fix Metro bundler error

  The `uuid` v10 package has ambiguous `exports` in its `package.json` which causes Metro (used by Expo/React Native) to resolve the wrong entry point, resulting in `Cannot read properties of undefined (reading 'v1')`. The `uuid` v11 package fixes its exports map to work correctly with Metro's package exports resolution.

- [#10165](https://github.com/langchain-ai/langchainjs/pull/10165) [`01a84ae`](https://github.com/langchain-ai/langchainjs/commit/01a84ae4b73ee2e3add940250aca2236213eaab2) Thanks [@hntrl](https://github.com/hntrl)! - fix(agents): propagate store and checkpointer in ReactAgent
  - Added `checkpointer` and `store` getter/setter pairs on `ReactAgent` that forward to the internal compiled graph. This fixes an issue where the LangGraph API server's checkpointer injection was silently ignored, causing thread state to be lost across server restarts.
  - Propagate `store` and `configurable` from the LangGraph config into the middleware `runtime` object. Previously, `runtime.store` was always `undefined` even when a store was provided to `createAgent()`.

- [#10146](https://github.com/langchain-ai/langchainjs/pull/10146) [`67cc069`](https://github.com/langchain-ai/langchainjs/commit/67cc069af19844d8c2676f6ae36117af7c7740d4) Thanks [@eddienubes](https://github.com/eddienubes)! - docs: update an outdated middleware example

- Updated dependencies [[`9f30267`](https://github.com/langchain-ai/langchainjs/commit/9f30267e95a2a42fac71f1d3674b84c5a190dbbc), [`403a99f`](https://github.com/langchain-ai/langchainjs/commit/403a99fd826383f30300809ae077e1c967023520), [`3b1fd54`](https://github.com/langchain-ai/langchainjs/commit/3b1fd5458a4aa29c398122829f383f21b5ac39da), [`77bd982`](https://github.com/langchain-ai/langchainjs/commit/77bd98274a885e947d76f7a9c6dd0b3763453218)]:
  - @langchain/core@1.1.29

## 1.2.27

### Patch Changes

- [#10137](https://github.com/langchain-ai/langchainjs/pull/10137) [`87a585e`](https://github.com/langchain-ai/langchainjs/commit/87a585e4b5d36fa2f82e34a92caa6e7c4289f370) Thanks [@hntrl](https://github.com/hntrl)! - fix(agents): infer StateSchema value types in middleware hooks

- Updated dependencies [[`10a876c`](https://github.com/langchain-ai/langchainjs/commit/10a876c7d5ff27d8f2889761ee20e95f76a50518), [`b46d96a`](https://github.com/langchain-ai/langchainjs/commit/b46d96a508a8bf212561dbb6f025e35c75f16257)]:
  - @langchain/core@1.1.28

## 1.2.26

### Patch Changes

- [#10108](https://github.com/langchain-ai/langchainjs/pull/10108) [`e7576ee`](https://github.com/langchain-ai/langchainjs/commit/e7576ee9e6408c399c08d271db43f63e622da10f) Thanks [@hntrl](https://github.com/hntrl)! - fix: replace retired Anthropic model IDs with active replacements
  - Update default model in ChatAnthropic from `claude-3-5-sonnet-latest` to `claude-sonnet-4-5-20250929`
  - Regenerate model profiles with latest data from models.dev API
  - Replace retired `claude-3-5-haiku-20241022`, `claude-3-7-sonnet-20250219`, `claude-3-5-sonnet-20240620`, and `claude-3-5-sonnet-20241022` in tests, docstrings, and examples

- [#10114](https://github.com/langchain-ai/langchainjs/pull/10114) [`0050c91`](https://github.com/langchain-ai/langchainjs/commit/0050c91481267327d88c430f5aacf57bd336facf) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): reset shared currentSystemMessage on middleware handler retry

## 1.2.25

### Patch Changes

- [#10083](https://github.com/langchain-ai/langchainjs/pull/10083) [`219b38d`](https://github.com/langchain-ai/langchainjs/commit/219b38d6d95977f3f9bf19625e72f42036d24df0) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): export ModelRequest interface

- [#10084](https://github.com/langchain-ai/langchainjs/pull/10084) [`7cae8a9`](https://github.com/langchain-ai/langchainjs/commit/7cae8a9ee6aece5f2df0fc6417ffd799bce4d1ca) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): persist underscore-prefixed middleware state fields across graph steps

- Updated dependencies [[`ed6ea53`](https://github.com/langchain-ai/langchainjs/commit/ed6ea53c38a004b65e30c0f5888a0ac7d8ee7028)]:
  - @langchain/core@1.1.26

## 1.2.24

### Patch Changes

- [#10034](https://github.com/langchain-ai/langchainjs/pull/10034) [`8eea9d4`](https://github.com/langchain-ai/langchainjs/commit/8eea9d462d2e782e62455973d6b0907aa917b41e) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): fix streaming types

## 1.2.23

### Patch Changes

- [#10010](https://github.com/langchain-ai/langchainjs/pull/10010) [`bf6eac4`](https://github.com/langchain-ai/langchainjs/commit/bf6eac40179f87da18fd122b4efc2ee48cd44b6c) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): fix structured output on opus

- Updated dependencies [[`70a4400`](https://github.com/langchain-ai/langchainjs/commit/70a440085b4bc2d036726ed12d9dc7841e914061)]:
  - @langchain/core@1.1.24

## 1.2.22

### Patch Changes

- [#10008](https://github.com/langchain-ai/langchainjs/pull/10008) [`5283513`](https://github.com/langchain-ai/langchainjs/commit/5283513740ff93063f9c1eaa583365f590ec5994) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): re-export missing interface

- Updated dependencies [[`71d08c0`](https://github.com/langchain-ai/langchainjs/commit/71d08c0a3a2597bd5a084eb35a7830e5ea1a2b29)]:
  - @langchain/core@1.1.23

## 1.2.21

### Patch Changes

- [#9996](https://github.com/langchain-ai/langchainjs/pull/9996) [`5113204`](https://github.com/langchain-ai/langchainjs/commit/5113204bba5eb1d0b23a4d3aae2c02ba3caf8b6b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): propagate subagent name in metadata

- Updated dependencies [[`8f166b1`](https://github.com/langchain-ai/langchainjs/commit/8f166b159343ae6fd0d6d44c0835ab56c0b153f4)]:
  - @langchain/core@1.1.22

## 1.2.20

### Patch Changes

- [#9983](https://github.com/langchain-ai/langchainjs/pull/9983) [`ea7e791`](https://github.com/langchain-ai/langchainjs/commit/ea7e791ff6c93188b7febbd98ea22359fa96eb20) Thanks [@encodedz](https://github.com/encodedz)! - Support concurrent `agentNode` dynamic prompts properly.

- [#9986](https://github.com/langchain-ai/langchainjs/pull/9986) [`8a52dad`](https://github.com/langchain-ai/langchainjs/commit/8a52dad1eace191b19d24cc5683acb22a5b9e964) Thanks [@brettshollenberger](https://github.com/brettshollenberger)! - Don't re-emit input state from middleware hooks that return void

- [#9982](https://github.com/langchain-ai/langchainjs/pull/9982) [`4558fa6`](https://github.com/langchain-ai/langchainjs/commit/4558fa6438749b9f6eca980099241d97c56b1a8f) Thanks [@hntrl](https://github.com/hntrl)! - feat(langchain): add optional `tools` parameter to `countTokensApproximately` for more accurate token estimates when tools are bound to a model

- [#9981](https://github.com/langchain-ai/langchainjs/pull/9981) [`5f28338`](https://github.com/langchain-ai/langchainjs/commit/5f283389c0dba1822f37139f0913016d77167734) Thanks [@hntrl](https://github.com/hntrl)! - feat(agents): support returning Command from wrapModelCall middleware

  Allow `wrapModelCall` middleware hooks to return `Command` objects for advanced
  control flow (routing, state updates), matching the existing `wrapToolCall`
  pattern. The framework tracks the effective AIMessage through the middleware
  chain so outer middleware always receive an AIMessage from `handler()`, even
  when an inner middleware returns a Command.

- Updated dependencies [[`d5e3db0`](https://github.com/langchain-ai/langchainjs/commit/d5e3db0d01ab321ec70a875805b2f74aefdadf9d)]:
  - @langchain/core@1.1.21

## 1.2.19

### Patch Changes

- [#9957](https://github.com/langchain-ai/langchainjs/pull/9957) [`71c3cba`](https://github.com/langchain-ai/langchainjs/commit/71c3cba843ab16d877299d158a1de0c7d22f3fb9) Thanks [@jacoblee93](https://github.com/jacoblee93)! - feat(langchain,core): Update prompt pulling params, LangSmith version

- Updated dependencies [[`71c3cba`](https://github.com/langchain-ai/langchainjs/commit/71c3cba843ab16d877299d158a1de0c7d22f3fb9)]:
  - @langchain/core@1.1.20

## 1.2.18

### Patch Changes

- [#9763](https://github.com/langchain-ai/langchainjs/pull/9763) [`8f0757f`](https://github.com/langchain-ai/langchainjs/commit/8f0757f06b2ed9fe810f636333fc71ffcedb3feb) Thanks [@AdamParker19](https://github.com/AdamParker19)! - fix(langchain): resolve className collision in MODEL_PROVIDER_CONFIG

  Refactored `getChatModelByClassName` to accept an optional `modelProvider` parameter for direct lookup, avoiding the className collision issue where multiple providers share the same className (e.g., `google-vertexai` and `google-vertexai-web` both use `"ChatVertexAI"`). When `modelProvider` is provided, the function uses direct config lookup instead of searching by className. Backward compatibility is maintained for existing callers that only pass `className`. This eliminates the duplicated import logic that was previously in `_initChatModelHelper`.

## 1.2.17

### Patch Changes

- [#9916](https://github.com/langchain-ai/langchainjs/pull/9916) [`3516592`](https://github.com/langchain-ai/langchainjs/commit/3516592c51c44eb55e477c0d63ffdab5672ae97e) Thanks [@hntrl](https://github.com/hntrl)! - feat(langchain): add withConfig() method to ReactAgent

  Adds a `withConfig()` method to ReactAgent following the same pattern as LangGraph's `Pregel.withConfig()`. This allows setting default configuration values (like `recursionLimit`, `tags`, or `configurable`) that get merged with invocation-time config.

- Updated dependencies [[`41bfea5`](https://github.com/langchain-ai/langchainjs/commit/41bfea51cf119573a3b956ee782d2731fe71c681)]:
  - @langchain/core@1.1.19

## 1.2.16

### Patch Changes

- [#9885](https://github.com/langchain-ai/langchainjs/pull/9885) [`913893f`](https://github.com/langchain-ai/langchainjs/commit/913893f678f1fae2a1b9bf964c9d93ffd0176680) Thanks [@Ambas-T](https://github.com/Ambas-T)! - fix: filter pregel keys from cache key to avoid OOM

- Updated dependencies [[`a9b5059`](https://github.com/langchain-ai/langchainjs/commit/a9b50597186002221aaa4585246e569fa44c27c8), [`a9b5059`](https://github.com/langchain-ai/langchainjs/commit/a9b50597186002221aaa4585246e569fa44c27c8)]:
  - @langchain/core@1.1.18

## 1.2.15

### Patch Changes

- [#9891](https://github.com/langchain-ai/langchainjs/pull/9891) [`983b212`](https://github.com/langchain-ai/langchainjs/commit/983b2123eeaa32011b7021643ba8c37698dfe2be) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): support StateSchema in middleware pipeline

## 1.2.14

### Patch Changes

- [#9870](https://github.com/langchain-ai/langchainjs/pull/9870) [`070b4d1`](https://github.com/langchain-ai/langchainjs/commit/070b4d14121ee5d1f24c8ad980cd4548f39f19eb) Thanks [@maahir30](https://github.com/maahir30)! - fix(langchain): StateSchema handling in AgentNode middleware
  - Added `toPartialZodObject` helper that correctly handles both Zod objects and LangGraph's StateSchema when parsing middleware state in AgentNode .

## 1.2.13

### Patch Changes

- [#9854](https://github.com/langchain-ai/langchainjs/pull/9854) [`160b5bf`](https://github.com/langchain-ai/langchainjs/commit/160b5bfe49f31190d28ec10a95075ef845c49fa3) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(anthropic): apply cache_control at final formatting layer

- [#9850](https://github.com/langchain-ai/langchainjs/pull/9850) [`b56bf9e`](https://github.com/langchain-ai/langchainjs/commit/b56bf9e05a3190be58835e336ffb775c5d585fca) Thanks [@joeljohn159](https://github.com/joeljohn159)! - Fix `ToolNode` to route missing tool errors through `handleToolErrors` instead of throwing, allowing agents to recover when LLMs hallucinate tool names.

- [#9847](https://github.com/langchain-ai/langchainjs/pull/9847) [`9cd35ae`](https://github.com/langchain-ai/langchainjs/commit/9cd35ae6ed878be88d3e68371df92cebc8d880ee) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): add dynamic tool registration via middleware

- [#9839](https://github.com/langchain-ai/langchainjs/pull/9839) [`134121a`](https://github.com/langchain-ai/langchainjs/commit/134121a6fc7e0ad1653bb5ce2ec347c61638c4df) Thanks [@maahir30](https://github.com/maahir30)! - ## Add state schema support to agents
  - Introduce `createAgentState` function replacing `createAgentAnnotationConditional`
  - Migrate from Zod-based schemas to LangGraph's native `StateSchema`, `ReducedValue`, and `UntrackedValue` primitives
  - Support both `StateSchema` and Zod v3/v4 objects as input schemas
  - Automatically merge state fields from user-provided schemas and middleware
  - Properly handle reducer metadata extraction from Zod v4 schemas via `schemaMetaRegistry`
  - Generate separate input/output schemas to avoid channel conflicts with reducers
  - Add comprehensive test coverage for state schema functionality

- Updated dependencies [[`05a9733`](https://github.com/langchain-ai/langchainjs/commit/05a9733448a10764c0bfd070af859c33e623b998)]:
  - @langchain/core@1.1.17

## 1.2.12

### Patch Changes

- [#9848](https://github.com/langchain-ai/langchainjs/pull/9848) [`3d26f58`](https://github.com/langchain-ai/langchainjs/commit/3d26f58e5b705f7c81104a6f1145c64e791d238b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): bubble up graph interrupt errors as is

## 1.2.11

### Patch Changes

- [#9812](https://github.com/langchain-ai/langchainjs/pull/9812) [`b8e3da6`](https://github.com/langchain-ai/langchainjs/commit/b8e3da6618ca27e86d384a3ce6d01520861916af) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): tag summarization model calls with lc_source metadata

- Updated dependencies [[`70387a1`](https://github.com/langchain-ai/langchainjs/commit/70387a144464539d65a546c8130cf51dfad025a1), [`a7c6ec5`](https://github.com/langchain-ai/langchainjs/commit/a7c6ec51ab9baa186ab5ebf815599c08f5c7e8ab), [`5e04543`](https://github.com/langchain-ai/langchainjs/commit/5e045435a783fdae44bc9a43e01a8e5eb7100db2), [`40b4467`](https://github.com/langchain-ai/langchainjs/commit/40b446762445575844610ee528abc77c247b2c43), [`17e30bd`](https://github.com/langchain-ai/langchainjs/commit/17e30bd7f4c7bdf87c9c30304b3b9e121cc1fbbc)]:
  - @langchain/core@1.1.16

## 1.2.10

### Patch Changes

- [#9806](https://github.com/langchain-ai/langchainjs/pull/9806) [`cf56fc9`](https://github.com/langchain-ai/langchainjs/commit/cf56fc98aebf55a798e80cbd5c30259ffaf8ca47) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): don't concatenate middleware error, instead use cause

- Updated dependencies [[`230462d`](https://github.com/langchain-ai/langchainjs/commit/230462d28c3a8b5ccadf433ea2f523eb6e658de6)]:
  - @langchain/core@1.1.15

## 1.2.9

### Patch Changes

- Updated dependencies [[`bd1ab45`](https://github.com/langchain-ai/langchainjs/commit/bd1ab45364391f69ce93ecba36a4a15dafca2b76)]:
  - @langchain/core@1.1.14

## 1.2.8

### Patch Changes

- [#9779](https://github.com/langchain-ai/langchainjs/pull/9779) [`bb80067`](https://github.com/langchain-ai/langchainjs/commit/bb80067874af359d7347b07b75f5caa0ee246aea) Thanks [@hntrl](https://github.com/hntrl)! - add lc_source tag to summarization middleware messages

- [#9784](https://github.com/langchain-ai/langchainjs/pull/9784) [`9a68b93`](https://github.com/langchain-ai/langchainjs/commit/9a68b933c13be556376244e83b18f61f051e6a5c) Thanks [@weiliddat](https://github.com/weiliddat)! - preserve reducer metadata in createAgent state schema

- Updated dependencies [[`3efe79c`](https://github.com/langchain-ai/langchainjs/commit/3efe79c62ff2ffe0ada562f7eecd85be074b649a), [`b8561c1`](https://github.com/langchain-ai/langchainjs/commit/b8561c17556bdf7a3ff8d70bc307422642a9172e)]:
  - @langchain/core@1.1.13

## 1.2.7

### Patch Changes

- [#9517](https://github.com/langchain-ai/langchainjs/pull/9517) [`23be5af`](https://github.com/langchain-ai/langchainjs/commit/23be5afd59b5f4806edef11937ce5e2ba300f7ee) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): Refactored agent generics to use type bag pattern (AgentTypeConfig, MiddlewareTypeConfig) with new helper types for type extraction.

- [#9761](https://github.com/langchain-ai/langchainjs/pull/9761) [`e0360d9`](https://github.com/langchain-ai/langchainjs/commit/e0360d9bdc0e7725d59625902bcfc98c39931e2a) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): ensure models only make on write_todo call at a time

- Updated dependencies [[`23be5af`](https://github.com/langchain-ai/langchainjs/commit/23be5afd59b5f4806edef11937ce5e2ba300f7ee)]:
  - @langchain/core@1.1.12

## 1.2.6

### Patch Changes

- Updated dependencies [[`a46a249`](https://github.com/langchain-ai/langchainjs/commit/a46a24983fd0fea649d950725a2673b3c435275f)]:
  - @langchain/core@1.1.11

## 1.2.5

### Patch Changes

- Updated dependencies [[`817fc9a`](https://github.com/langchain-ai/langchainjs/commit/817fc9a56d4699f3563a6e153b13eadf7bcc661b)]:
  - @langchain/core@1.1.10

## 1.2.4

### Patch Changes

- [#9739](https://github.com/langchain-ai/langchainjs/pull/9739) [`c28d24a`](https://github.com/langchain-ai/langchainjs/commit/c28d24a8770f6d0e543cde116b0e38b3baf21301) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(core): use getBufferString for message summarization

- [#9729](https://github.com/langchain-ai/langchainjs/pull/9729) [`536c7dd`](https://github.com/langchain-ai/langchainjs/commit/536c7ddacd6c7f80d2edf18ab9caeeab71827ccd) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): default strict to true in providerStrategy for OpenAI compatibility

- [#9728](https://github.com/langchain-ai/langchainjs/pull/9728) [`5cfbedf`](https://github.com/langchain-ai/langchainjs/commit/5cfbedf064ffdc960cb2e5a97e37d7a5900560de) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): support callbacks property in stream

- [#9715](https://github.com/langchain-ai/langchainjs/pull/9715) [`cc502e1`](https://github.com/langchain-ai/langchainjs/commit/cc502e1b67dbadcd123a7ea2964c791c2bbad581) Thanks [@jacoblee93](https://github.com/jacoblee93)! - fix(langchain): Add secretsFromEnv and secrets for prompt pulling
  fix(@langchain/classic): Add secretsFromEnv and secrets for prompt pulling

- [#9703](https://github.com/langchain-ai/langchainjs/pull/9703) [`5d1b874`](https://github.com/langchain-ai/langchainjs/commit/5d1b874e1bb0a038f23576d5ff6d8335b755bce1) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): properly pass on schema title

- [#9702](https://github.com/langchain-ai/langchainjs/pull/9702) [`bfcb87d`](https://github.com/langchain-ai/langchainjs/commit/bfcb87d23c580c7881f650960a448fe2e54a30b3) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(core): improve interop with Zod

- [#9714](https://github.com/langchain-ai/langchainjs/pull/9714) [`42b138f`](https://github.com/langchain-ai/langchainjs/commit/42b138fe20a8d711938058a69669e1c29b18bd50) Thanks [@r77wu](https://github.com/r77wu)! - use parseJumpToTarget in 'afterModel' Router

- [#9740](https://github.com/langchain-ai/langchainjs/pull/9740) [`f697451`](https://github.com/langchain-ai/langchainjs/commit/f6974516b041ed12befd26e1a4cbe457865a2780) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): keep tool call / AIMessage pairings when summarizing

- Updated dependencies [[`56600b9`](https://github.com/langchain-ai/langchainjs/commit/56600b94f8e185f44d4288b7a9b66c55778938dd), [`dc5c2ac`](https://github.com/langchain-ai/langchainjs/commit/dc5c2ac00f86dd2feeba9843d708926a5f38202e), [`c28d24a`](https://github.com/langchain-ai/langchainjs/commit/c28d24a8770f6d0e543cde116b0e38b3baf21301), [`bfcb87d`](https://github.com/langchain-ai/langchainjs/commit/bfcb87d23c580c7881f650960a448fe2e54a30b3)]:
  - @langchain/core@1.1.9

## 1.2.3

### Patch Changes

- [#9707](https://github.com/langchain-ai/langchainjs/pull/9707) [`e5063f9`](https://github.com/langchain-ai/langchainjs/commit/e5063f9c6e9989ea067dfdff39262b9e7b6aba62) Thanks [@hntrl](https://github.com/hntrl)! - add security hardening for `load`

- Updated dependencies [[`e5063f9`](https://github.com/langchain-ai/langchainjs/commit/e5063f9c6e9989ea067dfdff39262b9e7b6aba62), [`8996647`](https://github.com/langchain-ai/langchainjs/commit/89966470e8c0b112ce4f9a326004af6a4173f9e6)]:
  - @langchain/core@1.1.8

## 1.2.2

### Patch Changes

- [#9675](https://github.com/langchain-ai/langchainjs/pull/9675) [`af664be`](https://github.com/langchain-ai/langchainjs/commit/af664becc0245b2315ea2f784c9a6c1d7622dbb4) Thanks [@jacoblee93](https://github.com/jacoblee93)! - Bump LangSmith dep to 0.4.0

- [#9673](https://github.com/langchain-ai/langchainjs/pull/9673) [`ffb2402`](https://github.com/langchain-ai/langchainjs/commit/ffb24026cd93e58219519ee24c6e23ea57cb5bde) Thanks [@hntrl](https://github.com/hntrl)! - add `context` utility

- Updated dependencies [[`df9c42b`](https://github.com/langchain-ai/langchainjs/commit/df9c42b3ab61b85309ab47256e1d93c3188435ee), [`8d2982b`](https://github.com/langchain-ai/langchainjs/commit/8d2982bb94c0f4e4314ace3cc98a1ae87571b1ed), [`af664be`](https://github.com/langchain-ai/langchainjs/commit/af664becc0245b2315ea2f784c9a6c1d7622dbb4), [`ffb2402`](https://github.com/langchain-ai/langchainjs/commit/ffb24026cd93e58219519ee24c6e23ea57cb5bde)]:
  - @langchain/core@1.1.7

## 1.2.1

### Patch Changes

- Updated dependencies [[`a7b2a7d`](https://github.com/langchain-ai/langchainjs/commit/a7b2a7db5ef57df3731ae6c9931f4b663e909505), [`a496c5f`](https://github.com/langchain-ai/langchainjs/commit/a496c5fc64d94cc0809216325b0f1bfde3f92c45), [`1da1325`](https://github.com/langchain-ai/langchainjs/commit/1da1325aea044fb37af54a9de1f4ae0b9f47d4a2)]:
  - @langchain/core@1.1.6

## 1.2.0

### Minor Changes

- [#9651](https://github.com/langchain-ai/langchainjs/pull/9651) [`348c37c`](https://github.com/langchain-ai/langchainjs/commit/348c37c01a048c815fea1827c084878744e20742) Thanks [@christian-bromann](https://github.com/christian-bromann)! - feat(langchain): allow to set strict tag manually in providerStrategy #9578

## 1.1.6

### Patch Changes

- [#9586](https://github.com/langchain-ai/langchainjs/pull/9586) [`bc8e90f`](https://github.com/langchain-ai/langchainjs/commit/bc8e90f4f77d71f739c8faf3e6c22ab7e54ffc3c) Thanks [@hntrl](https://github.com/hntrl)! - patch prompts created from runs fix

- [#9623](https://github.com/langchain-ai/langchainjs/pull/9623) [`ade8b8a`](https://github.com/langchain-ai/langchainjs/commit/ade8b8af0b32a9afd5c5a0bf6c4543d3cb7fd848) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): properly retrieve structured output from thinking block

- [#9637](https://github.com/langchain-ai/langchainjs/pull/9637) [`88bb788`](https://github.com/langchain-ai/langchainjs/commit/88bb7882fadf185bad927277810c682c2eee8d01) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): Prevent functions from being accidentally assignable to AgentMiddleware

- [#8964](https://github.com/langchain-ai/langchainjs/pull/8964) [`38ff1b5`](https://github.com/langchain-ai/langchainjs/commit/38ff1b55d353196b8af7f64f7b854b8f643e3de9) Thanks [@jnjacobson](https://github.com/jnjacobson)! - add support for anyOf, allOf, oneOf in openapi conversion

- [#9640](https://github.com/langchain-ai/langchainjs/pull/9640) [`aa8c4f8`](https://github.com/langchain-ai/langchainjs/commit/aa8c4f867abe79b1c6de09a7b51a69163d0972aa) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): prevent summarization middleware from leaking streaming events

- [#9648](https://github.com/langchain-ai/langchainjs/pull/9648) [`29a8480`](https://github.com/langchain-ai/langchainjs/commit/29a8480799d4c3534892a29cef4a135c437deb9b) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): allow to set strict tag manually in providerStrategy #9578

- [#9630](https://github.com/langchain-ai/langchainjs/pull/9630) [`a2df2d4`](https://github.com/langchain-ai/langchainjs/commit/a2df2d422e040485da61120bbbda6ced543e578b) Thanks [@nephix](https://github.com/nephix)! - fix(summary-middleware): use summaryPrefix or fall back to default prefix

- Updated dependencies [[`005c729`](https://github.com/langchain-ai/langchainjs/commit/005c72903bcdf090e0f4c58960c8c243481f9874), [`ab78246`](https://github.com/langchain-ai/langchainjs/commit/ab782462753e6c3ae5d55c0c251f795af32929d5), [`8cc81c7`](https://github.com/langchain-ai/langchainjs/commit/8cc81c7cee69530f7a6296c69123edbe227b2fce), [`f32e499`](https://github.com/langchain-ai/langchainjs/commit/f32e4991d0e707324e3f6af287a1ee87ab833b7e), [`a28d83d`](https://github.com/langchain-ai/langchainjs/commit/a28d83d49dd1fd31e67b52a44abc70f2cc2a2026), [`2e5ad70`](https://github.com/langchain-ai/langchainjs/commit/2e5ad70d16c1f13eaaea95336bbe2ec4a4a4954a), [`e456c66`](https://github.com/langchain-ai/langchainjs/commit/e456c661aa1ab8f1ed4a98c40616f5a13270e88e), [`1cfe603`](https://github.com/langchain-ai/langchainjs/commit/1cfe603e97d8711343ae5f1f5a75648e7bd2a16e)]:
  - @langchain/core@1.1.5

## 1.1.5

### Patch Changes

- Updated dependencies [[`0bade90`](https://github.com/langchain-ai/langchainjs/commit/0bade90ed47c7988ed86f1e695a28273c7b3df50), [`6c40d00`](https://github.com/langchain-ai/langchainjs/commit/6c40d00e926f377d249c2919549381522eac8ed1)]:
  - @langchain/core@1.1.4

## 1.1.4

### Patch Changes

- Updated dependencies [[`bd2c46e`](https://github.com/langchain-ai/langchainjs/commit/bd2c46e09e661d9ac766c09e71bc6687d6fc811c), [`487378b`](https://github.com/langchain-ai/langchainjs/commit/487378bf14277659c8ca0ef06ea0f9836b818ff4), [`138e7fb`](https://github.com/langchain-ai/langchainjs/commit/138e7fb6280705457079863bedb238b16b322032)]:
  - @langchain/core@1.1.3

## 1.1.3

### Patch Changes

- [#9532](https://github.com/langchain-ai/langchainjs/pull/9532) [`3424293`](https://github.com/langchain-ai/langchainjs/commit/34242933ade61304481d055605af06ec54c8f5e4) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): improve state rendering in LangSmith studio

- [#9529](https://github.com/langchain-ai/langchainjs/pull/9529) [`0d2f74a`](https://github.com/langchain-ai/langchainjs/commit/0d2f74aeef2c05ddcf74dc6286bfa8eabf785ed2) Thanks [@christian-bromann](https://github.com/christian-bromann)! - fix(langchain): better detect invoke response in agent node

- [#9523](https://github.com/langchain-ai/langchainjs/pull/9523) [`95a8f78`](https://github.com/langchain-ai/langchainjs/commit/95a8f78179b38b69419079c7b9315844f49aab0c) Thanks [@LiteracyFanatic](https://github.com/LiteracyFanatic)! - Fix detection of models with native support for structured output

- Updated dependencies [[`833f578`](https://github.com/langchain-ai/langchainjs/commit/833f57834dc3aa64e4cfdd7499f865b2ab41462a)]:
  - @langchain/core@1.1.2

## 1.1.2

### Patch Changes

- Updated dependencies [[`636b994`](https://github.com/langchain-ai/langchainjs/commit/636b99459bf843362298866211c63a7a15c2a319), [`38f0162`](https://github.com/langchain-ai/langchainjs/commit/38f0162b7b2db2be2c3a75ae468728adcb49fdfb)]:
  - @langchain/core@1.1.1

## 1.1.1

### Patch Changes

- [#9487](https://github.com/langchain-ai/langchainjs/pull/9487) [`4827945`](https://github.com/langchain-ai/langchainjs/commit/48279457ee44f36cdde175a537e2b12f5866627f) Thanks [@hntrl](https://github.com/hntrl)! - constrain lower bound core peer dep

## 1.1.0

### Minor Changes

- [#9476](https://github.com/langchain-ai/langchainjs/pull/9476) [`2a47c77`](https://github.com/langchain-ai/langchainjs/commit/2a47c77c29a873c4c4d4940458e0d5fb3b2e45ce) Thanks [@christian-bromann](https://github.com/christian-bromann)! - add new modelRetryMiddleware

- [#9475](https://github.com/langchain-ai/langchainjs/pull/9475) [`708d360`](https://github.com/langchain-ai/langchainjs/commit/708d360df1869def7e4caaa5995d6e907bbf54cd) Thanks [@christian-bromann](https://github.com/christian-bromann)! - Support `SystemMessage` as `systemPrompt`

- [#9475](https://github.com/langchain-ai/langchainjs/pull/9475) [`708d360`](https://github.com/langchain-ai/langchainjs/commit/708d360df1869def7e4caaa5995d6e907bbf54cd) Thanks [@christian-bromann](https://github.com/christian-bromann)! - Add OpenAI content moderation middleware

### Patch Changes

- [#9467](https://github.com/langchain-ai/langchainjs/pull/9467) [`2750e08`](https://github.com/langchain-ai/langchainjs/commit/2750e08547614de366019584940fdb1ba93e581c) Thanks [@hntrl](https://github.com/hntrl)! - allow overriding profiles in `initChatModel`

- [#9467](https://github.com/langchain-ai/langchainjs/pull/9467) [`2750e08`](https://github.com/langchain-ai/langchainjs/commit/2750e08547614de366019584940fdb1ba93e581c) Thanks [@hntrl](https://github.com/hntrl)! - cache model instance imports for `initChatModel`

- [#9416](https://github.com/langchain-ai/langchainjs/pull/9416) [`0fe9beb`](https://github.com/langchain-ai/langchainjs/commit/0fe9bebee6710f719e47f913eec1ec4f638e4de4) Thanks [@hntrl](https://github.com/hntrl)! - fix 'moduleResultion: "node"' compatibility

- [#9467](https://github.com/langchain-ai/langchainjs/pull/9467) [`2750e08`](https://github.com/langchain-ai/langchainjs/commit/2750e08547614de366019584940fdb1ba93e581c) Thanks [@hntrl](https://github.com/hntrl)! - pass model profiles from chat models in `initChatModel`

## 1.0.6

### Patch Changes

- [#9434](https://github.com/langchain-ai/langchainjs/pull/9434) [`f7cfece`](https://github.com/langchain-ai/langchainjs/commit/f7cfecec29bf0f121e1a8b0baface5327d731122) Thanks [@deepansh946](https://github.com/deepansh946)! - Updated error handling behaviour of AgentNode

## 1.0.5

### Patch Changes

- [#9403](https://github.com/langchain-ai/langchainjs/pull/9403) [`944bf56`](https://github.com/langchain-ai/langchainjs/commit/944bf56ff0926e102c56a3073bfde6b751c97794) Thanks [@christian-bromann](https://github.com/christian-bromann)! - improvements to toolEmulator middleware

- [#9388](https://github.com/langchain-ai/langchainjs/pull/9388) [`831168a`](https://github.com/langchain-ai/langchainjs/commit/831168a5450bff706a319842626214281204346d) Thanks [@hntrl](https://github.com/hntrl)! - use `profile.maxInputTokens` in summarization middleware

- [#9393](https://github.com/langchain-ai/langchainjs/pull/9393) [`f1e2f9e`](https://github.com/langchain-ai/langchainjs/commit/f1e2f9eeb365bae78c8b5991ed41bfed58f25da6) Thanks [@christian-bromann](https://github.com/christian-bromann)! - align context editing with summarization interface

- [#9427](https://github.com/langchain-ai/langchainjs/pull/9427) [`bad7aea`](https://github.com/langchain-ai/langchainjs/commit/bad7aea86d3f60616952104c34a33de9561867c7) Thanks [@dqbd](https://github.com/dqbd)! - fix(langchain): add tool call contents and tool call ID to improve token count approximation

- [#9396](https://github.com/langchain-ai/langchainjs/pull/9396) [`ed6b581`](https://github.com/langchain-ai/langchainjs/commit/ed6b581e525cdf5d3b29abb1e17ca6169554c1b5) Thanks [@christian-bromann](https://github.com/christian-bromann)! - rename exit behavior from throw to error

## 1.0.4

### Patch Changes

- b401680: avoid invalid message order after summarization
- f63fc0f: fix(langchain): export ToolRuntime from langchain

## 1.0.3

### Patch Changes

- f1583cd: allow for model strings in summarization middleware
- e960f97: check message property when pulling chat models for vercel compat
- 66fc10c: fix(langchain): don't allow default or optional context schemas
- 0a8a23b: feat(@langchain/core): support of ToolRuntime
- b38be50: Add missing ToolMessage in toolStrategy structured output
- 42930b5: fix(langchain): improved state schema typing

## 1.0.2

### Patch Changes

- 2e45c43: fix(langchain): remove bad dynamic import for LS
- 28eceac: preserve full model name when deciding model provider

## 1.0.0

🎉 **LangChain v1.0** is here! This release provides a focused, production-ready foundation for building agents. We've streamlined the framework around three core improvements: **`createAgent`**, **standard content blocks**, and a **simplified package structure**. See the [release notes](https://docs.langchain.com/oss/javascript/releases/langchain-v1) for complete details.

### ✨ Major Features

#### `createAgent` - A new standard for building agents

`createAgent` is the new standard way to build agents in LangChain 1.0. It provides a simpler interface than `createReactAgent` from LangGraph while offering greater customization potential through middleware.

**Key features:**

- **Clean, intuitive API**: Build agents with minimal boilerplate
- **Built on LangGraph**: Get persistence, streaming, human-in-the-loop, and time travel out of the box
- **Middleware-first design**: Highly customizable through composable middleware
- **Improved structured output**: Generate structured outputs in the main agent loop without additional LLM calls

Example:

```typescript
import { createAgent } from "langchain";

const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5-20250929",
  tools: [getWeather],
  systemPrompt: "You are a helpful assistant.",
});

const result = await agent.invoke({
  messages: [{ role: "user", content: "What is the weather in Tokyo?" }],
});

console.log(result.content);
```

Under the hood, `createAgent` is built on the basic agent loop—calling a model using LangGraph, letting it choose tools to execute, and then finishing when it calls no more tools.

**Built on LangGraph features (work out of the box):**

- **Persistence**: Conversations automatically persist across sessions with built-in checkpointing
- **Streaming**: Stream tokens, tool calls, and reasoning traces in real-time
- **Human-in-the-loop**: Pause agent execution for human approval before sensitive actions
- **Time travel**: Rewind conversations to any point and explore alternate paths

**Structured output improvements:**

- Generate structured outputs in the main loop instead of requiring an additional LLM call
- Models can choose between calling tools or using provider-side structured output generation
- Significant cost reduction by eliminating extra LLM calls

Example:

```typescript
import { createAgent } from "langchain";
import * as z from "zod";

const weatherSchema = z.object({
  temperature: z.number(),
  condition: z.string(),
});

const agent = createAgent({
  model: "openai:gpt-4o-mini",
  tools: [getWeather],
  responseFormat: weatherSchema,
});

const result = await agent.invoke({
  messages: [{ role: "user", content: "What is the weather in Tokyo?" }],
});

console.log(result.structuredResponse);
```

For more information, see [Agents documentation](https://docs.langchain.com/oss/javascript/langchain/agents).

#### Middleware

Middleware is what makes `createAgent` highly customizable, raising the ceiling for what you can build. Great agents require **context engineering**—getting the right information to the model at the right time. Middleware helps you control dynamic prompts, conversation summarization, selective tool access, state management, and guardrails through a composable abstraction.

**Prebuilt middleware** for common patterns:

```typescript
import {
  createAgent,
  summarizationMiddleware,
  humanInTheLoopMiddleware,
  piiRedactionMiddleware,
} from "langchain";

const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5-20250929",
  tools: [readEmail, sendEmail],
  middleware: [
    piiRedactionMiddleware({ patterns: ["email", "phone", "ssn"] }),
    summarizationMiddleware({
      model: "anthropic:claude-sonnet-4-5-20250929",
      maxTokensBeforeSummary: 500,
    }),
    humanInTheLoopMiddleware({
      interruptOn: {
        sendEmail: {
          allowedDecisions: ["approve", "edit", "reject"],
        },
      },
    }),
  ] as const,
});
```

**Custom middleware** with lifecycle hooks:

| Hook            | When it runs             | Use cases                               |
| --------------- | ------------------------ | --------------------------------------- |
| `beforeAgent`   | Before calling the agent | Load memory, validate input             |
| `beforeModel`   | Before each LLM call     | Update prompts, trim messages           |
| `wrapModelCall` | Around each LLM call     | Intercept and modify requests/responses |
| `wrapToolCall`  | Around each tool call    | Intercept and modify tool execution     |
| `afterModel`    | After each LLM response  | Validate output, apply guardrails       |
| `afterAgent`    | After agent completes    | Save results, cleanup                   |

Example custom middleware:

```typescript
import { createMiddleware } from "langchain";

const contextSchema = z.object({
  userExpertise: z.enum(["beginner", "expert"]).default("beginner"),
});

const expertiseBasedToolMiddleware = createMiddleware({
  wrapModelCall: async (request, handler) => {
    const userLevel = request.runtime.context.userExpertise;
    if (userLevel === "expert") {
      const tools = [advancedSearch, dataAnalysis];
      return handler(request.replace("openai:gpt-5", tools));
    }
    const tools = [simpleSearch, basicCalculator];
    return handler(request.replace("openai:gpt-5-nano", tools));
  },
});

const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5-20250929",
  tools: [simpleSearch, advancedSearch, basicCalculator, dataAnalysis],
  middleware: [expertiseBasedToolMiddleware],
  contextSchema,
});
```

For more information, see the [complete middleware guide](https://docs.langchain.com/oss/javascript/langchain/middleware).

#### Simplified Package

LangChain v1 streamlines the `langchain` package namespace to focus on essential building blocks for agents. The package exposes only the most useful and relevant functionality (most re-exported from `@langchain/core` for convenience).

**What's in the core `langchain` package:**

- `createAgent` and agent-related utilities
- Core message types and content blocks
- Middleware infrastructure
- Tool definitions and schemas
- Prompt templates
- Output parsers
- Base runnable abstractions

### 🔄 Migration Notes

#### `@langchain/classic` for Legacy Functionality

Legacy functionality has moved to [`@langchain/classic`](https://www.npmjs.com/package/@langchain/classic) to keep the core package lean and focused.

**What's in `@langchain/classic`:**

- Legacy chains and chain implementations
- The indexing API
- [`@langchain/community`](https://www.npmjs.com/package/@langchain/community) exports
- Other deprecated functionality

**To migrate legacy code:**

1. Install `@langchain/classic`:

   ```bash
   npm install @langchain/classic
   ```

2. Update your imports:

   ```typescript
   import { ... } from "langchain"; // [!code --]
   import { ... } from "@langchain/classic"; // [!code ++]

   import { ... } from "langchain/chains"; // [!code --]
   import { ... } from "@langchain/classic/chains"; // [!code ++]
   ```

#### Upgrading to v1

Install the v1 packages:

```bash
npm install langchain@1.0.0 @langchain/core@1.0.0
```

### 📚 Additional Resources

- [LangChain 1.0 Announcement](https://blog.langchain.com/langchain-langchain-1-0-alpha-releases/)
- [Migration Guide](https://docs.langchain.com/oss/javascript/migrate/langchain-v1)
- [Agents Documentation](https://docs.langchain.com/oss/javascript/langchain/agents)
- [Middleware Guide](https://blog.langchain.com/agent-middleware/)

---

## 0.3.36

### Patch Changes

- cabd762: fix(langchain): add ChatMistralAI to well known models
- Updated dependencies [e63c7cc]
- Updated dependencies [b8ffc1e]
  - @langchain/openai@0.6.16

## 0.3.35

### Patch Changes

- fd4691f: use `keyEncoder` instead of insecure cache key getter
- 2f19cd5: feat: Add Perplexity support to universal chat model
- 3c94076: fix(langchain): Bind schemas for other types of pulled hub prompts
- Updated dependencies [d38e9d6]
  - @langchain/openai@0.6.14

## 0.3.34

### Patch Changes

- 6019a7d: update JSONL loader to support complex json structures
- caf5579: prevent ConfigurableModel mutation when using withStructuredOutput or bindTools
- d60f40f: infer mistralai models
- Updated dependencies [41bd944]
- Updated dependencies [707a768]
  - @langchain/openai@0.6.12

## 0.3.33

### Patch Changes

- d2c7f09: support prompts not created from RunnableBinding

## 0.3.32

### Patch Changes

- e0bd88c: add support for conversion of ref in array schema
- Updated dependencies [4a3f5af]
- Updated dependencies [424360b]
  - @langchain/openai@0.6.10