dima/langchain
1
0 Fork 0
mirror of https://github.com/langchain-ai/langchain.git synced 2026-10-05 09:25:14 +03:00
Code Issues Packages Projects Releases Wiki Activity
16,872 Commits 1,233 Branches 1,383 Tags
a9780cd3dd73135d21d7130b08711685f2700d51
Commit Graph
9925 Commits
Author SHA1 Message Date
ccurme d6167c0b0d release(anthropic): 1.7.5 (#40912) 2026-09-29 11:18:21 -04:00
ccurmeccurmeopen-swe[bot]
 		aaf25d0abd test(openai): drop retired completions live tests (#40910) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-29 10:27:52 -04:00
langchain-oss-model-profiles[bot]mdrxyMason Daughertyopen-swe[bot]
 		04ac76c07e chore(model-profiles): refresh model profile data (#40902) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**4 added · 3 removed · 18 changed** across 3 provider(s).

<details>
<summary>anthropic</summary>

**➕ 1 added**
- `claude-sonnet-5-5` — 1,000,000 ctx, 128,000 out, text+image+pdf in,
reasoning, tools

</details>

<details>
<summary>mistral</summary>

**➖ 1 removed**
- `magistral-small`

</details>

<details>
<summary>openrouter</summary>

**➕ 3 added**
- `anthropic/claude-sonnet-5.5` — 1,000,000 ctx, 128,000 out,
text+image+pdf in, reasoning, tools
- `nex-agi/nex-n2.5-mini` — 262,144 ctx, 235,929 out, text+image in,
reasoning
- `nex-agi/nex-n2.5-pro` — 262,144 ctx, 235,929 out, text+image in,
reasoning, tools

**➖ 2 removed**
- `deepseek/deepseek-r1-distill-llama-70b`
- `inclusionai/ling-3.0-flash-fin:free`

**✏️ 18 changed**
- `deepseek/deepseek-v3.1-terminus`: max output tokens 32,768 → 65,536
- `deepseek/deepseek-v3.2-exp`: max output tokens 65,536 → 147,456
- `deepseek/deepseek-v4-flash-vision-exp`: max output tokens 943,717 →
262,144
- `meta/muse-spark-1.1`: removed audio input
- `meta/muse-spark-1.2`: removed audio input
- `meta/muse-spark-1.2-contributor`: removed audio input
- `meta/muse-spark-1.3`: removed audio input
- `meta/muse-spark-1.3-contributor`: removed audio input
- `minimax/minimax-m2.7`: max output tokens 131,072 → 176,947
- `nvidia/nemotron-3.5-lightning`: max output tokens 131,072 → 32,768
- `qwen/qwen3-30b-a3b`: max output tokens 16,384 → 8,192
- `qwen/qwen3-30b-a3b-instruct-2507`: max output tokens 235,929 → 32,000
- `qwen/qwen3.8-27b`: max output tokens 131,072 → 235,929
- `z-ai/glm-5.2`: max output tokens 131,072 → 943,718
- `~anthropic/claude-sonnet-latest`: added temperature control
- `~deepseek/deepseek-pro-latest`: max output tokens 393,216 → 943,718
- `~z-ai/glm-flash-latest`: max output tokens 128,000 → 943,718
- `~z-ai/glm-latest`: max output tokens 943,718 → 131,072

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-29 10:14:22 -04:00
ccurme 08064f4859 release(core): 1.6.6 (#40906) 2026-09-29 09:25:34 -04:00
Hunter LovellHunter Lovellopen-swe[bot]ccurmeChester Curme
 		ce9066138d fix(anthropic): support Claude Sonnet 5.5 compatibility (#40882) 
Co-authored-by: Hunter Lovell <hntrl@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
 2026-09-29 09:05:37 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		78a3cbcc6b hotfix(fireworks): replace unavailable integration test model (#40890) 
The [Fireworks release
job](https://github.com/langchain-ai/langchain/actions/runs/36479033898/job/109120052547)
failed 55 tests because `kimi-k2p6` returned `404 NOT_FOUND`; the
`gpt-oss-120b` chat tests passed in that same job.

- Use `accounts/fireworks/models/gpt-oss-120b` across the affected chat,
completions, and standard integration tests, reusing the existing
chat-model constant.
- Preserve all assertions and coverage; leave production defaults and
release workflows unchanged.
- Fireworks [advertises GPT-OSS-120B as
serverless](https://fireworks.ai/models/fireworks/gpt-oss-120b). Live
completions and expanded chat coverage still need confirmation with CI
credentials: no Fireworks API key is available locally. This is not
evidence that Kimi was retired.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.langchain.dev/agents/b8a0feec-8262-501b-8d6f-204f467461e3)
· openai:gpt-6-astra (medium)

Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-28 16:41:02 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		316c045803 release(fireworks): 1.7.0 (#40889) 
Prepare the `langchain-fireworks` **1.7.0 minor release**, up from
1.6.3, for prompt-caching middleware support.

- Raise the existing `langchain` test dependency minimum to
`>=1.4.3,<2.0.0` for fallback-safe prompt caching; `langchain` remains a
lazy, optional runtime import, not a required package dependency.
- Refresh the lockfile for Fireworks 1.7.0 and LangChain 1.4.3,
including the latter's current package metadata. No publishing is
performed by this PR.

### PRs included since 1.6.3

- [#38823](https://github.com/langchain-ai/langchain/pull/38823): Add
prompt caching middleware.
- [#40874](https://github.com/langchain-ai/langchain/pull/40874):
Classify mid-stream read timeouts.
- [#40833](https://github.com/langchain-ai/langchain/pull/40833):
Refresh model profile data.

Existing lockfile caveat: `uv` warns that `pydantic==2.12.1` and
`pydantic-core==2.41.3` are yanked (the former references the latter;
the latter had a corrupted wheel upload). Those versions are unchanged
by this release PR.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.langchain.dev/agents/c2dee156-b2b1-573e-8f59-e498d251ae49)
· openai:gpt-6-astra (medium)

Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-28 16:24:46 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		be854a1301 release(langchain): 1.4.3 (#40888) 
Prepare the `langchain` 1.4.3 patch release. Bumps package metadata and
the lockfile from 1.4.2; no dependency changes or publishing are
performed by this PR.

### PRs included since 1.4.2

- [#40886](https://github.com/langchain-ai/langchain/pull/40886):
Sanitize cache settings for fallback models.
- [#40837](https://github.com/langchain-ai/langchain/pull/40837):
Support Bedrock Mantle chat models in `init_chat_model`.
- [#40844](https://github.com/langchain-ai/langchain/pull/40844):
Recognize GPT-6 structured output without profiles.
- [#40530](https://github.com/langchain-ai/langchain/pull/40530): Repair
invalid tool calls in `create_agent`.
- [#40713](https://github.com/langchain-ai/langchain/pull/40713): Remove
the commented-out Cohere extra.
- [#40626](https://github.com/langchain-ai/langchain/pull/40626): Bump
locked AnyIO from 4.11.0 to 4.14.2.
- [#40794](https://github.com/langchain-ai/langchain/pull/40794):
Correct repository setup guidance and package documentation.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.langchain.dev/agents/c2dee156-b2b1-573e-8f59-e498d251ae49)
· openai:gpt-6-astra (medium)

Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-28 16:14:05 -04:00
Mason Daugherty 2ade674ffc fix(langchain): sanitize cache settings for fallback models (#40886) 
`ModelFallbackMiddleware` now removes unsupported cache keys and
Fireworks session-affinity headers before fallback attempts, while
preserving cache settings supported by the selected fallback model.

---

When an agent falls back to another provider, explicitly supplied cache
settings in `ModelRequest.model_settings` can reach a model that does
not accept them and cause the fallback to fail.

`ModelFallbackMiddleware` now removes `x-session-affinity` for
non-Fireworks fallbacks and removes `prompt_cache_key` for providers
outside Fireworks, OpenAI, and Azure OpenAI. It preserves unrelated
settings and headers, retains the existing Anthropic cache-marker
handling, and leaves the original request unchanged. Unit tests cover
synchronous and asynchronous fallback, supported cache-key preservation,
and header cleanup.

Stacked on #38823. This PR contains only the fallback cleanup and its
tests. The Fireworks middleware in the base PR works independently
because its generated affinity is consumed directly by `ChatFireworks`.

Review focus: provider support is determined through `_llm_type`;
`prompt_cache_key` is shared by multiple providers and must not be
treated as Fireworks-only.
 2026-09-28 16:09:47 -04:00
Mason Daughertyandopen-swe[bot] 88b972731a feat(fireworks): add prompt caching middleware (#38823) 
Added `FireworksPromptCachingMiddleware` to improve prompt-cache reuse
across calls in the same agent thread. Explicit affinity settings take
precedence, and no affinity is generated without a thread ID.

---

Fireworks agents need consistent routing to reuse a replica's prompt
cache across turns. `FireworksPromptCachingMiddleware` supplies session
affinity from a SHA-256 hash of `config.configurable.thread_id`, while
respecting explicit `user`, `prompt_cache_key`, and `x-session-affinity`
settings on the selected model or request.

Affinity is scoped to the call and applied by `ChatFireworks` when
invoking the API. Generated affinity stays out of shared request
settings, and model-local headers remain scoped to their owning model,
including during fallback. This works with either ordering of the
caching and fallback middleware for a Fireworks primary model.

Related fallback cleanup for explicitly supplied cache settings is in
#40886, stacked on this PR. This PR works independently of that change.

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-28 16:06:40 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		60e57f5bc5 fix(fireworks): classify mid-stream read timeouts (#40874) 
`ChatFireworks` now reports mid-stream read timeouts as retryable
`ModelTimeoutError`s while remaining catchable as `httpx.ReadTimeout`.
Partial streams are not automatically replayed.

---

Fireworks stream read timeouts currently bypass LangChain's model-error
classification after the first chunk. Classify them as retryable
`ModelTimeoutError`s while preserving `httpx.ReadTimeout` compatibility,
the original cause, and request context when available.

- Covers synchronous and asynchronous stream consumption.
- Keeps setup retries unchanged. Does **not** replay partial streams:
doing so could duplicate text or corrupt tool-call arguments.
Whole-generation recovery remains the caller's responsibility.
- Built-in streaming retry and fallback behavior is unchanged:
`.with_retry()` does not retry streaming, and `.with_fallbacks()` only
switches models before the first chunk. Applications must explicitly
handle recovery after output has started.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/c2ba7088-1448-5e4d-9541-ae6f25b513c7)
· openai:gpt-6-astra (medium)

Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-28 14:18:14 -04:00
langchain-oss-model-profiles[bot]andmdrxy 213f230d64 chore(model-profiles): refresh model profile data (#40869) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**2 added · 0 removed · 10 changed** across 2 provider(s).

<details>
<summary>openai</summary>

**➕ 2 added**
- `gpt-daybreak-blue-latest` — 1,050,000 ctx, 128,000 out,
text+image+pdf in, reasoning, tools
- `gpt-daybreak-red-latest` — 400,000 ctx, 128,000 out, text+image+pdf
in, reasoning, tools

</details>

<details>
<summary>openrouter</summary>

**✏️ 10 changed**
- `deepseek/deepseek-chat`: max output tokens 16,384 → 16,000
- `deepseek/deepseek-v4-flash-vision-exp`: max output tokens 262,144 →
943,717
- `deepseek/deepseek-v4.1-flash`: max output tokens 384,000 → 943,718
- `minimax/minimax-m2`: max output tokens 131,072 → 176,947
- `minimax/minimax-m2.7`: max output tokens 176,947 → 131,072
- `qwen/qwen3-30b-a3b`: max output tokens 8,192 → 16,384
- `qwen/qwen3.5-35b-a3b`: max output tokens 16,384 → 65,536
- `z-ai/glm-5.3`: max output tokens 943,718 → 943,717
- `z-ai/glm-5.3-flash`: max output tokens 128,000 → 943,717
- `~deepseek/deepseek-flash-latest`: max output tokens 384,000 → 943,718

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
 2026-09-28 09:22:43 -04:00
langchain-oss-model-profiles[bot]andmdrxy d7508191bc chore(model-profiles): refresh model profile data (#40833) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**3 added · 11 removed · 27 changed** across 2 provider(s).

<details>
<summary>fireworks-ai</summary>

**➖ 7 removed**
- `accounts/fireworks/models/deepseek-v4-flash-0731`
- `accounts/fireworks/models/deepseek-v4-flash-vision-exp`
- `accounts/fireworks/models/deepseek-v4-pro-0813`
- `accounts/fireworks/models/minimax-m2p7`
- `accounts/fireworks/models/muse-glimmer-30b`
- `accounts/fireworks/models/qwen3p7-plus`
- `accounts/fireworks/routers/deepseek-pro-latest`

**✏️ 5 changed**
- `accounts/fireworks/models/deepseek-v4-pro`: attachments no → unset;
audio input no → unset; audio output no → unset; image input no → unset;
image output no → unset; last updated `2026-04-24` → unset; max input
tokens 1,000,000 → unset; max output tokens 384,000 → unset; display
name `DeepSeek V4 Pro` → unset; removed open weights; removed reasoning;
release date `2026-04-24` → unset; status `deprecated` → unset; removed
structured output; removed temperature control; removed text input;
removed text output; removed tool calling; video input no → unset; video
output no → unset
- `accounts/fireworks/models/glm-5p2`: attachments no → unset; audio
input no → unset; audio output no → unset; image input no → unset; image
output no → unset; last updated `2026-06-16` → unset; max input tokens
1,048,575 → unset; max output tokens 131,072 → unset; display name `GLM
5.2` → unset; removed open weights; removed reasoning; release date
`2026-06-16` → unset; status `deprecated` → unset; removed temperature
control; removed text input; removed text output; removed tool calling;
video input no → unset; video output no → unset
- `accounts/fireworks/models/kimi-k2p6`: removed attachments; audio
input no → unset; audio output no → unset; removed image input; image
output no → unset; last updated `2026-04-17` → unset; max input tokens
262,000 → unset; max output tokens 262,000 → unset; display name `Kimi
K2.6` → unset; removed open weights; removed reasoning; release date
`2026-04-17` → unset; status `deprecated` → unset; removed temperature
control; removed text input; removed text output; removed tool calling;
video input no → unset; video output no → unset
- `accounts/fireworks/models/kimi-k2p7-code`: removed attachments; audio
input no → unset; audio output no → unset; removed image input; image
output no → unset; last updated `2026-06-16` → unset; max input tokens
262,000 → unset; max output tokens 262,000 → unset; display name `Kimi
K2.7 Code` → unset; removed open weights; removed reasoning; release
date `2026-06-12` → unset; status `deprecated` → unset; removed
temperature control; removed text input; removed text output; removed
tool calling; video input no → unset; video output no → unset
- `accounts/fireworks/routers/glm-5p2-fast`: attachments no → unset;
audio input no → unset; audio output no → unset; image input no → unset;
image output no → unset; last updated `2026-06-26` → unset; max input
tokens 1,048,575 → unset; max output tokens 131,072 → unset; display
name `GLM 5.2 Fast` → unset; removed open weights; removed reasoning;
release date `2026-06-26` → unset; removed temperature control; removed
text input; removed text output; removed tool calling; video input no →
unset; video output no → unset

</details>

<details>
<summary>openrouter</summary>

**➕ 3 added**
- `mistralai/devstral-2512` — 262,144 ctx, 209,715 out, text+pdf in,
tools
- `mistralai/mistral-large-2512` — 262,144 ctx, 209,715 out,
text+image+pdf in, tools
- `perceptron/perceptron-mk1.5` — 36,864 ctx, 8,192 out,
text+image+audio+video in, reasoning, tools

**➖ 4 removed**
- `anthropic/claude-3-haiku`
- `nex-agi/nex-n2.5-mini:free`
- `nex-agi/nex-n2.5-pro:free`
- `z-ai/glm-5.2:free`

**✏️ 22 changed**
- `deepseek/deepseek-v4-flash`: max output tokens 384,000 → 131,072
- `deepseek/deepseek-v4-flash-vision-exp`: max output tokens 943,718 →
262,144
- `deepseek/deepseek-v4-pro-0813`: max output tokens 384,000 → 943,718
- `deepseek/deepseek-v4.1-flash`: max output tokens 131,072 → 384,000
- `minimax/minimax-01`: max output tokens 900,172 → 40,000
- `minimax/minimax-m2`: max output tokens 176,947 → 131,072
- `minimax/minimax-m2.7`: max output tokens 131,072 → 176,947
- `nvidia/nemotron-3.5-lightning`: max input tokens 262,144 → 1,000,000
- `qwen/qwen3-30b-a3b`: max output tokens 16,384 → 8,192
- `qwen/qwen3-vl-30b-a3b-instruct`: max output tokens 32,768 → 16,384
- `qwen/qwen3.5-122b-a10b`: max output tokens 65,536 → 235,929
- `qwen/qwen3.6-27b`: max output tokens 262,140 → 81,920
- `tencent/hy-mt2-30b-a3b`: max input tokens 32,768 → 8,192
- `thinkingmachines/inkling`: max input tokens 1,048,576 → 524,288
- `thinkingmachines/inkling-small`: max input tokens 1,048,576 → 524,288
- `xiaomi/mimo-v2.6-pro`: max input tokens 1,048,576 → 1,050,000
- `z-ai/glm-5.1`: max output tokens 128,000 → 131,072
- `z-ai/glm-5.3`: max output tokens 131,072 → 943,718
- `z-ai/glm-5.3-flash`: max output tokens 943,718 → 128,000
- `~deepseek/deepseek-flash-latest`: max output tokens 943,718 → 384,000
- `~deepseek/deepseek-pro-latest`: max output tokens 943,718 → 393,216
- `~z-ai/glm-latest`: max output tokens 131,072 → 943,718

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
 2026-09-27 15:48:38 -04:00
ccurmeccurmeopen-swe[bot]roshangardi
 		1ef23d6b7f fix(anthropic): serialize invalid tool calls as tool use on replay (#40864) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: roshangardi <34957845+roshangardi@users.noreply.github.com>
 2026-09-27 11:03:27 -04:00
ccurmeccurmeopen-swe[bot]
 		80b7409051 feat(langchain): support Bedrock Mantle chat models in init_chat_model (#40837) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-25 18:19:51 +00:00
xuanccurmeAman Gupta
 		40fe8d6e02 docs(core): fix docstring examples that don't run as copied (#40815) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: Aman Gupta <168967382+aman99dex@users.noreply.github.com>
 2026-09-25 14:18:56 -04:00
ccurmeandLucas Kim 021dce4671 fix(langchain): recognize GPT-6 structured output without profiles (#40844) 
Co-authored-by: Lucas Kim <41357160+kimnamu@users.noreply.github.com>
 2026-09-25 14:16:40 -04:00
ccurme e75dae1f53 release(fireworks): 1.6.3 (#40834) 2026-09-25 08:51:42 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		38cee0db98 fix(fireworks): declare native PDF inputs unsupported (#40814) 
Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-24 18:07:25 -04:00
Hunter Lovellhntrlopen-swe[bot]
 		846e161121 feat(openai): discover Azure workload identity (#40532) 
Co-authored-by: hntrl <hntrl@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-24 15:55:06 -04:00
Mason DaughertyMason Daughertyopen-swe[bot]
 		fbd70b73d4 fix(fireworks): preserve malformed tool arguments as diagnostic JSON (#40818) 
`ChatFireworks` can replay historical tool calls with malformed or
non-object JSON arguments without forwarding invalid argument strings to
Fireworks.

---

Fireworks rejects conversation history containing malformed or
non-object tool-call arguments, preventing an agent from recovering on
its next turn.

Wrap these arguments in a JSON object under
`__invalid_tool_call_arguments` when serializing history, preserving the
original payload, call IDs, and tool-result pairing. Apply this to
parsed and raw tool-call history without mutating messages or executing
repaired arguments; valid object strings stay unchanged.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/0da06f8d-3ea1-5f01-aa55-953acb9a71fa)
· openai:gpt-6-astra (medium)

Co-authored-by: Mason Daugherty <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-24 15:40:49 -04:00
ccurme c5ab14d42a release(core): 1.6.5 (#40816) 2026-09-24 14:03:30 -04:00
langchain-oss-model-profiles[bot]andmdrxy 5704d9d481 chore(model-profiles): refresh model profile data (#40804) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**8 added · 4 removed · 13 changed** across 3 provider(s).

<details>
<summary>deepseek</summary>

**✏️ 4 changed**
- `deepseek-flash`: max output tokens 384,000 → 393,216
- `deepseek-v4-flash`: max output tokens 384,000 → 393,216
- `deepseek-v4-flash-vision-exp`: max output tokens 384,000 → 393,216
- `deepseek-v4-pro`: max output tokens 384,000 → 393,216

</details>

<details>
<summary>fireworks-ai</summary>

**➕ 1 added**
- `accounts/fireworks/models/ember-1` — 1,048,576 ctx, 131,072 out,
text+image in, reasoning, tools

</details>

<details>
<summary>openrouter</summary>

**➕ 7 added**
- `aion-labs/aion-3.5` — 262,144 ctx, 32,768 out, reasoning, tools
- `aion-labs/aion-3.5-mini` — 262,144 ctx, 32,768 out, reasoning, tools
- `fireworks/ember-1` — 1,048,576 ctx, 943,718 out, text+image in,
reasoning, tools
- `qwen/qwen3.8-max-prime` — 1,000,000 ctx, 131,072 out,
text+image+video in, reasoning, tools
- `stealth/space-bunny-alpha` — 1,000,000 ctx, 524,288 out,
text+image+video in, reasoning, tools
- `upstage/solar-mini4` — 524,288 ctx, 131,072 out, reasoning, tools
- `z-ai/glm-5.3-prime` — 1,000,000 ctx, 131,072 out, reasoning, tools

**➖ 4 removed**
- `inclusionai/ling-3.0-flash-vl:free`
- `mistralai/devstral-2512`
- `nex-agi/nex-n2.5-mini`
- `nex-agi/nex-n2.5-pro`

**✏️ 9 changed**
- `deepseek/deepseek-v4.1-flash`: max output tokens 943,718 → 131,072
- `inclusionai/ling-3.0-flash-vl`: max input tokens 131,072 → 262,144
- `minimax/minimax-m2`: max output tokens 131,072 → 176,947
- `nvidia/nemotron-3.5-lightning`: max output tokens 235,929 → 131,072
- `qwen/qwen3-30b-a3b-instruct-2507`: max output tokens 32,000 → 235,929
- `qwen/qwen3-next-80b-a3b-instruct`: max output tokens 16,384 → 235,929
- `tencent/hy-mt2-30b-a3b`: max input tokens 8,192 → 32,768
- `~deepseek/deepseek-pro-latest`: max output tokens 384,000 → 943,718
- `~z-ai/glm-flash-latest`: max output tokens 131,072 → 128,000

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
 2026-09-24 10:36:28 -04:00
ccurme 7622d3dce7 release(openai): 1.6.6 (#40800) 2026-09-23 18:48:41 -04:00
Deepak Thorat 2dd956b8ad docs(infra): fix AGENTS.md root setup guidance and package doc accuracy (#40794) 
Closes #40793

---

Docs-only repair from a full audit of developer-facing markdown against
the repository as source of truth. A new contributor following
`AGENTS.md` from the repo root currently hits missing files and commands
that cannot run; package docs and a workflow comment also drift from
reality.

**What was wrong**

- `AGENTS.md` listed root-level `pyproject.toml`, `uv.lock`, and
`Makefile` as key config files — none exist at the repo root (config is
per package under `libs/*/`).
- Setup/test/lint examples (`uv sync --all-groups`, `make test` / `lint`
/ `format`) had no working directory, so they fail if copy-pasted from
the root.
- The monorepo structure tree omitted `openwiki/` and `AGENTS.md`.
- `libs/README.md` omitted the `model-profiles/` package from its
directory list.
- Root `README.md` linked Deep Agents with `http://` while every other
docs link uses `https://`.
- The PR-title paragraph claimed scopes are mandatory “with no
exceptions”, but `pr_lint.yml` sets `requireScope: false` (only empty
`type():` parens are rejected).
- Grammar (“require” → “requires”), an unfinished editable-installs
sentence, incomplete `pr_lint.yml` scope comment (missing `openrouter`,
`typesafe`), and a stale `make help` line pointing at a non-existent
top-level Makefile.

**What changed**

- Clarified per-package config layout and required `cd` into
`libs/<package>` before `uv` / `make` commands.
- Completed the structure diagram; added `model-profiles/` to
`libs/README.md`.
- Switched Deep Agents links to `https://`.
- Aligned the scope guidance with actual CI behavior (documented, did
not change `requireScope`).
- Tightened the `pr_lint.yml` comment and removed the dead Makefile help
line.

No runtime code, tests, or CI logic changed — comments and markdown
only.

AI assistance was used to prepare this change; I reviewed the diff
against the repository layout.
 2026-09-23 17:47:50 -04:00
ccurmeandAokiro 49f4b4016b fix(openai): raise on error events in stream path (#40791) 
Co-authored-by: Aokiro <mayanktharwani9@gmail.com>
 2026-09-23 15:47:14 -04:00
ccurmeccurmeopen-swe[bot]
 		19cadaa1a1 fix(core): abbreviate long tool IDs in XML buffer strings (#40792) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-23 15:36:26 -04:00
ccurme 798441e8b0 chore(anthropic): fix integration test cassette (#40790) 2026-09-23 13:52:22 -04:00
ccurme a476942bac release(openai): 1.6.5 (#40787) 2026-09-23 11:20:25 -04:00
ccurme 46c6bdf1b4 release(anthropic): 1.7.4 (#40786) 2026-09-23 11:19:43 -04:00
ccurmeccurmeopen-swe[bot]
 		290dabaff2 fix(anthropic): add Opus 5.5 and GPT-6 profile augmentations (#40785) 
Co-authored-by: ccurme <ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-23 11:09:07 -04:00
Richard Scarrottccurmeopen-swe[bot]Chester Curme
 		59baeb26d6 feat(anthropic,openai): mid-conversation tool changes on SystemMessage (#40758) 
Co-authored-by: ccurme <26529506+ccurme@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Chester Curme <chester.curme@gmail.com>
 2026-09-23 10:38:51 -04:00
langchain-oss-model-profiles[bot]andmdrxy 4b65996406 chore(model-profiles): refresh model profile data (#40780) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**7 added · 0 removed · 12 changed** across 2 provider(s).

<details>
<summary>deepseek</summary>

**✏️ 2 changed**
- `deepseek-v4-flash`: status unset → `deprecated`
- `deepseek-v4-flash-vision-exp`: status unset → `deprecated`

</details>

<details>
<summary>openrouter</summary>

**➕ 7 added**
- `anthropic/claude-opus-5.5` — 1,000,000 ctx, 128,000 out,
text+image+pdf in, reasoning, tools
- `cohere/command-a-plus` — 192,000 ctx, 64,000 out, text+image in,
reasoning, tools
- `openai/gpt-6-luna` — 1,050,000 ctx, 128,000 out, text+image+pdf in,
reasoning, tools
- `openai/gpt-6-luna-pro` — 1,050,000 ctx, 128,000 out, text+image+pdf
in, reasoning, tools
- `openai/gpt-6-sol` — 1,050,000 ctx, 128,000 out, text+image+pdf in,
reasoning, tools
- `openai/gpt-6-sol-pro` — 1,050,000 ctx, 128,000 out, text+image+pdf
in, reasoning, tools
- `qwen/qwen3.8-omni-flash` — 1,000,000 ctx, 131,072 out,
text+image+audio+video in, reasoning, tools

**✏️ 10 changed**
- `deepseek/deepseek-v4-pro-0813`: max output tokens 943,718 → 384,000
- `deepseek/deepseek-v4.1-flash`: max output tokens 384,000 → 943,718
- `openai/gpt-oss-20b`: max output tokens 117,964 → 32,768
- `qwen/qwen3.6-27b`: max output tokens 65,536 → 262,140
- `xiaomi/mimo-v2.6-flash`: added open weights
- `xiaomi/mimo-v2.6-pro`: added open weights
- `xiaomi/mimo-v2.6-pro-ultraspeed`: added open weights
- `~deepseek/deepseek-pro-latest`: max output tokens 943,718 → 384,000
- `~z-ai/glm-flash-latest`: max output tokens 943,718 → 131,072
- `~z-ai/glm-latest`: max output tokens 943,718 → 131,072

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
 2026-09-23 10:19:38 -04:00
ccurme 9fa192ea35 release(openai): 1.6.4 (#40775) 2026-09-22 18:28:20 -04:00
langchain-oss-model-profiles[bot]andccurme af6e0dbefd chore(model-profiles): refresh openai model profile data (#40774) 
Co-authored-by: ccurme <26529506+ccurme@users.noreply.github.com>
 2026-09-22 22:23:33 +00:00
ccurme ed8aad4e7b release(anthropic): 1.7.3 (#40773) 2026-09-22 18:03:16 -04:00
langchain-oss-model-profiles[bot]andccurme 2f1ecbf730 chore(model-profiles): refresh anthropic model profile data (#40772) 
Co-authored-by: ccurme <26529506+ccurme@users.noreply.github.com>
 2026-09-22 21:55:50 +00:00
ccurmeandopen-swe[bot] bfa03c8cf8 fix(langchain): repair invalid tool calls in create_agent (#40530) 
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-22 15:22:59 -04:00
ccurme fbac0b2d0c fix(anthropic): auto-route with_structured_output to method="json_schema" for fable and opus 5.5 (#40766) 2026-09-22 14:24:51 -04:00
ccurme 3971e49d24 chore(anthropic): update docs for Opus 5.5 (#40765) 2026-09-22 13:42:48 -04:00
langchain-oss-model-profiles[bot]andmdrxy 877ad8c7bf chore(model-profiles): refresh model profile data (#40747) 
Automated refresh of model profile data for all in-monorepo partner
integrations via `langchain-profiles refresh`.

🤖 Generated by the [`refresh_model_profiles`
workflow](https://github.com/langchain-ai/langchain/blob/master/.github/workflows/refresh_model_profiles.yml).

## Summary of changes

**9 added · 2 removed · 5 changed** across 3 provider(s).

<details>
<summary>huggingface</summary>

**➕ 1 added**
- `tencent/Hy4-preview` — 1,000,000 ctx, 64,000 out, reasoning, tools

</details>

<details>
<summary>openrouter</summary>

**➕ 6 added**
- `nex-agi/nex-n2.5-mini` — 262,144 ctx, 235,929 out, text+image in,
reasoning
- `nex-agi/nex-n2.5-pro` — 262,144 ctx, 235,929 out, text+image in,
reasoning, tools
- `x-ai/grok-4.7` — 500,000 ctx, 450,000 out, text+image+pdf in,
reasoning, tools
- `xiaomi/mimo-v2.6-flash` — 1,048,576 ctx, 131,072 out,
text+image+audio+video in, reasoning, tools
- `xiaomi/mimo-v2.6-pro` — 1,048,576 ctx, 131,072 out,
text+image+audio+video in, reasoning, tools
- `xiaomi/mimo-v2.6-pro-ultraspeed` — 1,048,576 ctx, 131,072 out,
text+image+audio+video in, reasoning, tools

**➖ 2 removed**
- `anthropic/claude-opus-4`
- `kwaipilot/kat-coder-pro-v2`

**✏️ 5 changed**
- `anthropic/claude-sonnet-4`: max input tokens 1,000,000 → 200,000
- `deepseek/deepseek-v4-pro-0813`: max output tokens 384,000 → 943,718
- `meta-llama/llama-3.1-70b-instruct`: max output tokens 8,192 → 16,384
- `qwen/qwen3-next-80b-a3b-thinking`: max output tokens 32,768 → 235,929
- `z-ai/glm-5.3-flash`: max output tokens 131,072 → 943,718

</details>

<details>
<summary>xai</summary>

**➕ 2 added**
- `grok-4.7` — 500,000 ctx, 500,000 out, text+image+pdf in, reasoning,
tools
- `grok-imagine-image-quality` — 16,000 ctx, text+image+pdf in

</details>

Co-authored-by: mdrxy <61371264+mdrxy@users.noreply.github.com>
 2026-09-22 10:22:17 -04:00
Mason DaughertyMasonopen-swe[bot]
 		9e4ea7a349 fix(fireworks): use current completions model in LLM tests (#40740) 
The first hotfix still selected a model unavailable to the legacy
`/v1/completions` endpoint, so Fireworks release integration tests
continued returning `NOT_FOUND`.

- use `kimi-k2p6`, the newest general-purpose Kimi model that Fireworks
marks Ready and Serverless
- preserve synchronous, asynchronous, streaming, and batch LLM coverage

AI-assisted contribution.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/4abc9cf1-22e0-5ce5-aba2-352047913ba0)
· openai:gpt-5.6-sol (medium)

---------

Co-authored-by: Mason <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-22 00:38:27 -04:00
Mason DaughertyMasonopen-swe[bot]
 		219b372051 fix(deepseek,infra): resolve compatible minimum OpenAI dependencies, bump min ver (#40738) 
- The [DeepSeek release
check](https://github.com/langchain-ai/langchain/actions/runs/35686046493/job/106613011631)
downgraded core to 1.4.7 while retaining local OpenAI 1.6.3, which
requires core 1.6.4. Include `langchain-openai` in release
minimum-version selection so both dependencies are resolved together;
preserve local packages during PR checks.
- Raise DeepSeek's OpenAI minimum to 1.3.1: testing the old 1.1.0 floor
exposes missing model-profile behavior and package-version metadata. Add
regression coverage for release and PR dependency selection.

## Release note
`langchain-deepseek` now requires `langchain-openai>=1.3.1` for model
profiles and package-version metadata.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/ed7f0398-e808-5f5b-9eb4-0521c82fff7f)
· openai:gpt-6-astra (low)

Co-authored-by: Mason <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-22 00:27:19 -04:00
Mason DaughertyMasonopen-swe[bot]
 		b976c82c58 hotfix(fireworks): use available model in LLM tests (#40737) 
The Fireworks release pipeline was blocked because the LLM integration
tests targeted `gpt-oss-20b`, which now returns `NOT_FOUND` from the
completions endpoint.

- replace the unavailable model with the serverless
`llama-v3p3-70b-instruct` model
- restore coverage for synchronous, asynchronous, streaming, and batch
LLM calls

AI-assisted contribution.

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/4abc9cf1-22e0-5ce5-aba2-352047913ba0)
· openai:gpt-5.6-sol (medium)

Co-authored-by: Mason <mdrxy@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-22 00:24:06 -04:00
Mason Daughertyandmdrxy e2eab823c7 release(deepseek): 1.1.1 (#40734) 
Bumps `langchain-deepseek` to 1.1.1, a patch release carrying the
changes merged since 1.1.0.

PRs included in this release:

- `fix(deepseek): route strict mode to the beta endpoint` (#40249)
- `fix(deepseek): map prompt_cache_hit_tokens to cache_read` (#39668)
- `chore(deps): bump anyio from 4.11.0 to 4.14.2 in
/libs/partners/deepseek` (#40641)
- `chore(model-profiles): refresh model profile data` (#40399)

Updates the package version, importable `__version__`, and lockfile
package entry. `uv lock` resolved successfully; its unrelated workspace
dependency churn was intentionally excluded.

This release PR was prepared with AI assistance (Open SWE).

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/c0f86652-eaed-5672-a4bc-94f03c88fe41)
· fireworks:accounts/fireworks/models/glm-5p3-flash (max)

Co-authored-by: mdrxy <mdrxy@users.noreply.github.com>
 2026-09-22 00:10:48 -04:00
Mason Daughertyandmdrxy 0683934fb2 release(fireworks): 1.6.2 (#40735) 
Bumps `langchain-fireworks` to 1.6.2, a patch release carrying
dependency and model profile updates merged since 1.6.1.

PRs included in this release:

- `chore(model-profiles): refresh model profile data` (#40665)
- `chore(deps): bump anyio from 4.11.0 to 4.14.2 in
/libs/partners/fireworks` (#40639)
- `chore(deps): bump urllib3 from 2.7.0 to 2.8.0 in
/libs/partners/fireworks` (#40587)
- `chore(deps): bump langsmith from 0.12.1 to 0.12.6 in
/libs/partners/fireworks` (#40586)
- `chore(deps): bump pygments from 2.20.0 to 2.21.0 in
/libs/partners/fireworks` (#40585)
- `chore(deps): bump idna from 3.19 to 3.20 in /libs/partners/fireworks`
(#40584)
- `chore(model-profiles): refresh model profile data` (#40541)
- `chore(model-profiles): refresh model profile data` (#40500)
- `chore(model-profiles): refresh model profile data` (#40452)
- `chore(model-profiles): refresh model profile data` (#40416)
- `chore(model-profiles): refresh model profile data` (#40399)
- `chore(model-profiles): refresh model profile data` (#40277)
- `chore(model-profiles): refresh model profile data` (#40217)
- `chore(model-profiles): refresh model profile data` (#40198)
- `chore(model-profiles): refresh model profile data` (#40171)
- `chore(model-profiles): refresh model profile data` (#40009)
- `chore(deps): bump orjson from 3.11.6 to 3.12.0 in
/libs/partners/fireworks` (#40129)
- `chore(deps): bump langsmith from 0.10.16 to 0.12.1 in
/libs/partners/fireworks` (#40130)

Updates the package version, importable `__version__`, and lockfile
package entry. `uv lock` resolved successfully; its unrelated workspace
dependency churn was intentionally excluded.

This release PR was prepared with AI assistance (Open SWE).

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/c0f86652-eaed-5672-a4bc-94f03c88fe41)
· fireworks:accounts/fireworks/models/glm-5p3-flash (max)

Co-authored-by: mdrxy <mdrxy@users.noreply.github.com>
 2026-09-22 00:10:41 -04:00
Mason Daughertyandmdrxy 79263361e8 release(openrouter): 0.2.9 (#40736) 
Bumps `langchain-openrouter` to 0.2.9, a patch release carrying
dependency and model profile updates merged since 0.2.8.

PRs included in this release:

- `chore(model-profiles): refresh model profile data` (#40705)
- `chore(model-profiles): refresh model profile data` (#40685)
- `chore(model-profiles): refresh model profile data` (#40665)
- `chore(deps): bump anyio from 4.13.0 to 4.14.2 in
/libs/partners/openrouter` (#40627)
- `chore(model-profiles): refresh model profile data` (#40600)
- `chore(model-profiles): refresh model profile data` (#40541)
- `chore(model-profiles): refresh model profile data` (#40500)
- `chore(model-profiles): refresh model profile data` (#40452)
- `chore(model-profiles): refresh model profile data` (#40436)
- `chore(model-profiles): refresh model profile data` (#40416)
- `chore(model-profiles): refresh model profile data` (#40399)
- `chore(model-profiles): refresh model profile data` (#40358)
- `chore(model-profiles): refresh model profile data` (#40317)
- `chore(model-profiles): refresh model profile data` (#40277)
- `chore(model-profiles): refresh model profile data` (#40258)
- `chore(model-profiles): refresh model profile data` (#40217)
- `chore(model-profiles): refresh model profile data` (#40198)
- `chore(model-profiles): refresh model profile data` (#40171)
- `chore(model-profiles): refresh model profile data` (#40009)
- `chore(model-profiles): refresh model profile data` (#39954)
- `chore(model-profiles): refresh model profile data` (#39928)
- `chore(model-profiles): refresh model profile data` (#39906)
- `chore(model-profiles): refresh model profile data` (#39875)
- `chore(model-profiles): refresh model profile data` (#39844)
- `chore(model-profiles): refresh model profile data` (#39824)
- `chore(model-profiles): refresh model profile data` (#39789)
- `chore(model-profiles): refresh model profile data` (#39751)
- `chore(model-profiles): refresh model profile data` (#39710)
- `chore(model-profiles): refresh model profile data` (#39692)
- `chore(model-profiles): refresh model profile data` (#39670)

Updates the package version, importable `__version__`, and lockfile
package entry. `uv lock` resolved successfully; its unrelated workspace
dependency churn was intentionally excluded.

This release PR was prepared with AI assistance (Open SWE).

Made by [Open SWE](https://github.com/langchain-ai/open-swe) · [view
thread](https://openswe.vercel.app/agents/c0f86652-eaed-5672-a4bc-94f03c88fe41)
· fireworks:accounts/fireworks/models/glm-5p3-flash (max)

Co-authored-by: mdrxy <mdrxy@users.noreply.github.com>
 2026-09-22 00:10:36 -04:00
ccurme 7e89d6c79c release(openai): 1.6.3 (#40719) 2026-09-21 14:23:30 -04:00
ccurme 99d0d06621 release(core): 1.6.4 (#40718) 2026-09-21 11:42:11 -04:00
ccurmeandopen-swe[bot] 17d3d892cd fix(openai): expose inferred Responses API routing at initialization (#40715) 
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-09-21 10:52:15 -04:00