langchain-ai
langchain
Blog
Docs
Changelog
Blog
Docs
Changelog
Overview
Branches
Benchmarks
Runs
Performance History
Latest Results
fix(langchain): persist context-editing clears via the actual model-call edit Address review feedback on the ContextEditingMiddleware persistence fix: recomputing the edit in after_model with count_tokens_approximately could disagree with wrap_model_call's decision under token_count_method="model" (different counters could clear a different set of messages, or none at all). Instead, wrap_model_call/awrap_model_call now persist the exact messages the edit changed for the real model request via an ExtendedModelResponse + Command update, so the checkpointer's view always matches what the model actually saw, regardless of token_count_method. Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
nm/langchain/context-editing-persist-clears
14 minutes ago
fix(langchain,anthropic): fix batch of near-trivial bugs Closes #34274 Closes #38718 Closes #36409 Closes #35852 Closes #38465 - LLMToolEmulator: remove redundant double-check when building tools_to_emulate, and make model a required keyword argument instead of silently defaulting to an Anthropic model that required langchain-anthropic to be installed. - Re-export PIIMatch from langchain.agents.middleware so custom PII detector authors don't need the private _redaction module. - Add a state_schema parameter to wrap_tool_call, matching the other middleware decorators. - Fix the Claude file-tool dispatch in langchain-anthropic to populate old_path alongside path, fixing a KeyError on every rename command. Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
nm/langchain/fix-batch-bug-reports
14 minutes ago
chore(langchain): export PIIMatch
hunter/export-pii-match
27 minutes ago
fix(langchain): preserve final repeated schema ordering
fix/langchain-last-schema-wins
35 minutes ago
fix(core): handle falsy chat model caches
fix/falsy-llm-cache
37 minutes ago
feat(langchain): filter internal middleware model calls from `messages` projection (#39252) Closes #34139 #34382 `SummarizationMiddleware`, `LLMToolEmulator`, and `LLMToolSelectorMiddleware` make internal model calls for bookkeeping, but these were indistinguishable from the agent's main model call in the event stream. This PR tags internal calls in config metadata and adds an `InternalCallTransformer` that filters them from `run.messages` and the raw event log. Note: If a user wants to stream the internal LLM calls, needs to build a custom transformer around it --------- Co-authored-by: Devbyteai <abud6673@gmail.com>
master
2 hours ago
Merge branch 'master' into nishitha/langchain/filter-internal-model-calls-from-messages
nishitha/langchain/filter-internal-model-calls-from-messages
2 hours ago
chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/langchain_v1 (#39242) Bumps [aiohttp](https://github.com/aio-libs/aiohttp) from 3.14.1 to 3.14.3. Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
master
3 hours ago
Latest Branches
CodSpeed Performance Gauge
0%
fix(langchain): persist `ClearToolUsesEdit` clears via `after_model`
#39265
17 minutes ago
ed4462a
nm/langchain/context-editing-persist-clears
CodSpeed Performance Gauge
0%
fix(langchain,anthropic): fix batch of near-trivial bugs
#39288
18 minutes ago
0854c51
nm/langchain/fix-batch-bug-reports
CodSpeed Performance Gauge
0%
chore(langchain): export PIIMatch
#39287
32 minutes ago
a2a41c9
hunter/export-pii-match
© 2026 CodSpeed Technology
Home
Terms
Privacy
Docs