langchain-ai
langchain
Blog
Docs
Changelog
Blog
Docs
Changelog
Overview
Branches
Benchmarks
Runs
Performance History
Latest Results
lint
imnishitha/core/typeddict-not-required
5 hours ago
nits
imnishitha/core/typeddict-not-required
5 hours ago
test(openai,langchain_v1): normalize function_call status in VCR matcher The status-preservation fix now always includes "status" (defaulting to None) on function_call/function_call_output Responses API items, which broke exact-body VCR cassette matching for cassettes recorded before that change existed. Rather than re-recording every affected cassette (no live API access), drop "status" from these items before comparing request bodies, matching the existing precedent of normalizing known-safe cross-version differences in this matcher. Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
imnishitha/openai/responses-api-function-call-status
5 hours ago
fix(core): respect Required/NotRequired and total=False in TypedDict tools _convert_any_typed_dicts_to_pydantic always marked TypedDict fields as required, ignoring `total=False` and explicit `NotRequired[...]` annotations, and crashed with a TypeError on `Required[...]`/ `NotRequired[...]` fields because it passed them through the generic subscripted-type recursion instead of unwrapping them. This now unwraps `Required`/`NotRequired` and derives each field's requiredness from the TypedDict's own `__required_keys__`, which already accounts for `total` and any per-field overrides. Closes #36490 Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
imnishitha/core/typeddict-not-required
7 hours ago
fix(core): preserve flat tool args schema for `RootModel` runnables (#39307) Closes #38713 Fixes `Runnable.as_tool()` advertising `TypedDict` inputs under a synthetic `root` key. When a runnable's `input_schema` is a `RootModel` (e.g. compiled `StateGraph`s), the tool schema is now built from the runnable's type hints instead, keeping the advertised schema consistent with the flat input the runnable actually expects. --------- Co-authored-by: shusnapx <ashu.kumarexam@gmail.com>
master
8 hours ago
fix(langchain): preserve history on `SummarizationMiddleware` summary failure (#39268) Closes #38867 Fixes `SummarizationMiddleware` replacing conversation history with error text when summary generation fails. Summary calls now retry transient failures up to 3 times via `Runnable.with_retry`. If all attempts fail, the underlying exception propagates and the original history remains untouched. `_create_summary`/`_acreate_summary` no longer convert exceptions into fabricated summaries. No new parameters or exception types are introduced. ### Release note `SummarizationMiddleware` now retries failed summary calls and propagates the underlying error if retries are exhausted, instead of replacing conversation history with error text. Thanks @harshxth for raising this! --------- Signed-off-by: Nishitha M <32355027+imnishitha@users.noreply.github.com>
master
8 hours ago
restore None mapping
cc/pydantic
10 hours ago
more realistic tests
cc/pydantic
10 hours ago
Latest Branches
CodSpeed Performance Gauge
0%
fix(core): fix issues around `Required/NotRequired` in TypedDict tools
#39336
5 hours ago
6cd9372
imnishitha/core/typeddict-not-required
CodSpeed Performance Gauge
0%
fix(openai): preserve `status` on `function_call` blocks
#39310
5 hours ago
00592c9
imnishitha/openai/responses-api-function-call-status
CodSpeed Performance Gauge
0%
fix(core): compat with pydantic 2.14
#39328
10 hours ago
b302857
cc/pydantic
© 2026 CodSpeed Technology
Home
Terms
Privacy
Docs