#40518 UsageMetadataCallbackHandler silently drops usage_metadata when model_name is missing

Submission checklist

This is a bug, not a usage question.
I added a clear and descriptive title that summarizes this issue.
I used the GitHub search to find a similar question and didn't find it.
I am sure that this is a bug in LangChain rather than my code.
The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
This is not related to the langchain-community package.
I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

Package (Required)

langchain
langchain-openai
langchain-anthropic
langchain-classic
langchain-core
langchain-model-profiles
langchain-tests
langchain-text-splitters
langchain-chroma
langchain-deepseek
langchain-exa
langchain-fireworks
langchain-groq
langchain-huggingface
langchain-mistralai
langchain-nomic
langchain-ollama
langchain-openrouter
langchain-perplexity
langchain-qdrant
langchain-xai
Other / not sure / general

Related Issues / PRs

PR #40513 (fix, ready — references this issue once it's reopened)

Reproduction Steps / Example Code (Python)

from langchain_core.callbacks import get_usage_metadata_callback
from langchain_core.language_models import GenericFakeChatModel
from langchain_core.messages import AIMessage
from langchain_core.messages.ai import UsageMetadata

usage = UsageMetadata(input_tokens=1, output_tokens=2, total_tokens=3)
llm = GenericFakeChatModel(messages=iter([AIMessage("hi", usage_metadata=usage)]))

with get_usage_metadata_callback() as cb:
    llm.invoke("hello")
    print(cb.usage_metadata)  # {} -- usage silently lost

Error Message and Stack Trace (if applicable)

Description

UsageMetadataCallbackHandler.on_llm_end (in langchain_core/callbacks/usage.py) only records token usage when the response's response_metadata also contains a model_name key:

if usage_metadata and model_name: ...

Any chat model whose AIMessage carries real usage_metadata but doesn't set response_metadata["model_name"] (custom BaseChatModel subclasses, or some third-party integrations) has its usage silently dropped — callback.usage_metadata stays {} with no error or warning. Since the entire purpose of this callback/context manager is usage tracking, this defeats it in a way that's easy to miss.

Expected behavior: Usage should still be tracked when model_name is absent, e.g. bucketed under an "unknown" key, rather than dropped entirely.

Proposed fix: Change the condition to only require usage_metadata, and key the dict with model_name or "unknown". This is backward compatible: models that do set model_name (the common case) are unaffected.

I have a fix with a regression test ready — PR #40513.

System Info

System Information

OS: Darwin OS Version: Darwin Kernel Version 27.0.0 Python Version: 3.14.6 (main, Jun 10 2026, 10:03:53) [Clang 21.0.0]

Package Information

langchain_core: 1.6.1 langsmith: 0.11.1 langchain_protocol: 0.0.18 langchain_tests: 1.1.9 langchain_text_splitters: 1.1.2

Social handles (optional)

No response