fix(langchain): anthropic cache token count #2414
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dosubot
bot
added
the
size:XL
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
...eninference-instrumentation-langchain/src/openinference/instrumentation/langchain/_tracer.py
Show resolved
Hide resolved
dosubot
bot
added
size:L
caroger
reviewed
Nov 10, 2025
| "cache_read_input_tokens" in obj | ||
| and isinstance(obj["cache_read_input_tokens"], int) | ||
| or "cache_creation_input_tokens" in obj | ||
| and isinstance(obj["cache_creation_input_tokens"], int) |
Collaborator
There was a problem hiding this comment.
nit: Add explicit parentheses in the type guard for clarity
return (
"input_tokens" in obj
and "output_tokens" in obj
and isinstance(obj["input_tokens"], int)
and isinstance(obj["output_tokens"], int)
and (
("cache_read_input_tokens" in obj and isinstance(obj["cache_read_input_tokens"], int))
or ("cache_creation_input_tokens" in obj and isinstance(obj["cache_creation_input_tokens"], int))
)
)
caroger
approved these changes
Nov 10, 2025
| ) | ||
| ): | ||
| return | ||
| keys: Sequence[str] |
There was a problem hiding this comment.
Bug: Anthropic Usage Metrics Disappear
Removing input_tokens and output_tokens from the generic token count extraction breaks basic Anthropic responses that don't include cache tokens or total_tokens. Previously, responses with just input_tokens and output_tokens (standard Anthropic format without caching) would be captured by the first loop. Now they're only handled by _is_raw_anthropic_usage_with_cache_read_or_write, which requires cache tokens to be present, causing basic Anthropic token counts to be lost.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
resolves #2381
Note
Improve token usage extraction to handle Anthropic cache read/write and LangChain UsageMetadata (with Bedrock heuristics), update deps, and add targeted tests.
_tracer.py):_is_lc_usage_metadata/_token_counts_from_lc_usage_metadatato maplangchain_core.messages.ai.UsageMetadata, including audio/reasoning and cache details with Bedrock-specific heuristics._is_raw_anthropic_usage_with_cache_read_or_write/_token_counts_from_raw_anthropic_usage_with_cache_read_or_writeto handle Anthropiccache_read_input_tokens/cache_creation_input_tokensand emit detailed cache attributes._token_counts; remove previous key-based Anthropic handling.TypedDict,TypeGuard,UsageMetadata).tests/test_token_counts.pycovering LC usage metadata and Anthropic cache scenarios.tests/test_instrumentor.pyAnthropic expectations (LLM_TOKEN_COUNT_PROMPTfrom22to33).langchain_coreto>= 0.3.9(instruments) and== 0.3.9(type-check).tox.ini: adduv pip list -vincommands_pre.Written by Cursor Bugbot for commit 30a6e95. This will update automatically on new commits. Configure here.