[chelsealong]

What does this PR do?

Fixes langfuse/langfuse#19452.

Map Anthropic’s 5-minute and 1-hour cache writes to their usage buckets and retain only unattributed writes in cache_creation_input_tokens. This prevents the aggregate and tier counts from counting the same tokens twice.

The reported example now totals 5,701 tokens instead of 11,245. Regression tests cover that example, an unattributed remainder, missing aggregate, and zero-value tiers.

Prepared with Codex assistance.

Type of change

• Bug fix

Verification

uv run --frozen pytest -q tests/unit/test_parse_usage_model.py tests/unit/test_langchain.py # 39 passed
uv run --frozen ruff check . # passed
uv run --frozen ruff format --check langfuse/langchain/CallbackHandler.py tests/unit/test_parse_usage_model.py # passed
uv run --frozen mypy langfuse --no-error-summary # passed
git diff --check # passed

Checklist

• I added or updated tests for behavior changes.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

Confidence Score: 5/5

The PR appears safe to merge; no actionable issues were found.

Summary

Maps Anthropic’s cache writes to separate 5-minute and 1-hour buckets. Only writes not covered by those buckets remain in cache_creation_input_tokens.

• LangChain usage records separate Anthropic cache tiers without repeating their tokens.

_{Reviews (1) · Last reviewed commit: "fix(langchain): avoid double counting An..." · Reviewed by Greptile}

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}