fix(anthropic): include cache tokens in total usage

[gambletan]

Summary

• include Anthropic cache-read and cache-creation tokens in provider-level total usage
• preserve the cache-inclusive total through shared UsageMetrics normalization and Flow aggregation
• keep OpenAI-style cached prompt accounting unchanged by applying the extra counters only to the Anthropic input_tokens shape

Problem

Anthropic reports cache reads and cache writes separately from input_tokens. CrewAI exposed those counters but calculated total_tokens from input_tokens + output_tokens only, substantially undercounting cached workloads. The shared UsageMetrics normalizer then recomputed the same incomplete total, so public LLM summaries and Flow metrics were affected as well.

Testing

• 76 Anthropic provider and Flow usage tests passed
• 29 usage-shape and LLM usage-event tests passed
• ruff check passed for all changed files
• ruff format --check passed for changed source files
• git diff --check passed
• two independent Codex review passes; the first identified the shared normalizer gap, which is covered by the final tests

Fixes #6768

[gambletan]

Closing as a duplicate of #6770, which was opened first and covers the same issue and end-to-end UsageMetrics path. This PR was created after a final duplicate-check/create command did not abort when the issue timeline returned #6770. Thanks @Anai-Guo for the thorough fix.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: ca5349df-3a00-49ed-bce9-3b573c43437e

📥 Commits

Reviewing files that changed from the base of the PR and between c8f441c and 7d5bbd0.

📒 Files selected for processing (4)

• lib/crewai/src/crewai/llms/providers/anthropic/completion.py
• lib/crewai/src/crewai/types/usage_metrics.py
• lib/crewai/tests/llms/anthropic/test_anthropic.py
• lib/crewai/tests/test_flow_usage_metrics.py

---

📝 Walkthrough

Walkthrough

Changes

Anthropic usage accounting

Layer / File(s)	Summary
Include cache tokens in provider totals lib/crewai/src/crewai/llms/providers/anthropic/completion.py, lib/crewai/src/crewai/types/usage_metrics.py	Anthropic totals now include input, output, cache-read, and cache-creation tokens. Normalized usage preserves the cache counters.
Validate cache token accounting lib/crewai/tests/llms/anthropic/test_anthropic.py, lib/crewai/tests/test_flow_usage_metrics.py	Anthropic fixtures include cache fields. Tests verify extraction, aggregate totals, and normalized usage metrics.

Suggested reviewers: lorenzejay

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}