fix web search tool cost

[mubashir1osmani]

Relevant issues

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Apr 2, 2026 9:59pm

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing mubashir1osmani:fix_web_search_pricing (7280721) with main (9c5fda4)}

[greptile-apps]

Greptile Summary

This PR fixes a gap in Anthropic web search cost tracking: when a real ModelResponse was returned from an Anthropic completion that used the web search tool, the cost was silently dropped to zero because the ModelResponse branch of response_object_includes_web_search_call only checked for url_citation message annotations and usage.prompt_tokens_details.web_search_requests — it never checked the Anthropic-specific usage.server_tool_use.web_search_requests field.

Key changes:

• tool_call_cost_tracking.py: Adds a server_tool_use.web_search_requests check inside the isinstance(response_object, ModelResponse) branch, mirroring the identical check that already existed in the elif usage is not None: fallback branch (used when no ModelResponse is present).
• test_tool_call_cost_tracking.py: Adds a focused regression test (test_get_cost_for_anthropic_web_search_with_model_response) that constructs a ModelResponse + Usage object with server_tool_use.web_search_requests=2, asserts the detection returns True, and asserts the resulting cost is non-zero. Only local/mock objects are used — no real network calls.

Confidence Score: 5/5

• Safe to merge — a small, targeted bug fix with a clear regression test and no side effects.
• The change is minimal (6 lines of production code, a comment update, and one new test). It mirrors an identical check that already existed in the fallback branch, uses a model that is present in the cost map, and the new test exercises the exact previously-broken path without making real network calls. No custom rules are violated.
• No files require special attention.

Important Files Changed

Filename	Overview
litellm/litellm_core_utils/llm_cost_calc/tool_call_cost_tracking.py	Adds usage.server_tool_use.web_search_requests detection to the ModelResponse branch of response_object_includes_web_search_call, fixing a gap where Anthropic web search costs were missed when a real ModelResponse object was passed.
tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py	Adds a regression test (test_get_cost_for_anthropic_web_search_with_model_response) that constructs a real ModelResponse with usage.server_tool_use.web_search_requests=2 and verifies both detection and non-zero cost; uses only local objects, no network calls.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[response_object_includes_web_search_call] --> B{isinstance ModelResponse?}
    B -->|Yes| C{has url_citation\nannotations?}
    C -->|Yes| D[return True]
    C -->|No| E{usage is not None?}
    E -->|Yes| F{usage.prompt_tokens_details\n.web_search_requests?}
    F -->|Yes| G[return True]
    F -->|No| H{usage.server_tool_use\n.web_search_requests?\n🆕 NEW CHECK}
    H -->|Yes| I[return True]
    H -->|No| J[return False]
    E -->|No| J
    B -->|No, ResponsesAPIResponse| K{has web_search_call\nin output?}
    K -->|Yes| D
    K -->|No| J
    B -->|No, other| L{usage.server_tool_use\n.web_search_requests?\nexisting check}
    L -->|Yes| D
    L -->|No| M{usage.prompt_tokens_details\n.web_search_requests?}
    M -->|Yes| D
    M -->|No| J

    style H fill:#90EE90,stroke:#006400
    style I fill:#90EE90,stroke:#006400

Loading

_{Reviews (1): Last reviewed commit: "fix web search tool cost" | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[mubashir1osmani]

fixed in #25052