fix(anthropic): preserve optional Responses tool properties

[eugene-yao-zocdoc]

Relevant issues

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

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

Screenshots / Proof of Fix

A live OpenAI Responses API probe using a synthetic search-tool schema reproduced the provider behavior behind this bug:

• When strict was omitted, Responses normalized the tool to strict: true, changed required from ["query"] to ["query", "cursor", "context_channel_id"], and generated empty strings for the optional fields.
• With strict: false, Responses preserved required: ["query"] and generated only {"query":"zz_responses_probe"}.

The patched adapter now emits strict: false for ordinary Anthropic tools while preserving an explicit top-level Anthropic strict: true request. The sibling Chat Completions adapter also keeps an explicit strict value at function level rather than placing it inside the JSON Schema.

After rebasing onto litellm_internal_staging, focused verification against commit 6dbd305f passed:

........................................................................ [ 40%]
........................................................................ [ 80%]
...................................                                      [100%]
179 passed, 1 warning in 3.84s

The edited modules pass py_compile, and git diff --check reports no errors.

Type

🐛 Bug Fix

Changes

• Explicitly disable the OpenAI Responses API's strict-by-default normalization when translating ordinary Anthropic tools.
• Preserve explicitly requested Anthropic strict tool validation across both Responses and Chat Completions routes.
• Add the top-level strict field to the Anthropic tool type.
• Add regressions proving optional properties stay optional, explicit strict validation remains enabled, and strict does not leak into the parameters schema.

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[eugene-yao-zocdoc]

@greptileai

[greptile-apps]

Greptile Summary

This PR preserves Anthropic tool strictness across OpenAI-compatible adapters

• Adds an optional top-level strict field to the Anthropic tool type
• Maps explicit strict validation to Chat Completions function tools
• Disables Responses API strict-by-default behavior for ordinary tools while preserving explicit strict: true
• Adds regression coverage for optional properties and strict-field placement

Confidence Score: 5/5

The PR appears safe to merge

No blocking failure remains

Important Files Changed

Filename	Overview
litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py	Moves an explicit Anthropic strict setting to the OpenAI function-tool level without leaking it into the parameters schema
litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py	Emits explicit Responses function-tool strictness, defaulting ordinary Anthropic tools to non-strict handling
litellm/types/llms/anthropic.py	Extends the Anthropic custom-tool TypedDict with the optional top-level strict field
tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py	Verifies explicit strict validation remains at the Chat Completions function level
tests/test_litellm/llms/anthropic/experimental_pass_through/responses_adapters/test_responses_adapters_transformation.py	Verifies optional properties remain optional and explicit Responses strict validation is preserved

_{Reviews (2): Last reviewed commit: "fix(anthropic): preserve optional Respon..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing eugene-yao-zocdoc:fix/anthropic-responses-optional-tool-schema (6dbd305) with litellm_internal_staging (bb6bb66)}