fix(translation): preserve native Anthropic thinking settings

[grahamking]

AnthropicMessagesCodec::encode_request (crates/switchyard-translation/src/codecs/anthropic/buffered.rs:294-308) now re-emits the caller's native thinking object on the reconstructed path, and only falls back to {"type": "adaptive"} when effort is set and no native thinking exists.

Previously, any reasoning.effort clobbered thinking with adaptive, so reconstruction disagreed with verbatim replay (a caller sending thinking: disabled + output_config.effort replayed as disabled but reconstructed as adaptive).

GLM 5.3's review:

Real bug, correctly fixed. README claims match the code, including omit-before-merge ordering and reasoning_effort only applying to OpenAI backends.

Fixes: https://linear.app/nvidia/issue/SWITCH-1490/

Assisted-by: Pi:GPT 6 Astra medium
Reviewed-by: Pi:GLM 5.3 high
Signed-off-by: Graham King grahamk@nvidia.com

Summary by CodeRabbit

• Bug Fixes
  • Anthropic requests now preserve native thinking settings, including disabled, enabled, and adaptive modes, when translated for a new model or target prompt.
  • Reasoning effort is kept separate from native thinking settings and selects adaptive thinking only when no native setting is provided.
  • OpenAI targets can apply their configured reasoning effort when preparing requests.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: NVIDIA-NeMo/Switchyard/.coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 3217fa9e-62b0-4630-8cf7-22d33cf99f3a

📥 Commits

Reviewing files that changed from the base of the PR and between fbabf51 and c1aba3d.

📒 Files selected for processing (3)

• crates/libsy-llm-client/README.md
• crates/switchyard-translation/src/codecs/anthropic/buffered.rs
• crates/switchyard-translation/tests/request_translation.rs

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

Walkthrough

Anthropic request encoding now preserves native thinking settings separately from output_config.effort. A reconstruction test covers disabled, enabled, and adaptive thinking. The README documents field replacement behavior and target-specific effort overrides.

Changes

Anthropic thinking reconstruction

Layer / File(s)	Summary
Preserve thinking during reconstruction crates/switchyard-translation/src/codecs/anthropic/buffered.rs, crates/switchyard-translation/tests/request_translation.rs, crates/libsy-llm-client/README.md	Anthropic encoding copies raw reasoning to thinking when present. It adds adaptive thinking only when thinking is absent and keeps effort in output_config.effort. The test covers reconstruction across disabled, enabled, and adaptive settings. The README describes field replacement and target effort overrides.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Merge Risk: ⚪ Minimal · up to c1aba

The change preserves native Anthropic thinking settings while keeping output effort separate. No merge-blocking issue was identified; merge after normal checks pass.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 75.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 4 functions across 2 files. (1 skipped: 1…	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: preserving native Anthropic thinking settings during translation.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

Docstring coverage is 75.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 4 functions across 2 files. (1 skipped: 1 unsupported.)

• Fix all pre-merge checks with AI

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Autopilot is currently an internal CodeRabbit preview.

---

_{A rabbit checks the thinking field,
Native settings stay revealed.
Effort keeps its separate place,
Three test cases trace the case.
Hop, the request returns in form!}

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