feat: support reasoning, text.format, and service_tier in /v1/responses API

[vvagias]

Summary

• Remove 501 "Not Implemented" rejections for reasoning, text, and service_tier parameters in the /v1/responses endpoint
• Map reasoning.effort → reasoning_effort on the internal chat completions request
• Map text.format → response_format (text / json_object / json_schema) on the internal chat completions request
• Map service_tier → service_tier on the internal chat completions request
• Echo all three parameters back in the response object per the OpenAI spec
• Fix OutputMessage deserialization: make id and status fields #[serde(default)] so clients can send assistant messages in input without these fields
• Add 11 unit tests covering all new mappings and echo-back behavior

Motivation

The /v1/responses API currently rejects several valid OpenAI parameters with HTTP 501, even though the underlying chat completions backend already supports them. This PR removes those artificial gates and wires up the translation layer.

Test plan

• cargo test -p dynamo-llm — all new and existing tests pass
• cargo clippy -p dynamo-llm — no warnings
• Deployed to staging with TRT-LLM backend (Nemotron Nano 30B FP8) and verified:
  • reasoning.effort accepted and echoed back
  • text.format accepted and echoed back
  • service_tier accepted and echoed back
  • OutputMessage deserialization works for assistant messages without id/status

Summary by CodeRabbit

Release Notes

• New Features

  • Added support for OpenAI Responses API fields: reasoning, text formatting options, and service tier configuration.

• Improvements

  • Enhanced message handling flexibility with sensible defaults, enabling reuse of assistant messages without explicit metadata.
  • Improved API compatibility between Responses and Chat Completions endpoints.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

👋 Hi vvagias! Thank you for contributing to ai-dynamo/dynamo.

Just a reminder: The NVIDIA Test Github Validation CI runs an essential subset of the testing framework to quickly catch errors.Your PR reviewers may elect to test the changes comprehensively before approving your changes.

🚀

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between a35df0f and 0b7b937.

📒 Files selected for processing (3)

• lib/async-openai/src/types/responses/response.rs
• lib/llm/src/http/service/openai.rs
• lib/llm/src/protocols/openai/responses/mod.rs

---

Walkthrough

The changes extend OpenAI API response handling by adding support for reasoning, text, and service_tier fields in the Responses API. Default behaviors are introduced for OutputStatus and OutputMessage deserialization, while conversion logic maps these new fields between the Responses and Chat Completions API formats.

Changes

Cohort / File(s)	Summary
Response Type Defaults lib/async-openai/src/types/responses/response.rs	Added Default derivation to OutputStatus with Completed as default variant. Annotated OutputMessage fields (id, status) with #[serde(default)] to improve deserialization flexibility. Added documentation explaining default field behavior.
OpenAI Service Integration lib/llm/src/http/service/openai.rs, lib/llm/src/protocols/openai/responses/mod.rs	Added reasoning, text, and service_tier fields to ResponseParams struct and extended request/response plumbing to propagate these fields between Responses and Chat Completions API formats. Introduced conversion helpers for text format and service tier mapping. Removed validation checks blocking these fields. Added comprehensive tests for new field mappings and echo behavior.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

🐰 Defaults bloom and fields align,
Responses flow through the service line,
Reasoning, text, and tiers now play,
Conversions hop in every way! 🌟

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'feat: support reasoning, text.format, and service_tier in /v1/responses API' clearly and concisely describes the main feature addition across all three changed files.
Description check	✅ Passed	The description comprehensively covers the Summary, Motivation, and Test plan sections with detailed technical changes and verification details matching the template's requirements.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

---

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 and usage tips.}

[vvagias]

Superseded by a new PR that stacks on top of #6599 to avoid conflicts.