feat: add OpenAI Responses API route

[maxpetrusenkoagent]

Summary

• add a Responses API request model that maps Responses payloads onto the existing chat workflow contract
• register /v1/responses-compatible routing when openai_api_v1_path ends with /responses
• return Responses API-shaped non-streaming responses and SSE lifecycle events for streaming responses

Fixes #1278

Verification

• uv run --extra test --extra async_endpoints pytest packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py -q
• uv run --extra test --extra async_endpoints ruff check packages/nvidia_nat_core/src/nat/data_models/api_server.py packages/nvidia_nat_core/src/nat/front_ends/fastapi/routes/chat.py packages/nvidia_nat_core/src/nat/front_ends/fastapi/routes/v1_responses.py packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py

[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.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: a375f250-4a23-432e-b76a-526922170b07

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[maxpetrusenkoagent]

Review-fix sweep update:

Summary:

• Added DCO sign-off to the PR commit.
• Fixed Responses API structured input conversion so untyped message items preserve content.
• Preserved structured message roles when converting Responses API input into the existing ChatRequest workflow contract.
• Added regression coverage for both cases.

Verification:

• uv run --extra test --extra async_endpoints pytest packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py::test_responses_request_preserves_structured_message_roles packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py::test_responses_request_converts_untyped_message_input_to_chat_request packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py::test_responses_request_converts_to_chat_request -q -> 3 passed
• uv run --extra test --extra async_endpoints pytest packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py -q -> 17 passed
• uv run --extra test --extra async_endpoints ruff check packages/nvidia_nat_core/src/nat/data_models/api_server.py packages/nvidia_nat_core/src/nat/front_ends/fastapi/routes/chat.py packages/nvidia_nat_core/src/nat/front_ends/fastapi/routes/v1_responses.py packages/nvidia_nat_core/tests/nat/front_ends/fastapi/test_openai_compatibility.py -> All checks passed
• Independent final autoreview on the exact updated diff -> no blocking findings

Remaining blocker I could not clear from the contributor account:

• Label Checker reports missing category/breaking labels. I attempted to add Feature and non-breaking, but GitHub returned: maxpetrusenkoagent does not have the correct permissions to execute AddLabelsToLabelable.

Confidence: high on the code/check fixes. Label checker needs a maintainer/vetter to apply the required labels.

[maxpetrusenkoagent]

Review-fix sweep found no actionable code blockers from the contributor side.

Current blockers appear repository-maintainer gated:

• Label Checker: missing category/breaking labels. I retried adding Feature + non-breaking; GitHub returned maxpetrusenkoagent does not have the correct permissions to execute AddLabelsToLabelable.
• PR remains draft / review-required, with no maintainer review yet.

No new code changes made in this sweep.

[maxpetrusenkoagent]

Review-Fix Summary

Applied 3 targeted fixes to the OpenAI Responses API route (v1_responses.py, chat.py):

Fix	Change	Files
Interactive extension flag	enable_interactive threaded through add_v1_responses_route → post_responses_api_endpoint → _responses_stream call chain	v1_responses.py, chat.py
Status code	WORKFLOW_ERROR handler now returns 500 (server error) instead of 422 (input validation failure) — matches OpenAI API contract and the RESPONSE_500 status already registered on the route	v1_responses.py
Formatting	StreamingResponse call re-wrapped to pass ruff E501 (line-length)	v1_responses.py

Pre-existing safe code confirmed (no changes needed):

• response.choices guard at v1_responses.py:73 — already present
• output_item/content_part initialized before the stream loop — already present

---

Verification

• Tests: pytest test_openai_compatibility.py — 19/19 passed
• Lint: ruff check on all touched files — All checks passed

---

Second-Agent Review (hermes chat -Q / MiniMax-M2.7)

Verdict: CLEAN (1 minor non-blocking note)

Minor note (non-blocking): MCPOAuth2ProviderConfig.preflight_auth: Literal[False] type override from base bool — code smell, MCP auth path no-ops regardless, no functional breakage.

No blocking findings. The type mismatch in MCPOAuth2ProviderConfig.preflight_auth is a code-smell but does not cause functional breakage in practice, as the MCP authenticate() path will no-op regardless. The Responses API, preflight auth, and list-content flatten changes are all properly implemented with reasonable test coverage.

---

Confidence: High. All blocking findings from the prior review are resolved. Tests pass, lint clean, second-agent confirms CLEAN.

[maxpetrusenko]

Label Checker is failing because this PR is missing category and breaking-change labels. Suggested labels for this change: feature request and non-breaking.