RSPEED-2402: Enable MCP tool use on rlsapi /v1/infer endpoint

[major]

Description

RSPEED-2402

Allow the rlsapi v1 (/v1/infer) endpoint to use MCP servers transparently.
The client sends a plain question and gets a plain text answer, but behind the
scenes the LLM can call MCP tools (e.g., a RAG knowledge base) to enrich its
response.

The query v2 endpoint already supports MCP. The rlsapi v1 endpoint uses the
same underlying responses.create() API — it just never passed the tools
parameter. This change closes that gap by reusing get_mcp_tools() from
query v2.

Type of change

• Refactor
• New feature
• Bug fix
• CVE fix
• Optimization
• Documentation Update
• Configuration Update
• Bump-up service version
• Bump-up dependent library
• Bump-up library or tool used for development (does not change the final image)
• CI configuration change
• Konflux configuration change
• Unit tests improvement
• Integration tests improvement
• End to end tests improvement
• Benchmarks improvement

Tools used to create PR

• Assisted-by: Claude Opus 4
• Generated by: Claude Opus 4

Related Tickets & Documents

• RSPEED-2402 (parent epic: RSPEED-2229)

Checklist before requesting a review

• I have performed a self-review of my code.
• PR has passed all pre-merge test jobs.
• If it is a core feature, I have added thorough tests.

Testing

1. Configure mcp_servers in lightspeed-stack.yaml:

   mcp_servers:
     - name: "rag-knowledge-base"
       url: "http://rag-mcp-server:9000/sse"

2. Send a POST to /v1/infer with a question that benefits from tool use
3. Verify the response contains information enriched by the MCP tool
4. Send a POST to /v1/infer with no mcp_servers configured — verify
   identical behavior to before this change (regression test)

Summary by CodeRabbit

Release Notes

• New Features
  • MCP tool support is now integrated into the RLS API v1. Tools from configured MCP servers are automatically available during response generation.
  • The system seamlessly handles tool passthrough to the language model, enabling richer interactions with available tools from your server configuration.

[coderabbitai]

Walkthrough

This pull request adds MCP (Model Context Protocol) tool support to the RLS API v1 retrieve flow. The retrieve_simple_response function is updated to accept an optional tools parameter, which is passed to the Llama Stack client. The infer_endpoint function is modified to compute MCP tools from configuration and forward them through the retrieve flow.

Changes

Cohort / File(s)	Summary
Core API Endpoint src/app/endpoints/rlsapi_v1.py	Updated retrieve_simple_response signature to accept optional tools parameter and wire tools into Llama Stack client calls. Modified infer_endpoint to compute MCP tools via get_mcp_tools and pass them to retrieve_simple_response.
Integration Tests tests/integration/endpoints/test_rlsapi_v1_integration.py	Added two integration tests: one verifying empty tools list passthrough when no MCP servers are configured, and another verifying MCP tool definitions are forwarded to the LLM responses endpoint.
Unit Tests tests/unit/app/endpoints/test_rlsapi_v1.py	Added helper function _setup_responses_mock_with_capture and three unit tests covering tool parameter passthrough in retrieve_simple_response, default empty tools behavior, and infer_endpoint calling get_mcp_tools.

Sequence Diagram(s)

sequenceDiagram
    participant Client
    participant infer_endpoint
    participant get_mcp_tools
    participant retrieve_simple_response
    participant LlamaStackClient as Llama Stack Client

    Client->>infer_endpoint: POST /infer<br/>(question, instructions)
    infer_endpoint->>get_mcp_tools: Call with<br/>configuration.mcp_servers
    get_mcp_tools-->>infer_endpoint: Return tools list
    infer_endpoint->>retrieve_simple_response: Call with<br/>(question, instructions, tools)
    retrieve_simple_response->>LlamaStackClient: client.responses.create<br/>(..., tools=tools)
    LlamaStackClient-->>retrieve_simple_response: Response with<br/>generated text
    retrieve_simple_response-->>infer_endpoint: Return response text
    infer_endpoint-->>Client: Return response

Loading

Estimated Code Review Effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly Related PRs

• feat(rlsapi): add v1 /infer endpoint module #902 — Directly extends the /infer endpoint functionality introduced in that PR by adding MCP tool passthrough capability.
• feat(rlsapi): implement LLM integration for v1 /infer endpoint #916 — Modifies the same retrieve_simple_response function and Llama Stack wiring that this PR extends with MCP tools support.

Suggested Labels

ok-to-test

Suggested Reviewers

• tisnik

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main change: enabling MCP tool use on the rlsapi /v1/infer endpoint, which aligns perfectly with the primary objective of the pull request.
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.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

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.}

[tisnik]

LGTM, nice work