RSPEED-2408: fix(rlsapi): use customization.system_prompt in /v1/infer

[major]

Description

The /v1/infer endpoint hardcodes constants.DEFAULT_SYSTEM_PROMPT ("You are a helpful assistant") in _build_instructions(). The /v1/query and /v1/streaming_query endpoints already support configurable system prompts via customization.system_prompt in lightspeed-stack.yaml, but /v1/infer ignores this config entirely.

This change makes _build_instructions() check configuration.customization.system_prompt first, falling back to DEFAULT_SYSTEM_PROMPT when unset. Follows the same precedence pattern as get_system_prompt() in src/utils/prompts.py, simplified for rlsapi (no per-request prompt, no custom profiles).

Type of change

• Bug fix
• Unit tests improvement

Tools used to create PR

• Assisted-by: Claude (claude-opus-4-6)
• Generated by: N/A

Related Tickets & Documents

• Related Issue: RSPEED-2408
• 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

• uv run pytest tests/unit/app/endpoints/test_rlsapi_v1.py -v -k "build_instructions" — all 6 tests pass (3 existing + 3 new)
• uv run make verify — all linters clean
• New test cases:
  • test_build_instructions_customization_system_prompt_set — custom prompt used when configured
  • test_build_instructions_customization_system_prompt_none — falls back to DEFAULT_SYSTEM_PROMPT when system_prompt is None
  • test_build_instructions_no_customization — falls back when customization itself is None

Summary by CodeRabbit

• New Features

  • System prompts can now be customized via configuration; the app uses a provided custom prompt or falls back to the default when none is set.

• Tests

  • Added tests verifying custom prompt usage and correct fallback to the default prompt.

[coderabbitai]

Walkthrough

_build_instructions now uses configuration.customization.system_prompt when present; otherwise it falls back to constants.DEFAULT_SYSTEM_PROMPT. Tests were added to verify both the custom-prompt path and the fallback when customization is absent.

Changes

Cohort / File(s)	Summary
System Prompt Configuration src/app/endpoints/rlsapi_v1.py	Modify _build_instructions to select configuration.customization.system_prompt if available, else use constants.DEFAULT_SYSTEM_PROMPT.
Prompt Configuration Tests tests/unit/app/endpoints/test_rlsapi_v1.py	Add parametrized test verifying custom system_prompt is used and contains OS info; add fallback test verifying default prompt when customization is None.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• LCORE-574: unit tests for customization #504: Adds/implements the Customization model and resources for system_prompt, which is directly related to the configurable prompt usage validated by these tests.

🚥 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: implementing support for customization.system_prompt in the /v1/infer endpoint.
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.}

[major]

The e2e failure looks unrelated 🤔

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

tests/unit/app/endpoints/test_rlsapi_v1.py (1)

155-167: ⚠️ Potential issue | 🟠 Major

test_build_instructions lacks mocking for configuration — tests will fail.

The existing test at lines 155–167 has no mocker parameter and doesn't patch configuration, but _build_instructions now accesses configuration.customization at runtime. This will hit the uninitialized singleton and raise a LogicError.

The two new tests correctly mock configuration using mocker.patch(). The existing parametrized test should either:

• Accept mocker: MockerFixture and patch configuration before calling _build_instructions, or
• Use the mock_configuration fixture already defined in this test file (line 76)

🧹 Nitpick comments (1)

src/app/endpoints/rlsapi_v1.py (1)

95-101: Consider caching the configuration.customization access in a local variable.

configuration.customization is accessed twice, and each access goes through the property getter (which includes a None-check on _configuration). A local variable avoids the redundant property call and reads more cleanly.

♻️ Suggested refactor

-    if (
-        configuration.customization is not None
-        and configuration.customization.system_prompt is not None
-    ):
-        base_prompt = configuration.customization.system_prompt
-    else:
-        base_prompt = constants.DEFAULT_SYSTEM_PROMPT
+    customization = configuration.customization
+    if customization is not None and customization.system_prompt is not None:
+        base_prompt = customization.system_prompt
+    else:
+        base_prompt = constants.DEFAULT_SYSTEM_PROMPT

[tisnik]

LGTM