perf: use module-scoped fixture for RAG tests

[planetf1]

Optimize RAG Test Performance

Type of PR

• Bug Fix
• New Feature
• Documentation
• Other (Performance Optimization)

Description

• Link to Issue: tests: speed up RAG tests #336 (discovered during feat: add pytest markers for test categorization (#322) #326)

This PR optimizes the RAG intrinsic test suite by changing the backend fixture from function-scoped to module-scoped. The backend fixture loads a 3B parameter model (granite-4.0-micro) which was being reloaded for each of the 7 tests in the file, causing significant overhead (a big proportion of the overal tests runnable on this system as per #326)

Fixes #336

Performance Impact:

• Before: 17 minutes
• After: 3.5 minutes
• Improvement: ~5x faster

Safety Justification:

1. Tests remain isolated: All 7 tests perform read-only operations on the backend
2. No state mutation: Each test uses different intrinsics (different adapters)
3. Cleanup preserved: Teardown code still executes once at module end
4. Assertions unchanged: All test logic and expected outputs remain identical
5. Verified: All 7 tests pass with same results

The change is a single line addition (scope="module") to the fixture decorator, so should be a low-risk optimization with significant benefits.

Testing

• Tests added to the respective file if code was changed (N/A - no new tests needed)
• New code has 100% coverage if code was added (N/A - only fixture scope changed)
• Ensure existing tests and github automation passes
  • All 7 RAG tests pass: uv run pytest test/stdlib/components/intrinsic/test_rag.py -v
  • Pre-commit hooks pass (ruff, mypy, codespell)
  • Test execution verified:

NOTE: Testing was done on a base of #326 , but have created a clean delta here - will rebase/retest once that merged

[github-actions]

The PR description has been updated. Please fill out the template for your PR to be reviewed.

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert|release)(?:\(.+\))?:

[planetf1]

Rebased and retested with the pytest markers fix (now in main)

• overall time down from ~20 mins to ~6 mins - some tests still a bit of a stretch for 32GB, but getting better.

=================== 218 passed, 59 skipped, 2 xpassed, 104 warnings in 367.86s (0:06:07) ====================

for full run.