ci: add coverage measurement and gate to backend-test

[lucastononro]

Summary

• backend-test had no coverage visibility at all — pytest tests/ -v --tb=short ran with no --cov, and pytest-cov wasn't even a dependency.
• This adds coverage measurement, a terminal report, an uploaded XML artifact, and a fail-under floor so future coverage regressions in untested modules (e.g. runner, sandbox) can't merge silently.

Changes

• backend/requirements.txt: add pytest-cov>=6.0.0.
• .github/workflows/ci.yml (backend-test): run pytest tests/ -v --tb=short --cov=. --cov-report=xml --cov-report=term-missing --cov-fail-under=60, then upload backend/coverage.xml via actions/upload-artifact@v4 (if: always() so failed runs still publish the report).
• Floor is set at 60%, a few points below the measured baseline (65.81% total, 296 passed / 8 skipped) — enough margin to avoid flaking on incidental drift while still catching real regressions.

Test plan

Existing

• Ran the full backend suite locally in a fresh venv (Python 3.12, deps from requirements.txt): 296 passed, 8 skipped, unchanged from current main.

New

• Ran the exact new CI command locally: pytest tests/ -v --tb=short --cov=. --cov-report=xml --cov-report=term-missing --cov-fail-under=60 → passes with Required test coverage of 60% reached. Total coverage: 65.81%.

Closes #116

🤖 Generated with Claude Code

Greptile Summary

This PR adds pytest-cov coverage measurement and a fail-under gate to the backend-test CI job, with test files excluded from the metric via a new .coveragerc so the floor reflects untested production code only.

• .github/workflows/ci.yml: Expands the pytest invocation with --cov=., XML + terminal reports, and a --cov-fail-under=50 gate; uploads backend/coverage.xml as an artifact on every run (including failures) via actions/upload-artifact@v4.
• backend/.coveragerc: New config that omits tests/*, keeping the coverage percentage focused on application code (~51% baseline vs ~66% when tests were included).
• backend/requirements.txt: Adds pytest-cov>=6.0.0 as a test dependency.

Confidence Score: 5/5

This is a safe, additive CI change with no modifications to application logic.

The change only adds coverage instrumentation to the test job. The .coveragerc omit pattern is appropriate for the flat tests/ structure, the artifact upload path is correct relative to the workspace root, and the 50% floor gives a few points of headroom above the measured ~51% production-only baseline so the gate won't flake on minor drift.

No files require special attention.

Important Files Changed

Filename	Overview
.github/workflows/ci.yml	Adds --cov flags, coverage fail-under=50 gate, and artifact upload to backend-test job; artifact path backend/coverage.xml is correct relative to workspace root
backend/.coveragerc	New file; omits tests/* from coverage measurement — pattern is correct since all test files are flat under backend/tests/
backend/requirements.txt	Adds pytest-cov>=6.0.0 as a test dependency; straightforward addition

_{Reviews (2): Last reviewed commit: "fix(review): address Greptile findings o..." | Re-trigger Greptile}

[greptile-apps]

Artifact name not unique across concurrent runs

name: backend-coverage is a fixed string. If two PRs or branches trigger backend-test simultaneously (e.g. a push to main and an open PR), both jobs will try to upload an artifact with the same name into the same Actions scope. actions/upload-artifact@v4 will reject the second upload with a conflict error. Consider parameterising with the run id, e.g. name: backend-coverage-${{ github.run_id }}.

[lucastononro]

Greptile review findings addressed (commit cd58511)

Fixed — coverage source includes test files (ci.yml:48): Valid. Measured it: with --cov=. the total was 65.81%, inflated by test modules at ~100%. Added backend/.coveragerc with [run] omit = tests/* so the metric reflects production code only, and adjusted the gate --cov-fail-under=60 → 50 to match real production coverage (51.27% measured locally; 60 would have failed every build once tests were excluded). The gate stays a meaningful ratchet just below current coverage.

Dismissed — artifact name not unique across concurrent runs (ci.yml:53): False positive. GitHub Actions artifacts are scoped per workflow run; the upload-artifact@v4 name-conflict error only occurs when the same run uploads the same artifact name twice. This workflow uploads backend-coverage exactly once per run, so a concurrent main push and PR run (distinct run IDs, distinct artifact namespaces) cannot collide. A stable name is also easier to fetch programmatically.

Verification (local, Python 3.12, fresh venv from backend/requirements.txt):

• Exact CI command pytest tests/ -v --tb=short --cov=. --cov-report=xml --cov-report=term-missing --cov-fail-under=50: 296 passed, 8 skipped, coverage 51.27%, gate reached
• coverage.xml verified to contain zero tests/ entries after the omit
• ruff check . and ruff format --check .: clean (142 files)
• ci.yml parses as valid YAML

[lucastononro]

Merged into integration branch staging-v0.0.5 (see the STAGING-v0.0.5.md ledger on that branch for merge order, Greptile follow-ups and test results). These changes will land on main via the staging merge — closing to clear the queue.