Ci Fix

Diagnose and fix all CI pipeline failures automatically. Runs in a loop until CI is green.

## Phase 1: Diagnose

Get the latest CI run and identify all failures:

```bash
RUN=$(gh run list --workflow ci.yml --limit 1 --json databaseId,conclusion --jq '.[0]')
echo "$RUN"

If the conclusion is "success", report that CI is green and stop.

If the conclusion is "failure", get per-job status:

RUN_ID=$(echo "$RUN" | jq -r '.databaseId')
gh api repos/Acumenus-Data-Sciences/Parthenon/actions/runs/$RUN_ID/jobs \
  --jq '.jobs[] | "\(.name): \(.conclusion)"'

For each failed job, get the error details:

JOB_ID=$(gh api repos/Acumenus-Data-Sciences/Parthenon/actions/runs/$RUN_ID/jobs \
  --jq '.jobs[] | select(.conclusion=="failure") | .id')
gh api repos/Acumenus-Data-Sciences/Parthenon/actions/jobs/$JOB_ID/logs 2>/dev/null \
  | grep "FAILED\|error TS\|##\[error\]\|SQLSTATE\|Error:" | head -30

Classify each error into one of: Pint, PHPStan, Pest, TypeScript, ESLint, Vitest, mypy, pytest, Migration, Build.

Phase 2: Fix by Category

Pint (PHP code style) — ALWAYS auto-fix

# Use Docker Pint (matches CI version) — CRITICAL for version parity
docker compose exec -T php sh -c "cd /var/www/html && vendor/bin/pint"

If Docker is not running, use local: cd backend && vendor/bin/pint Note: Local Pint version may differ from CI. Docker Pint matches CI exactly.

PHPStan (PHP static analysis)

• If errors are from code YOU changed: fix the actual type issues
• If errors are pre-existing or from strict_types additions: regenerate baseline

  cd backend && vendor/bin/phpstan analyse --generate-baseline
  

• Never add inline @phpstan-ignore — fix the code or use baseline

Pest (PHP tests)

• Read the failing test AND the controller/service it tests
• Common causes:
  • Migration failures: CI missing a schema → add to .github/workflows/ci.yml "Create schemas" step
  • Mock return types: Service returns object but mock returns array → cast with (object)
  • Route binding: Model uses getRouteKeyName() → ensure test uses the right key (slug vs id)
  • RBAC: User missing permissions → seed RolePermissionSeeder and assign correct role
  • PostGIS unavailable: Use SAVEPOINT in migration to prevent transaction abort

TypeScript Build Errors

CRITICAL: CI uses npm run build (Vite) which is STRICTER than tsc --noEmit. Common CI-only failures that pass locally:

• useRef<T>() without argument → useRef<T | undefined>(undefined) (React 19 strict)
• Unused imports → remove entirely (don't prefix with _)
• as Record<string, ElementType> → cast through unknown first
• Accessing .data on paginated responses → check if type uses .items or .data

Fix, then verify with BOTH:

cd frontend && npx tsc --noEmit    # Type check
cd frontend && npx vite build      # Full build (catches more)

ESLint

cd frontend && npx eslint . --fix  # Auto-fix what's possible

Manual fixes needed for:

• react-hooks/rules-of-hooks: Move early returns AFTER all hooks
• react-hooks/exhaustive-deps: Add missing deps or restructure
• react-refresh/only-export-components: Split exports into separate files

mypy (Python)

cd ai && mypy app/

Add type annotations. Use Any only for dynamic third-party types.

pytest (Python)

cd ai && PYTEST_CURRENT_TEST=1 pytest -v

Common fixes: mock ChromaDB client, fix import paths, match service return types.

Phase 3: Verify Locally

Run ALL checks that CI runs, in order:

# Backend
docker compose exec -T php sh -c "cd /var/www/html && vendor/bin/pint --test" \
  && echo "Pint: PASS" || echo "Pint: FAIL"

cd /home/smudoshi/Github/Parthenon/backend \
  && vendor/bin/phpstan analyse \
  && echo "PHPStan: PASS" || echo "PHPStan: FAIL"

# Frontend
cd /home/smudoshi/Github/Parthenon/frontend \
  && npx tsc --noEmit \
  && echo "TSC: PASS" || echo "TSC: FAIL"

# Python
cd /home/smudoshi/Github/Parthenon/ai \
  && python3 -c "import py_compile; py_compile.compile('app/main.py', doraise=True)" \
  && echo "Python syntax: PASS" || echo "Python syntax: FAIL"

Phase 4: Commit, Push, Monitor

Stage ONLY the files you changed (not .env or secrets):

git add <specific-files>
git commit -m "fix(ci): <description of what was fixed>"
git push origin main

Then monitor the CI run:

sleep 210  # CI takes ~3.5 minutes
RUN=$(gh run list --limit 1 --json databaseId --jq '.[0].databaseId')
gh run view "$RUN" --json jobs --jq '.jobs[] | "\(.name): \(.conclusion)"'

Phase 5: Loop Until Green

If any jobs still fail, repeat from Phase 1 with the NEW error details. Common iteration patterns:

• Fix 1 reveals Fix 2 (e.g., fixing TS exposes ESLint issue)
• CI version differs from local (Pint 1.27 vs 1.29 — always use Docker Pint)
• Cascading model changes (fixing a type breaks a downstream mock)

Reporting

After CI is green, summarize all fixes in a table:

Job	Error	Fix	Files Changed
Backend/Pint	concat_space	Auto-fixed	3 files
Frontend/Build	TS6133 unused import	Removed import	1 file

Rules

1. Fix the code, not the linter — don't add ignores unless it's a genuine false positive
2. Use Docker Pint for version parity with CI
3. Never modify test expectations to make tests pass — fix the implementation
4. Never skip hooks (--no-verify) — fix the underlying issue
5. Stage specific files — never git add . or git add -A
6. One commit per fix iteration — makes rollback easy
7. Check for .env files before staging — never commit secrets