First CI Heal

Trigger and observe GitWire's self-healing CI pipeline.

Prerequisites

• GitWire is deployed with a repo that has GitHub Actions workflows
• The workflow_run event is subscribed on your GitHub App
• The repo has a CI workflow that can intentionally fail

Step 1: Create a Failing CI Run

Introduce a deliberate error in your repo (e.g., a lint error):

// Add this to a JS file to trigger ESLint
const  x  =  1  // extra spaces → lint error

Commit and push. GitHub Actions will run and fail.

Step 2: Watch the Webhook

GitHub sends a workflow_run webhook with conclusion: "failure".

Step 3: Watch the Diagnosis

The CI Heal Worker will:

1. Fetch the failure logs
2. Send to Claude for diagnosis
3. Classify the failure type (e.g., lint_error)
4. If healable → fetch the failing file → generate fix → open PR

Step 4: Check the PR

If the failure is healable, GitWire will:

1. Create a branch: gitwire/heal/{run_id}
2. Commit the fixed file
3. Open a PR with:
   • Title: fix(ci): resolve lint_error in path/to/file.js
   • Body: Root cause, fix applied, confidence level

Check https://gitwire.yourdomain.com/ci for the healing status.

Step 5: Review and Merge

1. Review the auto-generated PR on GitHub
2. If the fix looks correct, merge it
3. If not, close the PR and fix manually

Dashboard Verification

# Check CI stats
curl https://gitwire.yourdomain.com/api/ci/stats \
  -H "Authorization: Bearer YOUR_API_KEY"

# Check heal history
curl https://gitwire.yourdomain.com/api/heal/stats \
  -H "Authorization: Bearer YOUR_API_KEY"

Troubleshooting

Problem	Solution
No healing triggered	Verify workflow_run event is subscribed
Failure not diagnosed	Check ANTHROPIC_API_KEY is valid
Fix PR not created	Failure type may not be healable (check heal_status: skipped)
PR has bad fix	Check heal_confidence — low confidence fixes need more review

→ First Contributor Fix