blackroad-os-prism-console/.github/workflows/auto-fix-deployment.yml

name: Auto-Fix Failed Deployments

on:
  workflow_run:
    workflows: ["Railway Deploy"]
    types: [completed]

jobs:
  detect-and-fix:
    runs-on: ubuntu-latest
    if: ${{ github.event.workflow_run.conclusion == 'failure' }}

    steps:
      - name: Checkout
        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
        with:
          ref: ${{ github.event.workflow_run.head_branch }}
          fetch-depth: 10

      - name: Setup Node.js
        uses: actions/setup-node@1e60f620b9541d16bece96c5465dc8ee9832be0b
        with:
          node-version: '20'

      - name: Analyze Failure
        id: analyze
        run: |
          echo "Analyzing deployment failure..."

          # Common failure patterns and fixes
          NEEDS_DEPS=false
          NEEDS_BUILD_FIX=false
          NEEDS_ENV_FIX=false
          NEEDS_RAILWAY_LINK=false

          # Check if package-lock exists but pnpm is used
          if [ -f "pnpm-lock.yaml" ] && [ -f "package-lock.json" ]; then
            echo "fix_type=lock_conflict" >> $GITHUB_OUTPUT
            echo "Detected lock file conflict"
          elif [ ! -f "package-lock.json" ] && [ ! -f "pnpm-lock.yaml" ] && [ -f "package.json" ]; then
            echo "fix_type=missing_lock" >> $GITHUB_OUTPUT
            echo "Missing lock file"
          elif [ -f "railway.json" ] || [ -f "railway.toml" ]; then
            echo "fix_type=railway_config" >> $GITHUB_OUTPUT
            echo "Railway config needs verification"
          else
            echo "fix_type=generic" >> $GITHUB_OUTPUT
            echo "Generic deployment issue"
          fi

      - name: Fix Lock File Conflict
        if: steps.analyze.outputs.fix_type == 'lock_conflict'
        run: |
          echo "🔧 Fixing lock file conflict..."
          rm -f package-lock.json
          npm install -g pnpm
          pnpm install --frozen-lockfile || pnpm install
          git config user.name "BlackRoad Auto-Fix Bot"
          git config user.email "bot@blackroad.systems"
          git add pnpm-lock.yaml
          git diff --staged --quiet || git commit -m "🤖 Auto-fix: Remove package-lock.json conflict

          Removed npm lock file in favor of pnpm-lock.yaml

          Auto-fix triggered by failed deployment

          Co-Authored-By: Claude <noreply@anthropic.com>"

      - name: Fix Missing Lock File
        if: steps.analyze.outputs.fix_type == 'missing_lock'
        run: |
          echo "🔧 Generating lock file..."
          if [ -f "pnpm-lock.yaml" ]; then
            npm install -g pnpm
            pnpm install
          else
            npm install
          fi
          git config user.name "BlackRoad Auto-Fix Bot"
          git config user.email "bot@blackroad.systems"
          git add package-lock.json pnpm-lock.yaml 2>/dev/null || true
          git diff --staged --quiet || git commit -m "🤖 Auto-fix: Generate missing lock file

          Created lock file for dependency consistency

          Auto-fix triggered by failed deployment

          Co-Authored-By: Claude <noreply@anthropic.com>"

      - name: Verify Railway Config
        if: steps.analyze.outputs.fix_type == 'railway_config'
        run: |
          echo "🔧 Verifying Railway configuration..."

          # Ensure railway.json or railway.toml is valid
          if [ -f "railway.json" ]; then
            node -e "JSON.parse(require('fs').readFileSync('railway.json'))" || {
              echo "Invalid railway.json detected"
              # Create basic valid config
              cat > railway.json << 'RAILWAYCONFIG'
              {
                "$schema": "https://railway.app/railway.schema.json",
                "build": {
                  "builder": "NIXPACKS"
                },
                "deploy": {
                  "restartPolicyType": "ON_FAILURE",
                  "restartPolicyMaxRetries": 10
                }
              }
              RAILWAYCONFIG
              git config user.name "BlackRoad Auto-Fix Bot"
              git config user.email "bot@blackroad.systems"
              git add railway.json
              git commit -m "🤖 Auto-fix: Repair railway.json configuration

              Fixed invalid Railway configuration

              Auto-fix triggered by failed deployment

              Co-Authored-By: Claude <noreply@anthropic.com>"
            }
          fi

      - name: Generic Fix - Clean and Rebuild
        if: steps.analyze.outputs.fix_type == 'generic'
        run: |
          echo "🔧 Attempting generic fixes..."

          # Clean common build artifacts
          rm -rf node_modules .next .out dist build .cache

          # Ensure package.json has basic scripts
          if [ -f "package.json" ]; then
            node -e "
              const pkg = require('./package.json');
              if (!pkg.scripts) { pkg.scripts = {}; }
              if (!pkg.scripts.build) { pkg.scripts.build = 'echo No build needed'; }
              if (!pkg.scripts.start) { pkg.scripts.start = 'echo No start command'; }
              require('fs').writeFileSync('package.json', JSON.stringify(pkg, null, 2));
            "
          fi

      - name: Push Fixes
        run: |
          git config user.name "BlackRoad Auto-Fix Bot"
          git config user.email "bot@blackroad.systems"

          if git diff --staged --quiet && git diff --quiet; then
            echo "No fixes needed or already applied"
          else
            git add -A
            git commit -m "🤖 Auto-fix: Resolve deployment issues

            Automatic remediation of deployment failure

            Fix type: ${{ steps.analyze.outputs.fix_type }}

            Co-Authored-By: Claude <noreply@anthropic.com>" || true
            git push origin ${{ github.event.workflow_run.head_branch }} || echo "Nothing to push"
          fi

      - name: Retry Deployment
        env:
          RAILWAY_TOKEN: ${{ secrets.RAILWAY_TOKEN }}
        run: |
          echo "🔄 Retrying deployment after fixes..."
          npm install -g @railway/cli
          railway up --detach || {
            echo "⚠️ Retry failed - manual intervention may be needed"
            exit 1
          }
          echo "✅ Deployment retry successful!"

      - name: Notify on Persistent Failure
        if: failure()
        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea
        with:
          script: |
            const issue = {
              owner: context.repo.owner,
              repo: context.repo.repo,
              title: '🚨 Auto-Fix Failed: Deployment Issue Requires Manual Intervention',
              body: `## Deployment Failure Alert

            **Branch:** \`${{ github.event.workflow_run.head_branch }}\`
            **Workflow:** Railway Deploy
            **Auto-Fix Attempt:** Failed

            ### Analysis
            - Fix type attempted: \`${{ steps.analyze.outputs.fix_type }}\`
            - Automatic remediation was unsuccessful

            ### Next Steps
            1. Review deployment logs: ${{ github.event.workflow_run.html_url }}
            2. Check Railway dashboard for detailed error messages
            3. Review recent commits for breaking changes
            4. Verify environment variables and secrets

            ### Workflow Run
            - Run ID: ${{ github.event.workflow_run.id }}
            - Commit: ${{ github.event.workflow_run.head_sha }}

            This issue was automatically created by the Auto-Fix workflow.

            cc: @blackboxprogramming

            ---
            🤖 Auto-generated issue`
            };

            // Check if similar issue already exists
            const { data: issues } = await github.rest.issues.listForRepo({
              owner: context.repo.owner,
              repo: context.repo.repo,
              state: 'open',
              labels: 'deployment-failure'
            });

            const existingIssue = issues.find(i =>
              i.title.includes('Auto-Fix Failed') &&
              i.body.includes('${{ github.event.workflow_run.head_branch }}')
            );

            if (!existingIssue) {
              const { data: newIssue } = await github.rest.issues.create({
                ...issue,
                labels: ['deployment-failure', 'auto-fix-failed', 'urgent']
              });
              console.log('Created issue:', newIssue.html_url);
            } else {
              console.log('Similar issue already exists:', existingIssue.html_url);
              await github.rest.issues.createComment({
                owner: context.repo.owner,
                repo: context.repo.repo,
                issue_number: existingIssue.number,
                body: '🔄 Another auto-fix attempt failed. Still requires manual intervention.'
              });
            }