Rollback Triggers & Multi-Stage Workflows

Rollback triggers use Codex to analyze production incidents and recommend whether to roll back a deployment. When monitoring detects an anomaly (error rate spike, latency increase, health check failure), a CI workflow triggers Codex to analyze the most recent deployment diff, correlate it with the incident symptoms, and produce a rollback recommendation. The recommendation includes: likely root cause (which change in the diff could cause this symptom), confidence level, and suggested mitigation (rollback vs. hotfix).

Multi-stage workflows chain multiple Codex operations in a single pipeline. Stage 1: review the diff for correctness. Stage 2: generate tests for untested code paths. Stage 3: validate documentation updates. Stage 4: perform the deployment gate check. Each stage uses a different profile (model, timeout, rules) optimized for its task. Stages can run in parallel when independent or sequentially when one stage's output feeds the next. The workflow should short-circuit on failure — if Stage 1 finds a critical issue, skip Stage 2-4 to save time and cost.

name: Multi-Stage Codex Pipeline
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  stage-1-review:
    runs-on: ubuntu-latest
    outputs:
      has-critical: ${{ steps.review.outputs.has-critical }}
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - run: npm install -g @openai/codex
      - name: Critical Review
        id: review
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
        run: |
          codex --profile review --approval full-auto \
            "Check for critical issues only: security, data loss, breaking changes. 
             Output JSON: {critical: boolean, issues: string[]}" \
            < <(git diff origin/${{ github.base_ref }}...HEAD) > review.json
          echo "has-critical=$(jq -r '.critical' review.json)" >> "$GITHUB_OUTPUT"

  stage-2-tests:
    needs: stage-1-review
    if: needs.stage-1-review.outputs.has-critical != 'true'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npm install -g @openai/codex
      - name: Test Coverage Check
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
        run: echo "Stage 2: checking test coverage..."

  stage-3-gate:
    needs: [stage-1-review, stage-2-tests]
    if: needs.stage-1-review.outputs.has-critical != 'true'
    runs-on: ubuntu-latest
    steps:
      - run: echo "Stage 3: deployment gate check..."