autopilot-demo

Demo target for the Coding-Autopilot-System AI repair pipeline - triggers intake workflows when CI fails, demonstrating the end-to-end path from failure detection to pull request.

Part of the Coding-Autopilot-System autonomous CI repair platform. The control plane lives in autopilot-core, and the runner-hosted worker/runtime pattern lives in ci-autopilot.

Repo boundary

• autopilot-demo is not the operator and not the worker host. It is the demonstration target repo.
• autopilot-core owns queue scanning, Codex invocation, and PR creation.
• ci-autopilot shows the worker/runtime implementation used to process queued repair tasks.

How the demo works

flowchart LR
    A[Trigger Demo CI\nworkflow_dispatch] --> B[demo-ci.yml runs\nubuntu-latest]
    B --> C{CI result}
    C -->|failure| D[autopilot-create-issue.yml\ncreates intake issue]
    D --> E[Issue: autofix + queued]
    E --> F[autopilot-core operator\npicks up issue]
    F --> G[Codex generates fix]
    G --> H[Pull Request opened]

Loading

1. Trigger Demo CI via workflow_dispatch to simulate a CI failure.
2. autopilot-create-issue.yml detects the failure and creates an issue labeled autofix + queued.
3. The autopilot-core operator scans for the issue and invokes Codex.
4. Codex generates a targeted fix and opens a pull request in this repo.

Running the demo

# Trigger the Demo CI workflow (simulates a failure)
gh workflow run demo-ci.yml -R Coding-Autopilot-System/autopilot-demo

# Watch for the intake issue to be created
gh issue list -R Coding-Autopilot-System/autopilot-demo --label autofix --label queued

# Monitor autopilot-demo for the fix PR
gh pr list -R Coding-Autopilot-System/autopilot-demo

Demo runbook

1. Trigger .github/workflows/demo-ci.yml with simulate_failure=true to produce a known failure signal. Pushes and default dispatches remain green.
2. Confirm .github/workflows/autopilot-create-issue.yml creates an autofix + queued issue.
3. Watch autopilot-core pick up the issue and open a PR back into this repo.
4. Use this repo's issue, branch, and PR history as the audit trail for the demo.

Enterprise proof points

• Demonstrates bounded blast radius: the platform proposes a fix in a dedicated target repo before broader rollout.
• Produces an auditable story for reviewers: failure event, queued issue, operator pickup, and PR are visible artifacts.
• Keeps the demo reproducible with workflow-dispatch entry points rather than hidden local steps.

Workflows

Workflow	Purpose
ci.yml	Portfolio CI - YAML validation (always passes)
demo-ci.yml	Demo trigger - simulates CI activity to test intake flow
autopilot-create-issue.yml	Intake - creates autofix + queued issue on workflow failure

Documentation

• Wiki - setup guide, architecture, configuration reference
• autopilot-core - operator control plane
• ci-autopilot - worker/runtime reference
• Coding-Autopilot-System org

Prerequisites and expected result

• Authenticate GitHub CLI with gh auth status and confirm Actions are enabled for this repository.
• Run the autopilot-core operator with access to this repository before triggering the failure.
• Expect Demo CI to fail, Autopilot Issue Intake to create or reopen one autofix + queued issue, and the operator to propose a pull request.

Reset and troubleshooting

1. Close the completed intake issue and merge or close its repair pull request before the next demonstration.
2. Re-run with simulate_failure=true; intake reopens the matching issue when the same commit is demonstrated again.
3. If no issue appears, inspect gh run list -R Coding-Autopilot-System/autopilot-demo --workflow autopilot-create-issue.yml and confirm the failed run was named Demo CI.
4. If the issue remains queued, verify the autopilot-core operator is running and can read issues and create branches and pull requests in this repository.
5. If CI fails before the demo step, run python -m unittest discover -s tests -v locally and repair the workflow contract violation first.