Compass Ultra AI DevOps

AI DevOps chatbot and automated release-loop test agent for Compass Ultra.

This repo has two jobs:

1. Provide a local AI DevOps chatbot for release-readiness questions.
2. Run Playwright checks against the live Compass Ultra demo app.

Demo target:

https://www.compassultra.com/app?demo=true

Chatbot

Run the chatbot locally:

npm install
npm start

Then open:

http://localhost:8787

The chatbot can help with:

• Release risk analysis
• Policy gates
• Snapshot diffs
• PDF runbooks
• Rollback planning
• GitHub, Jira, and Slack workflow payloads
• CI release gates
• Launch readiness checklists

By default it uses a local built-in release assistant, so it works without an API key.

To connect it to an OpenAI-compatible provider, set these environment variables before running npm start:

AI_API_KEY=your_key
AI_BASE_URL=https://api.openai.com/v1
AI_MODEL=gpt-4.1-mini

OPENAI_API_KEY also works if AI_API_KEY is not set.

Automated AI DevOps Test

Run the external release-loop test locally:

npm install
npx playwright install chromium
npm test

The test checks:

• Demo workspace loads
• Core release-review buttons are visible
• Feature flags can be toggled
• AI risk analysis can run
• Snapshot diff opens
• PDF runbook export starts
• Policy gates are visible
• GitHub, Jira, and Slack workflow areas are reachable
• Console errors are collected for review

A local run writes:

• ai-devops-report.html
• risk-analysis-failure.png if a failure occurs
• downloaded-runbook.pdf if export succeeds

GitHub Actions

The workflow in .github/workflows/ai-devops-test.yml runs daily at 6:00 UTC and can also be triggered manually from the Actions tab.

Repo Structure

.
├── .github/workflows/ai-devops-test.yml
├── public/
│   ├── app.js
│   ├── index.html
│   └── styles.css
├── tests/test-release-loop.js
├── package.json
├── server.js
└── README.md

Notes

This repo is intentionally separate from the main Compass Ultra product repo. It acts like an external AI DevOps operator: chat with it, run release checks, and use the output as launch-readiness evidence.