Reverse engineered prompt

Build me a self hosted AI code reviewer called Mira. I want something a team can run on its own infrastructure that connects to GitHub first, and ideally GitLab too, reviews every pull request or merge request with the team’s own LLM key, and leaves concise, useful comments instead of noisy spam. It should let people pick different model providers through config, support custom review rules in plain language, learn from merged PRs and rejected comments over time, and answer inline questions when someone tags it.

I also want a clean dashboard that stores data in SQLite or Postgres and shows review history, webhook and event activity, token and cost tracking, package inventory across repos, CVE alerts from dependency scanning, and dependency or blast radius views powered by a code index. Make it easy to run with Docker and friendly to Railway, Fly, or Render. Include sample env and config files, sensible defaults, and setup docs so someone can self host it quickly. Look up current docs online if you need to.

Want more depth? Deep Reverse