AI AdminPanel Documentation

AI-Powered Deployment

AI Deploy analyzes a public GitHub repository and uses the Git build pipeline to deploy it. Configure Panel AI under Settings → AI → AI Providers first; customer BYOK keys do not configure this analysis service.

AI-Powered Deploy form with repository URL, branch and customer fields; AI Provider Not Configured is shown and Analyze is disabled.

This capture shows the provider prerequisite before analysis. It does not show an analysis result or successful deployment. The form's promise to review and edit configuration is subject to the v2.12.8 review limitation below.

Analyze a repository

  1. Open Deploy → AI Deploy.
  2. Enter a public https://github.com/ repository URL, select the branch and customer assignment, then choose Analyze.
  3. Read the analysis and check the detected build method, application port and required configuration. Generated output can be wrong; inspect the project's requirements before deploying it.

A Compose detection banner may offer a switch to Compose Deploy for public GitHub repositories. It is a best-effort check: no banner does not prove that the repository lacks Compose files. Use Git Deploy for repository builds and Compose Deploy for reviewed configurations using pre-built images.

Review limitation in v2.12.8

Compose and environment review edits are not submitted by AI Deploy. The Deploy action uses the repository URL, branch, customer, resource limits, and the analysis-derived build method and port. It generates a service name from the repository name plus a suffix. It does not deploy the edited review YAML or send the review environment-variable map.

If your application needs those edits, use Git Deploy with the required changes in the repository, or Compose Deploy with pre-built images. Do not put required configuration into the AI review editor and assume it reaches the container.

Deploy and inspect

Choose resource limits and deploy only after checking the inputs above. Follow build progress, inspect errors and open the service when ready. Verify the app's own behavior and HTTPS access; a build result is not application acceptance. Private repositories use Git Deploy's Deploy Key field. Arbitrary documentation URLs and guaranteed multi-service AI orchestration are outside this flow.