Deploy To P3

Deploy to p3

This skill wires up automated deployment for any app to the self-hosted experimental server at *.p3.geoint.mx.

Server Architecture

• Reverse proxy: Caddy, config at deployment/server_config/Caddyfile
• Container registry: GitHub Container Registry (ghcr.io/<owner>/<repo>:<sha>)
• Networking: Shared Docker network named gateway; all containers join it and Caddy routes by container name
• Runner: Self-hosted GitHub Actions runner (reachable via Tailscale)
• Apps path: /home/deploy/apps/<app-name> on host experimentos
• Domain: *.p3.geoint.mx

Asset Templates

This skill includes templates in ${CLAUDE_SKILL_DIR}/assets/:

• ${CLAUDE_SKILL_DIR}/assets/docker-compose.prod.yml — Production Docker Compose template
• ${CLAUDE_SKILL_DIR}/assets/deploy.yml — GitHub Actions deploy workflow template
• ${CLAUDE_SKILL_DIR}/assets/caddyfile-block.txt — Caddyfile reverse proxy block template

Workflow

1. Invocation: User invokes /deploy-to-p3 in a project directory.

2. Information Gathering: Ask the user for:

   • App name (e.g. spotmet) — used as the subdomain and container name prefix. Use $ARGUMENTS if provided.
   • Internal port the app listens on inside the container (e.g. 8000).
   • GitHub owner/repo (e.g. plablo/spotmet) — used to construct the ghcr.io image path.
   • Does the project already have a Dockerfile? If not, warn the user they must add one before deployment will work.
   • Does the project already have a CI workflow? If yes, the deploy workflow should use workflow_run trigger after CI succeeds; if no CI exists, deploy triggers directly on push to main.
   • Any environment variables the app needs at runtime (names only — values go in GitHub Secrets).

3. File Generation: Read the templates from ${CLAUDE_SKILL_DIR}/assets/, replace all placeholders, and write:

   • docker-compose.prod.yml in the project root
   • .github/workflows/deploy.yml in the project

4. Caddyfile Block: Print the filled-in Caddyfile block to the conversation — do NOT write it to disk (it lives on the server).

5. Post-generation Checklist: After generating files, tell the user:

   • Add any secret env vars to GitHub repository secrets.
   • SSH into experimentos as deploy and run mkdir -p ~/apps/{{APP_NAME}}.
   • If this is the very first app ever: docker network create gateway (only needed once on the server).
   • Append the Caddyfile block to /home/deploy/apps/server_config/Caddyfile on experimentos and reload: docker exec caddy caddy reload --config /etc/caddy/Caddyfile.
   • Push to main to trigger the first deploy.

Placeholder Reference