AI-Powered Deployment

AI Deploy uses a language model to analyze a project and generate a Docker Compose configuration automatically. Paste a URL, review the generated config, and deploy.

Prerequisites

AI Deploy requires an AI provider to be configured in Settings > AI Provider:

Provider	Configuration
OpenAI	API key, model name (e.g., gpt-4o)
LiteLLM	Proxy URL, API key, model name
Ollama	Server URL (e.g., http://ollama:11434), model name

Any OpenAI-compatible API endpoint works.

How It Works

Step 1: Provide a URL

Navigate to Deploy > AI Deploy and enter:

Field	Description
URL	GitHub repo URL, documentation link, or project website
Service Name	Name for the service (becomes the subdomain)
Customer	Which customer owns this service

Step 2: AI Analysis

The panel sends the URL to the configured AI model, which:

1. Fetches and reads the repository README, Dockerfile, compose files, and package manifests
2. Identifies the tech stack, dependencies, and required services
3. Generates a Docker Compose configuration with:
   • Correct base images and versions
   • Required environment variables with sensible defaults
   • Volume mounts for persistent data
   • Health check endpoints
   • Multi-service setups (e.g., app + database)
4. Returns the configuration with an explanation of design decisions

Analysis progress streams to the UI via WebSocket so you can watch the AI's reasoning.

Step 3: Review and Edit

The generated Docker Compose configuration is displayed in an editor. You can:

• Review the AI's explanation of why it chose specific images and settings
• Modify environment variables, image versions, or resource limits
• Add or remove services
• Adjust volume mounts or networking

The editor supports YAML syntax highlighting and validation.

Step 4: Deploy

Click Deploy to execute the generated configuration. From this point, the process is identical to a Compose Deploy:

1. Traefik labels are injected for routing
2. Images are pulled
3. Containers are started
4. Health checks are monitored
5. SSL certificates are provisioned

Use Cases

AI Deploy works well for:

• Unknown projects — paste a GitHub URL and let AI figure out how to run it
• Quick prototyping — get a service running without reading deployment docs
• Complex stacks — AI handles multi-service setups (app, database, cache, worker)
• Documentation links — paste a project's install docs and AI extracts the deployment steps

Limitations

• AI-generated configurations should always be reviewed before deploying
• Quality depends on the AI model — larger models produce better results
• Private repositories require the Git Deploy method instead
• Very new or niche projects may not be well-known to the AI model
• The AI does not have access to the repository's actual source code beyond what is publicly readable

Fallback

If the AI-generated configuration does not work:

1. Check the container logs for errors
2. Edit the generated compose file in the service settings
3. Use Git Deploy or Compose Deploy for more control