llama-farm /llamafarm:start - Start LlamaFarm Services
Start the LlamaFarm backend services required for AI chat, RAG, and project management.
Usage
/llamafarm:start # Start all services
/llamafarm:start --no-browser # Start without opening Designer UI
/llamafarm:start --runtime # Also start Universal Runtime (ML/embeddings)
What This Command Does
- Checks prerequisites - Verifies LlamaFarm CLI is installed
- Starts services - Launches FastAPI server and RAG worker
- Waits for health - Polls until services are ready
- Reports status - Shows component health and URLs
Implementation
When the user runs /llamafarm:start, follow these steps:
Step 1: Check CLI Installation
which lf || echo "NOT_FOUND"
If not found, provide installation instructions:
LlamaFarm CLI not found. Install it with:
# Using Homebrew (recommended)
brew install llama-farm/tap/lf
# Or download from GitHub releases
https://github.com/llama-farm/llamafarm/releases
Step 2: Check for Existing Services
lf services status 2>/dev/null || echo "SERVICES_NOT_RUNNING"
If services are already running, report status and skip startup.
Step 3: Start Services
lf start
This command:
- Starts the FastAPI server on port 8000
- Starts the Celery RAG worker for async processing
- Opens the Designer UI in browser (unless
--no-browser)
Step 4: Wait for Health
Poll the health endpoint until ready (up to 30 seconds):
curl -s http://localhost:14345/health | jq .
Expected response:
{
"status": "healthy",
"components": {
"server": "healthy",
"celery": "healthy"
}
}
Step 5: Report Results
Present results to user:
LlamaFarm Services Started
==========================
Status: Running
Server: http://localhost:14345
Designer UI: http://localhost:14345 (opened in browser)
API Docs: http://localhost:14345/docs
Components:
- Server: healthy
- RAG Worker: healthy
MCP Server: http://localhost:14345/mcp (available for Claude Code)
Next steps:
- Create a project: lf init my-project
- Or use /llamafarm:config to generate a configuration
Options
| Option | Description |
|---|---|
--no-browser |
Don't open Designer UI in browser |
--runtime |
Also start Universal Runtime for ML/embeddings |
--timeout <seconds> |
Health check timeout (default: 30) |
Starting Universal Runtime
If --runtime is specified or user has Universal Runtime configured:
# Check if Universal Runtime is configured
grep -q "universal" llamafarm.yaml && echo "UNIVERSAL_CONFIGURED"
# Start Universal Runtime (separate terminal/process)
lf runtime start
Universal Runtime provides:
- Local embedding models (nomic-embed-text-v2-moe)
- ML inference (anomaly detection, classification)
- OCR capabilities
Error Handling
Port Already in Use
lsof -i :14345
If port is occupied, suggest:
Port 8000 is already in use by another process.
Options:
1. Stop the existing process: kill <PID>
2. Use a different port: LF_SERVER_PORT=8001 lf start
Missing Dependencies
If startup fails due to missing Python packages:
Some LlamaFarm dependencies are missing.
Run:
cd /path/to/llamafarm
uv sync --all-extras
Config Not Found
If no llamafarm.yaml exists:
No llamafarm.yaml found in current directory.
Options:
1. Initialize a new project: lf init my-project
2. Use /llamafarm:config to generate a configuration
3. Navigate to an existing LlamaFarm project directory
Related Commands
/llamafarm:stop- Stop services/llamafarm:status- Check service health/llamafarm:logs- View service logs