Ollama

Server Commands

Run Ollama as a background server for API access. Configure host bindings, manage server lifecycle, and enable remote connections for multi-client setups.

6 commands

Pro Tips

Set OLLAMA_HOST=0.0.0.0:11434 to allow connections from other machines.

Ollama server starts automatically on first 'ollama run' if not running.

Use OLLAMA_MODELS to customize where models are stored on disk.

Common Mistakes

Binding to 0.0.0.0 exposes the server - use firewall rules or authentication proxy.

Server keeps models in memory for fast inference - monitor RAM usage.

Commands

Start server

$ ollama serve

Start Ollama API server on default port 11434.

Bind to all interfaces

$ OLLAMA_HOST=0.0.0.0:11434 ollama serve

Allow connections from other machines.

Custom models directory

$ OLLAMA_MODELS=/path/to/models ollama serve

Store models in a custom location.

Check server status

$ curl http://localhost:11434

Verify the server is running.

Enable GPU layers

$ OLLAMA_NUM_GPU=35 ollama serve

Configure number of layers to offload to GPU.

Set max loaded models

$ OLLAMA_MAX_LOADED_MODELS=2 ollama serve

Limit concurrent models in memory.