Ollama

Provider value: ollama

Run models locally with Ollama. No API key required.

Setup

1. Install Ollama
2. Pull a model: ollama pull llama3
3. Run ra:

ra --provider ollama --model llama3 "Write a haiku"

Environment variables

Variable	Required	Description
RA_OLLAMA_HOST	No	Ollama host (default: http://localhost:11434)

Remote Ollama

Point ra at an Ollama instance running on another machine:

export RA_OLLAMA_HOST=http://my-server:11434
ra --provider ollama --model llama3 "Hello"

Or via CLI flag:

ra --provider ollama --ollama-host http://my-server:11434 --model llama3 "Hello"

See also

• Configuration — provider credentials reference