Ollama (Offline)

Run Axiom entirely on-device with no API key using Ollama.

Setup

1. Install Ollama from ollama.com
2. Pull a model:

ollama pull llama3.2

3. Run axiom-wiki init and select Ollama (local)

Axiom connects to http://localhost:11434 by default and validates the connection during setup.

Docker + Ollama

services:
  axiom-wiki:
    image: axiomwiki/axiom-wiki
    volumes:
      - ./wiki:/app/wiki
      - ./raw:/app/raw
    environment:
      - OLLAMA_BASE_URL=http://ollama:11434/api
  ollama:
    image: ollama/ollama
    volumes:
      - ollama_data:/root/.ollama
volumes:
  ollama_data:

Model recommendations

For wiki use, models with strong instruction following and JSON output work best:

• llama3.2 — good balance of quality and speed
• mistral — fast, good for smaller documents
• qwen2.5 — strong multilingual support