Guide · Updated 2026-05-31
Ollama + Liminal
Point AGENT_API_BASE_URL at Ollama's OpenAI-compatible server.
1. Start Ollama
Install from ollama.com and pull a code-capable model (e.g. qwen2.5-coder).
2. Configure Liminal
AGENT_API_BASE_URL=http://127.0.0.1:11434/v1
AGENT_MODEL=your-ollama-model-tag
Leave AGENT_API_KEY empty or set a placeholder if required.
FAQ
Common questions
Is local inference slower?
Depends on GPU/CPU. You trade cloud latency for data staying on your network.