2026-11-15
How to run a chat model on your own machine
Install Docker, then from this project run docker compose up -d. Pull a small model with docker compose exec ollama ollama pull llama3.2. In .dev.vars set OLLAMA_BASE_URL=http://127.0.0.1:11434 and MOCK_AI=false.
Restart the app and pick Llama 3.2 (Ollama). Account history still sits in the local database. The prompt is sent to Ollama on your machine. The footer still has to show “Built with Llama” because the licence requires it.
That setup is self-hosted inference. It is not a second copy of a hosted assistant, and it does not remove the model’s own acceptable-use rules.