Search intent answer
Ollama is a common local LLM route for this project, but the practical failures are small and painful: wrong model name, closed Ollama process, proxy interception, or a model too large for VRAM.
How the console handles it
- Enter your Ollama base URL and model name.
- Confirm the server responds before the VTuber backend starts.
- Compare model size with available GPU memory.
- Check proxy bypass for localhost services.
- Unlock a clean conf.yaml patch plan.