Reference

Troubleshooting

The most common problems and how to deal with them.

SmartScreen warns on install

Installers are unsigned, so Windows SmartScreen shows a warning on first run. Choose “More info” → “Run anyway”.

llama-server not found

Install llama.cpp in Settings or point the app at an existing install. The app does not build llama.cpp — it downloads a prebuilt release for your GPU.

Model won't start or the server crashes

Most often the model doesn't fit in VRAM. Lower GPU layers (-ngl), context size or KV cache quantization, pick a smaller model quantization, or run Auto-tune. Details are in the server log.

Phone can't reach remote access

Make sure both devices are on the same Wi‑Fi and the router doesn't have client isolation on. The PIN may have expired — generate a new one.

Privacy

Models, chats and speech stay on your computer. Only these go over the network: model and engine downloads (HuggingFace, GitHub), update checks, and requests to an AI provider if you configure a key.