Brain on the stick — or in the cloud
Cloud or local brain
You choose where thinking happens: a model file on the machine, or a frontier cloud model with your own API keys. Files, logins, and memory stay on bia either way.
Local mode loads a GGUF model from the stick’s home. It can work without the internet and is tuned for the machine’s memory. Context size is adjustable; a failed resize rolls back instead of bricking the session.
Frontier Model Mode (FMM) points at OpenRouter, OpenAI, Anthropic, or xAI. Default cloud room to think is 128K tokens — the scale used in testing. Keys live in bia’s state; status screens mask them.
The same system skill also shares a browser console link, file download links, and a polite “ask the human” path for captchas — so bia escalates instead of guessing.
In everyday life
- Ships with a local model path, or a lighter stick that expects cloud keys.
- Switches between cloud and local without moving your files or messenger sessions.
- Offers an HTTPS web console with agent CA and client-certificate auth.
- Asks a person for captchas and hard logins instead of hallucinating them.
Unlike a normal chatbot
SaaS chatbots keep your life in their cloud. Here the home directory is yours; only the thinking engine is optional and swappable.