Project 27 of 39 · AI systems
AI Chat Assistant
Chat backend that runs a locally hosted PyTorch model by default and falls back to OpenRouter when the local path is unavailable.
01Overview
The chat backend behind the assistant on this site. It runs a locally hosted model by default, and falls back to a hosted provider when the local path is unavailable.
Keeping a local path first means the common case does not depend on someone else's uptime.
02How it is built
- Local model by default, hosted provider as fallback.
- Session handling so a conversation has continuity.
- Streams replies rather than waiting for a complete answer.
03Screens
