This is a fork of lm-sys/FastChat, a handy tool for locally running Large Language Models. Into the inference I've hacked internet access. This was not intended for real use, but as a tech demo that was slapped together on a wednesday night.
Changes compared to the original repo:
- Vicuna model system prompt specifies how to use the internet
- inference.py chat_loop is heavily modified to interupt excecution and fetch internet information
- find_website.py was added for retrieving information from web.
Start guide and more can be found in FASTCHAT_README.md