ChatCat (currently under review) is a software suite consisting of a library server, API, and web frontend for conducting laboratory experiments and user studies with generative retrieval systems. You can try it online or watch a short screencast.
Components:
- chat-backend contains the server backend
- chat-ui contains the user interface
- custom-backends dynamically add new chat backends via websockets
- chatnoir-api provides an API client to chat with the cat
- llms the supported llms that are available (you can dynamically add more via custom-backends)
ChatCat has a focus on research-oriented laboratory experiments, so you can dynamically connect custom backends and/or pipelines for generative retrieval systems. Please see the tutorial on how to add a custom backend. We also provide a set of recipies/example backends that are always available online.
You can access the LLMs in ChatCat via an API (please install it via pip install chatnoir-api).
The models alpaca-en-7b and gpt2-xl are public and do not require an API-Key, for other models (including custom backends), you need an API key for authentication (please request one from the admins).
For python, we provide a client library so that you can chat with the cat:
from chatnoir_api.chat import ChatNoirChatClient
chat_client = ChatNoirChatClient()
response = chat_client.chat("how are you?")