Experimenting with other LLMs

[AshishSardana]

LangchainJS provides HuggingFaceInference class that can use the LLMs from HF Hub.

The documentation for using it is here and the source code here.

Has anyone tried replacing the OpenAIChat LLM with one of HuggingFace's LLM?

I've tried by making simple changes to makeChain.ts and adding HUGGINGFACE_HUB_API_KEY in the .env file, but I'm receiving empty response from the chatbot.

[AshishSardana]

@mayooear - have you had a chance to try other LLMs? It'd be great if you can give it a quick try and share the necessary changes required.
I'd love to add the functionality where the user could select the LLM from the GUI in order to compare the response generated by them. For starters, the embeddings could be used from OpenAI's LLM but the response completion LLM can be dynamic.

[BaltekLabs]

Now with the ability to run models like llama and alpaca locally, would be nice to pull from local models like with oobabooga