nomic-ai / gpt4all

GPT4All: Run Local LLMs on Any Device. Open-source and available for commercial use.
https://nomic.ai/gpt4all
MIT License
69.72k stars 7.63k forks source link

[Feature] Control which RAG Encoding model is used #2552

Open johmicrot opened 3 months ago

johmicrot commented 3 months ago

Feature Request: Direct Control Over Specific Model for RAG Encoding

I would like to request the addition of a feature that allows direct control over the selection of a specific model for RAG encoding. Currently, there doesn't seem to be an option to choose a particular model for this purpose, and having this capability would greatly enhance flexibility and performance for various use cases.

I am using Ubuntu 23.10.

manyoso commented 3 months ago

We've found this is just not an easy thing to allow while keeping the app usable for the vast majority of users. The model that is included for embeddings right now is extremely powerful and state-of-the-art. Can you say more about why you wish to use an alternative embedding model?

johmicrot commented 2 months ago

It can be helpful if a user can have the choice stronger or weaker models, comparing different models can help debug/evaluate performance. Perhaps a user guide allowing advanced users to change this feature would be helpful.