Closed kingjulio8238 closed 1 month ago
see PR #19 for reference.
Update
Instead of replacing GPT 4 and gpt-4-preview with Llama 3 8B and LLaVa respectively, include a model setup functionality for users to choose which models and keys they want to provide. Allows for more flexibility for end users.
Started on pr #32 but a little stuck for now
memary uses GPT 4 for most LLM calls which doesn't allow the community to use memary for free. It would be highly beneficial to see a shift from using GPT 4 to Llama 3 8B. We would also like to see LLaVA being used for the agent's vision tool.
Additionally, we would like the community to be able to run these models locally and there's no better way than using Ollama. LLaVa (original and fine-tuned) can be found here: https://ollama.com/library?q=llava
Main Changes: