Open ImVexed opened 6 months ago
I have quite limited knowledge on TS or yarn. By poking around I came to the following minimal changes required to talk to a local LLM hosted with Ollama (in my case a Gemma 2B).
https://github.com/Omniplex-ai/omniplex/compare/main...jank:omniplex:main
Please note that you need a an OpenAI key and Bing key in your environment. OpenAI Key is required as GPT-3 is used for the 'tool' selection process (e.g. if to use search or stocks or weather). Replacing that by Ollama will likely require a fine-tuned prompt.
Also note that I had to comment out the Firebase Authentication. Otherwise the web app would not correctly start.
Would be nice to support self hosted LLMs. It doesn't have to be ollama, but it seems to be fairly easy to interface with.