Does it work with local LLMs?

[serfriz]

Instead of using the OpenAI API, can we use local models with for example Ollama?

[schelskedevco]

It's not something that I've tested but I don't think it should be too big of a lift.

The LLM API calls are all just in this file: https://github.com/schelskedevco/ignidash/blob/main/convex/use_openai.ts. This is in Convex which is self-hosted when you're self-hosting the app.

Maybe the Vercel AI SDK, which (as I understand it) helps provide a unified interface for different LLM APIs including Ollama, could be used here to make it easier to swap out APIs of one's choosing?

I don't have time to prioritize this at the moment, and probably won't for the next few weeks, but if you want to try something like this for yourself I'm happy to help.