2 ms·
I think this looks nice and compelling, but I wonder how you lock down the seats here. For an organization, they're going to be thinking "why not just get one
by nkozyra 3y ago
I think this looks nice and compelling, but I wonder how you lock down the seats here.
For an organization, they're going to be thinking "why not just get one seat and share it?"
It's also a hard sell to developers, who can spin up a LLM+RAG a lot easier than folks in spaces being similarly served by an OpenAI-as-a-service proxy app. Pointing ollama at my codebase is fairly trivial.
Not trying to rain on any parades; there are a lot of AI-powered projects like this and the ones that are selling are those that have proprietary data and serve less technical spaces.
- replete 3y agoWhat method do you use for using oIlama with RAG like this? I
- nkozyra 3y agoVector storage of choice + Langchain
- dakshgupta 3y agoThis is the case with most seat-based tools, except here you authenticate via your code hosting service, so it only lets you query the repos you have read perms for. Of course it’s not fool proof. IMO it’s really easy to build a basic version of this that answers basic questions. We are attempting to build something that can answer the types of questions a good developer at a company would have about the company codebase. In our experience that’s a much harder problem than it looks.