-
Notifications
You must be signed in to change notification settings - Fork 198
Hackage AI support iframe? #1368
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
Are you talking about just an iframe with any stock AI chatbot? Or about developing a custom-trained Hackage-specific one? |
Stock AI chatbot (have the user load an API key) is easier to implement, but long-term, Mistral 7B is open-source, is from a country few have objections to, and claims to be very low-cost to run (3B is cheaper, but lacks reinforcement learning). I'm actually messing around with a local clone right now, in hopes of seeing what an API-key chatbot might look like. There are actual real issues with this, given that yes, stock AI chatbots have dumped bad monad tutorials when prompted. However, Deepseek helped me figure out how to get a frame-delay timer on my Bricks toy tetris program up (i.e, how to use the BChan to set and update a custom event), so it's not completely useless. |
Stock AI are reasonably good, but I already have one in my text editor and another on GitHub, both equipped with intimate knowledge about what I’m working on. I don’t see an appeal of another one embedded into Hackage. As for a self-trained custom AI, I’d like to see a prototype. The problem is that you’d have to train it on as much Haskell as possible, not just Hackage packages themselves, otherwise it’s unlikely to be useful. |
I think it's a difference in use-cases; after all, you're on CLC, a professional Haskeller, and demonstrably a great programmer. Less-skilled users (and others have also reported this as an issue) have difficulty navigating and comprehending the material on Hackage, and it'd be an ergonomic improvement to have access to a chatbot in-line, especially if, say, they're on a coffee break or a lunch break and are reading up on libraries on mobile. |
Given variable levels of documentation in the Haskell library ecosystem, one possibility would be to simply add AI assistance to Hackage and let the LLM / reasoning agent handle it.
Of course, the realistic problem is that LLMs have variable quality, and while occasionally useful, are generally unreliable.
How would Hackage maintainers feel about adding an AI support iframe to Hackage? How would Hackage users feel about adding AI support iframes to Hackage?
How about support for linking it to maintainer AI?
The text was updated successfully, but these errors were encountered: