search

AllYourBot / hostedgpt

An open version of ChatGPT you can host anywhere or run locally.
MIT License
304 stars 122 forks source link

Add language models to seed #384

Closed nwarwick closed 1 month ago

nwarwick commented 1 month ago

I was running into an issue setting up the application locally, specifically at the seeding step (see image below). I realized it was due to the fact that it was trying to create assistant records without having language model records to connect them to.

Screenshot 2024-05-30 at 2 23 07 PM
krschacht commented 1 month ago

@mwarwick I really appreciate you submitting a fix for this! I inadvertently broke this when I merged in a recent PR. Adding a test case for seed is exactly what I needed.

BTW, what's motivated you to set up HostedGPT?

And as you get started using it, please let me know how it's working out for you. I monitor the Discussions (and PRs) closely and am working on this project a lot so I love to hear any feedback about what could improve it for you.

nwarwick commented 1 month ago

My pleasure @krschacht!

BTW, what's motivated you to set up HostedGPT?

I primarily just wanted to take a peek at how you solved a few things (It looks/works great btw). I am building a new Rails app which includes a chat interface with an LLM.

krschacht commented 1 month ago

@nwarwick Nice. It may be worth you knowing this then... one thing I really want to do is basically take almost all the code within this app and put it into an engine. That way people who want to build a new app like, say, "ChatGPT for data analytics" can mount the engine, override a few views, add some appropriate hooks like, say, after_message_processed.

I know of a few people using this in their startups, but they need to make so many changes/additions that they basically copy all code into their own app. And while I don't mind this at all, people who do this then won't benefit from all the future downstream changes I make.

So if this HostedGPT codebase had an appropriate abstraction in place (and I think engine's are the right abstraction) then it would set people up for deep customization without permanently forking/copying.

I don't know when I'll get to that, it'll be at least a couple months. But I mention this in case it impacts your plans! And, in case, maybe, possibly, perhaps, you decide you want to help with that. ;)

nwarwick commented 1 month ago

That's a great idea @krschacht! For our usecase we need to provide extra context via a RAG system which is in another app and deals with hitting OpenAI. I will definitely be keeping a close eye on this project and helping where I can!

krschacht commented 1 month ago

Nick, I've been considering adding langchainrb as the core library that this HostedGPT is built on top of. It has RAG built in. There was another contributor investigating this to see what would be involved in integrating langchainrb. Anyway, I just mention that in case it's helpful. I don't have a timeline for you on that one.\

Message ID: @.***>