Hi! I’ve tried to use deepwiki several times, and it’s wrong so much of the time I think having the link on our page may be actively harmful, as it will confuse people. Asking OpenHands is way more effective at getting good answers. Does anyone else use deepwiki? If not maybe we could remove the badge from our README?

Any specific bug reports we should file to them? Would be nice if it "just worked", since it seems to be kinda popular.

It’s just not thorough. It misses important information all the time

I’ve used deepwiki several times (on other repos), but only the “deep research” feature since it’s free.

Did it work well for you?

It works when I know something exists and I just need to find where the code is

OK, interesting. It doesn’t even work for me in that case (unless I know the right keywords, in which case I can just use github search)

it doesn’t always work but I mean, it’s free 🤣 github search returns too many results for a huge repo even if I know the keyword

https://github.com/AsyncFuncAI/deepwiki-open ? 🙂 and we can customize the thoroughness

I’ve had good results with Docusaurus while using it as a context source, not to suggest a replacement, ymmv, etc.

Hey @polite-architect-52194, yeah, the deepwiki-open looked pretty cool. tbh I’d love for us to just be able to ask OpenHands a question (which would consolidate our tech stack), but that still may be a bit slower than deepwiki no matter how hard we try

https://github.com/langchain-ai/chat-langchain for quick Q&A, this could work? (I know you're not a fan of RAG, but it might be faster? 😉) and for active OpenHands Q&A while inside OpenHands when developing OpenHands, perhaps https://github.com/langchain-ai/mcpdoc could come in useful targeted at OpenHands' docs?

Yeah possibly! I'm a fan of rag when it has benefits, so an examination of any of these would be welcome!