Find answers from the community

Updated 2 months ago

hey @Logan M I am trying to run the code

hey I am trying to run the code for image to image retrieval and reasoning using GPT4V, and I am using LanceDB instead of Qdrant. but while retrieving it shows this error:
this error doesn't come up with qdrant, but pops up in lancedb. Any hel would be great.
Attachment
imgtoimg.PNG
k
L
A
19 comments
what i think is that, the integration hasn't added support for file_path or img_path to be taken as an input.
The LanceDBVectorStore implementation doesn't seem to be retrieving metadata properly

Probably the class needs a little bit of TLC 😅

Ideally we can get the _node_content field being stored and reconstruct the node completely. But the code is currently not doing that
let's work on the lancedb class then
It's not quite as popular as other vector dbs, which is why this probably got missed.

If you want to open a PR, that would be amazing ❤️
https://github.com/run-llama/llama_index/blob/main/llama_index/vector_stores/lancedb.py
hey @Logan M , it would be great if u could help on this, it would be great coz..neither i wrote the integration nor i know llama-index from inside that much.
i am getting this error when i try the pdf tables example:
Attachment
theh.PNG
and i get this error, while trying the image captioning example:
Attachment
egerg.PNG
I think whoever wrote the integration for lancedb into llama index..missed some aspects/ functionalities.
it would be really helpful if you could help me on with this, so we can wrap this up sooner.
That second error is unrelated to lancedb, I think it was actually just fixed and merged into main (it happens when tables don't have the same number of columns extracted per row).
Most of our vector stores our community contributed -- I am definitely not a lancedb expert, and it's also not as popular
Its a free and open source package, so anyone is free to improve it. We are all software developers here 😉

I can try to take a look I'm a bit, no promises
I understand..I am not pointing it to anyone
i am from the lancedb team and we have been trying to improve our usage
thanks ..i'll send you the PR soon
@AyushC take a look here.
Hey @Logan M , sorry for the confusion here. I'll take over all lancedb related issues from now. I think the integration was done by community members so we haven't been able to track it. Some of our community members have been including Kaushal here have been reporting some errors so I thought of joining this discord.
Please feel free to tag me here or on github (ayushexel) if something comes up. For now I think we're good based on the above messages?
Thanks @AyushC ! Yea just let me know if you have any questions. Happy to merge an PRs you make though ❤️
Add a reply
Sign up and join the conversation on Discord