[Question & Feat] Will LEANN support video rag?

[AgainstEntropy]

What problem does this solve?

Video RAG

Proposed solution

video indexing + VLMs that support video inputs

Example usage

[yichuan-w]

Yeah, we will support soon, right now the image support is ready, we incorporate multivector embedding, which is much more fine-grained

https://github.com/yichuan-w/LEANN/blob/main/apps/multimodal/vision-based-pdf-multi-vector/multi-vector-leann-similarity-map.py

We are actively developing more features to support more modalities