Metadata search improvements

[khoroshevskyi]

How to improve metadata search in qdrant?

This issue is brainstorming possible solutions.

1. Add more text (metadata). -> problem is if text is too long, it will be truncated by model.
   Solution: multiple embeddings per bedfile. Chunk the metadata into chunks of 128/256. Then embed + upsert with bed_id.

2. Try another bigger models for search. e.g. BAAI/bge-large-en-v1.5 . Small ?

3. Divide search into multiple queries and then retrieve best results. I think qdrant allows you to do that.

4. ??

[khoroshevskyi]

In the release: v0.12.1 hybrid qdrant search was implemented that improved search results. Closing this issue for now