Hybrid Search in Qdrant

Introduction

I am currently working on Mandelbaum and we're using qdrant as our vector database. We're storing products in our qdrant db in order to allow users to semantically search through their product catalogue to give them the best search experience in ecommerce stores. We use the product title and description as well as some meta information to create embeddings from this data and store those in qdrant. If the user then conducts a search we use the user's query, embed this as well using the same embedding model and then use qdrants search engine to find the most relevant results.

This works quite well from the get-go but has some serious caveats: Since