Module 1: Multi-Vector Embeddings in Qdrant

Connect to a local Qdrant instance or Cloud, depending on when you run your Qdrant instance.

Create a collection with a named multi-vector configuration. The multivector_config with MAX_SIM comparator tells Qdrant to use MaxSim for similarity computation.

Define the same documents from the previous notebook to index into Qdrant.

Upsert documents using models.Document for automatic FastEmbed embedding. Qdrant generates ColBERT multi-vectors without requiring to interact with the model directly.

Define the search query.

Search the collection using MaxSim. The query is also embedded via models.Document.

Storing vectors on disk