Prev Next

Database / Weaviate Vector database Interview questions

Explain the lifecycle of a hybrid search query in Weaviate?

A hybrid search request involves running two independent searches internally and then fusing their results, all within the scope of handling that single client call.

sequenceDiagram participant Client participant Weaviate participant Vectorizer as Vectorizer Module participant VectorIdx as Vector Index (HNSW/Flat) participant Inverted as Inverted Index (BM25) Client->>Weaviate: hybrid(query, alpha, limit) Weaviate->>Vectorizer: Embed the query text Vectorizer-->>Weaviate: Query vector Weaviate->>VectorIdx: Vector similarity search VectorIdx-->>Weaviate: Ranked vector results Weaviate->>Inverted: BM25 keyword search Inverted-->>Weaviate: Ranked keyword results Weaviate->>Weaviate: Fuse both ranked lists (rankedFusion/relativeScoreFusion, weighted by alpha) Weaviate-->>Client: Final fused top-K results

When a hybrid query arrives, Weaviate first embeds the query text using the collection's configured vectorizer (unless a raw vector was supplied directly), then runs a vector similarity search against the vector index and, in parallel, a BM25 keyword search against the inverted index. Each produces its own independently ranked list of candidate objects.

Those two lists are then combined using the configured fusion algorithm (rankedFusion or relativeScoreFusion), weighted according to the alpha parameter, and the resulting single fused ranking is what's actually returned to the client as the final top-K results, meaning the client never sees the two intermediate result lists separately, only the merged outcome.

During a hybrid search, Weaviate runs:
The fusion step that combines vector and BM25 results is weighted by:

More Related questions...

What is Weaviate? What is a Collection in Weaviate? What is a vectorizer module in Weaviate? What is hybrid search in Weaviate? What is the HNSW index in Weaviate? What is a Flat index in Weaviate? What is the Dynamic index in Weaviate? What is a cross-reference in Weaviate? What are named vectors in Weaviate? What is Weaviate Cloud? What is quantization/compression in Weaviate, and why is it used? What is Rotational Quantization (RQ) in Weaviate? What is generative search (RAG) in Weaviate? What is multi-tenancy in Weaviate? What is Object TTL in Weaviate? What is the Weaviate Query Agent? What is a UUID's role for objects in Weaviate? What are properties in a Weaviate collection? What is BM25 in the context of Weaviate? What is the alpha parameter in Weaviate hybrid search? What are the main deployment options for Weaviate? What is replication in Weaviate? What is sharding in Weaviate? What are the main use cases for Weaviate? What is the difference between bringing your own vectors and using a vectorizer module? Explain the data flow of an object being vectorized and indexed in Weaviate? Why does Weaviate combine BM25 and vector search instead of using vector search alone? How does Weaviate differ from Milvus? What is the difference between HNSW and the HFresh index in Weaviate? How do you choose between PQ, BQ, SQ, and RQ quantization? When should you use the Dynamic index instead of always using HNSW? How do you troubleshoot poor recall after enabling quantization in Weaviate? What is the difference between rescoring and raw compressed-vector search? How does Weaviate's multi-tenancy isolate tenant data on disk? Explain the internal working of Weaviate's rescoring mechanism for quantized vectors? What is the difference between Weaviate and Pinecone? How do you implement RAG using Weaviate's generative search module? Why use cross-references sparingly, according to Weaviate's own guidance? What is the difference between rankedFusion and relativeScoreFusion in hybrid search? How does the Weaviate Query Agent route natural-language questions across collections? When would you choose multi-vector (ColBERT-style) embeddings over single-vector embeddings? How do you configure replication factor for high availability in Weaviate? What is the difference between Weaviate's REST/GraphQL API and its gRPC API? Explain the lifecycle of a hybrid search query in Weaviate? How do you optimize Weaviate for memory efficiency at large scale? What is the difference between PQ and RQ quantization internally? How does Weaviate decide which shard(s) to query for a given request? Why should you avoid excessive cross-reference traversal in a single query? What is the difference between Weaviate Database (self-hosted) and Weaviate Cloud? How do you troubleshoot a Weaviate collection running out of memory at scale?
Show more question and Answers...


Comments & Discussions