What is Hybrid Search?

Hybrid search combines lexical and vector search into a single query, so your application can handle any kind of search your users throw at it, whether they know exactly what they want or are just describing it.

Rather than choosing between precision and meaning, hybrid search gives you both. A user searching for a specific SKU gets exact-match results. A user searching for "something warm and waterproof for camping" gets semantically relevant ones. The same search bar, the same query pipeline, handling both.

How it Works

Hybrid search isn't a separate feature to enable. Rather, it's a query-level decision. As long as your index is configured with both lexical and vector fields, your application can combine them in a single query, weight them against each other, and return a unified set of results.

This also means you're not locked into always doing both. Your application can make smart decisions based on what the user actually typed.

One word? Lean on lexical. Vector search on a single term casts too wide a net.

A full sentence? Pure vector search will pick up the nuance better.

Somewhere in between? Run both and combine the results. The flexibility is there; how you use it depends on what your users need.

When Hybrid Search is the Right Choice

Hybrid search is a good fit when your users don't all search the same way, which applies to most applications. If your catalog includes both structured identifiers and rich descriptions, or if your users range from power users who know exactly what they want to casual browsers describing a feeling, hybrid search lets you serve all of them well without compromising on either end.

What You'll Need

To use hybrid search, your index needs to be configured for both lexical and vector search from the start. The fields, mappings, and query structure all need to account for both.

See Vector Search for what that involves on the vector side.