Skip to content

Hybrid Indexing

Combining multiple index structures — such as a vector graph index alongside an inverted keyword index — to serve different query types from the same data store.

Hybrid indexing means maintaining more than one type of index over the same data so that different query types can each be served efficiently. A common example is keeping a vector graph index for semantic similarity alongside an inverted keyword index for exact-term matching, both built over the same documents.

This is what makes hybrid search practical. To combine dense vector retrieval with sparse keyword retrieval, the database needs both a structure optimised for nearest-neighbour search and one optimised for term lookups. By maintaining both, it can run each kind of query against the index best suited to it, then fuse the results into a single ranking.

Hybrid indexing can also refer to combining multiple vector index strategies — for instance pairing a coarse cluster-based index with a fine graph index, or mixing approximate and exact execution depending on filter selectivity. In all cases the principle is the same: rather than forcing one index to serve every query, the system keeps complementary structures so each query takes the most efficient path, at the cost of extra storage and maintenance.