memsolus
Blog›Engenharia›Hybrid Search: Combining Semantics and Keywords
Engenharia

Hybrid Search: Combining Semantics and Keywords

How Memsolus hybrid search combines pgvector, tsvector and reranking for precise results.

EM
Equipe Memsolus
20 Mar 2026·7 min read·Engenharia
userIduser_ab291fativo
query"preferências de linguagem do usuário"busca semântica
score0.94 · cosine similaritytop 1
latência112ms · índice HNSWp99

Semantic search with pgvector

pgvector stores vector embeddings that enable similarity-based semantic search, going beyond exact word matching.

Keyword search with tsvector

PostgreSQL tsvector complements semantic search with precise term matching.

Reranking

A reranking model combines the results of both searches and produces a list ordered by real relevance.

Tópicos:#Busca Híbrida#pgvector#Pipeline
EM
Equipe Memsolus
Time de engenharia e produto do Memsolus.
Read in another language
PortuguêsEnglishEspañol

Stay in the loop.
No spam, just content.

New articles about memory for AI, tutorials and product updates — straight to your inbox, every week.

Your data is safe. Unsubscribe anytime.
Ready to get started?

Give persistent memory to your AI agents.

Open-source SDK, REST API, and vector search. Integration in 3 lines of code. Free to start, no credit card required.