About brinicle

brinicle is a disk-first retrieval engine for building fast search without running a heavy search database.

It started as a low-RAM vector index engine, but it has grown into a broader retrieval layer for modern applications: vector search, structured item search, lexical search, semantic search, hybrid search, autocomplete, and query suggestions.

The idea behind brinicle is simple: many products need powerful retrieval, but they do not always need a full database platform around it.

A lot of teams already have their own metadata store, API layer, permissions, business logic, and deployment model. For them, adding a full vector or search database can mean extra memory usage, extra operations, and extra complexity for features they may not actually use.

brinicle takes a different path.

It gives you the search engine layer directly: a C++ core with a simple Python API, disk-first indexing, predictable memory usage, streaming-first ingest, and support for real update workflows like insert, upsert, delete, rebuild, and optimization.

Use VectorEngine when you already have embeddings.

Use ItemSearchEngine when you want lexical, semantic, or hybrid search over catalog-like records.

Use AutocompleteEngine when you need low-RAM query or title suggestions.

brinicle is not trying to be a full vector database. It does not try to own your metadata, authentication, replication, dashboards, or entire service layer.

That boundary is intentional.

brinicle is for builders who want retrieval to be local, controllable, resource-efficient, and easy to embed inside their own systems.

If you need a complete managed database platform, use one.

If you need a fast retrieval engine that stays close to your product and respects your RAM budget, brinicle is built for that.