GPU acceleration for PostgreSQL vector index builds using NVIDIA CUDA/cuVS and VectorChord.
pgext.cloud · Source code · Documentation · License text · Control file
| Package | pgpu |
|---|---|
| Version | 2.1.0 |
| PostgreSQL versions | 14 · 15 · 16 · 17 · 18 |
| Repository | Unspecified |
| Extension type | Standard |
| Project status | Active |
| Requires CREATE EXTENSION | Yes |
| Requires preload | No |
| Trusted | No |
| Relocatable | No |
| Schema | pgpu |
| Stars | 42 |
| Latest commit | 2026-01-26 |
| Latest release | 2026-01-13 |
| Catalogue updated | 2026-08-30 |
| vector | vector data type and ivfflat and hnsw access methods |
| pg_cuvs | ⚡ Blazing-fast GPU vector search for PostgreSQL powered by NVIDIA cuVS. |
| pgcontext | Vector, filter-aware HNSW, and hybrid retrieval over authoritative PostgreSQL tables. |
| vectorize | The simplest way to do vector search on Postgres |
| pgmnemo | Single-plan multimodal agent memory for PostgreSQL |
| vchord | Vector database plugin for Postgres, written in Rust |
| smlar | Effective similarity search |
| pg_bestmatch | Generate BM25 sparse vector inside PostgreSQL |
| pg_similarity | support similarity queries |
| pg_sorted_heap | Sorted heap table AM with zone map scan pruning and built-in vector search |