Timescale

Timescale is a PostgreSQL-based cloud database company best known for its time-series work, but since 2024 it has pushed hard into AI with two open-source extensions: pgvectorscale, which adds a StreamingDiskANN index and statistical binary quantization for scalable vector search on top of pgvector, and pgai, which brings embedding management and model calls directly into SQL for RAG and agent workloads.

The pitch is consolidation: instead of running a separate vector database alongside Postgres, developers keep transactional data, time-series data, and embeddings in one familiar database. pgvectorscale is PostgreSQL-licensed and, per the company's benchmarks, targets 50M+ vector scale with disk-efficient indexes.

Limitations to note: performance claims come from vendor benchmarks and should be validated against your workload; pgvector itself has matured significantly (0.8.x with iterative scans and quantization), narrowing the gap; and at modest vector counts a plain pgvector setup is simpler.

Use Timescale Cloud, or install the open-source pgvectorscale and pgai extensions on self-hosted Postgres.