PostgresML
Postgres extension and platform that runs machine learning and LLM inference inside the database to simplify AI infrastructure.
What it does
PostgresML turns a Postgres database into an ML/AI platform by letting users train, serve, and query models directly with SQL. It combines vector search, embedding generation, and open-source LLM inference (like Llama and Mistral) in one system instead of stitching together separate microservices.
Core features
In-database model training and inference (regression, classification, clustering)
Vector embedding generation and storage
Fast ANN/KNN vector search with HNSW or IVFFlat indexing
Open-source LLM serving (Llama, Mistral, Mixtral)
Korvus SDK for building RAG pipelines
PgCat connection pooler for deployment
Best for
→Teams wanting to avoid separate vector database and ML microservices
→Developers building RAG search or chatbots on existing Postgres data
→Companies needing VPC-hosted AI infrastructure for compliance
→Engineers fine-tuning open-source LLMs on their own data
Tutorials
Step-by-step: exactly how to get things done with it.