Show HN: Graph RAG in Postgres – new documents retire older facts. No graph DB

Category: ai-ml

Tags: rag, graph-database, postgresql, pgvector, knowledge-graph, temporal-reasoning, llm

Score: 7.3/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 6)

post-graph-rag is a Python library that implements Graph RAG entirely on PostgreSQL using pgvector, with a novel temporal fact-retirement mechanism so newer documents can invalidate older facts. It combines vector search, property-graph traversal, and full-text search in one database, and ships with LongMemEval and ECT-QA benchmark results plus a companion post-graph library. The approach is compelling for teams that want graph-augmented RAG without operating a separate graph or vector store.

Target audience: backend developers, AI/ML engineers, data engineers

Repository: https://github.com/crajah/post-graph-rag · Python · Apache-2.0 · 10 stars

View on Hacker News