post-graph-rag
github.com/crajah/post-graph-rag- Category
- Developer Tools
- Rank
- No. 1194Tools index
- Pricing
- Open Source
- Platform
- cli
- Type
- TOOL
- GitHub
- 3 stars
- Date
About
A Python library that implements GraphRAG natively inside PostgreSQL, using an LLM to extract entities and relations into a property graph stored alongside pgvector embeddings, then answering queries by fusing vector similarity, graph traversal, and full-text search. Its key feature is temporal fact resolution — newer documents can supersede or close earlier extracted facts rather than creating contradictory edges — and it reports benchmark results beating Graphiti and other GraphRAG systems on LongMemEval and ECT-QA.
What it can do
Extract entities and relations from documents using an LLM
Text documents → Property graph of entities and relations
Store knowledge graph and vector embeddings natively in PostgreSQL
Extracted graph data and embeddings → PostgreSQL database records
Answer queries by fusing vector similarity, graph traversal, and full-text search
User query → Query answer
Resolve temporal facts by superseding or closing earlier extracted facts with newer documents
New documents and existing facts → Updated fact graph without contradictions
Benchmark performance against other GraphRAG systems
Benchmark datasets (LongMemEval, ECT-QA) → Comparative performance results
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.