Hey HN,
We just launched VeritasGraph, an open-source RAG framework that swaps out pure vector search for a knowledge graph backbone—and the response has been exciting: 3,300+ visitors and 130 GitHub stars in just 5 days.
Why we built it:
Vector RAG ≠ enough → Great for simple lookups, but it breaks on multi-hop questions that need connecting facts across docs.
Verifiable attribution → Every AI output is traced to its original source, making results auditable and reducing hallucinations.
Data sovereignty → 100% on-prem, no vendor lock-in, no third-party data exposure.
We want enterprise AI to be trustworthy, private, and explainable—not just “accurate enough.”
Repo: github.com/bibinprathap/VeritasGraph
Would love HN’s thoughts:
Are you running into the limits of vector-only RAG?
How are you tackling explainability and trust in your AI apps?
We just launched VeritasGraph, an open-source RAG framework that swaps out pure vector search for a knowledge graph backbone—and the response has been exciting: 3,300+ visitors and 130 GitHub stars in just 5 days.
Why we built it:
Vector RAG ≠ enough → Great for simple lookups, but it breaks on multi-hop questions that need connecting facts across docs.
Verifiable attribution → Every AI output is traced to its original source, making results auditable and reducing hallucinations.
Data sovereignty → 100% on-prem, no vendor lock-in, no third-party data exposure.
We want enterprise AI to be trustworthy, private, and explainable—not just “accurate enough.”
Repo: github.com/bibinprathap/VeritasGraph
Would love HN’s thoughts:
Are you running into the limits of vector-only RAG?
How are you tackling explainability and trust in your AI apps?