Hey HN, we want to share HelixDB helix db, a project our team has been building to rethink how we handle AI data. Why do we need yet another database? Standard RAG pipelines often force developers to juggle a vector store, a graph database, and a full-text search engine simultaneously. This fragmentation leads to complex sync logic, data staleness, and massive operational overhead. HelixDB solves this by providing a fully native Graph-Vector Database that handles relationships, semantic vectors, and full-text search in one engine. Our benchmarking shows that we are on par with Pinecone for vector latency, and our graph traversal performance is up to three orders of magnitude faster than Neo4j in high-concurrency environments. - Knowledge Graph RAG: You can execute multi-hop reasoning by traversing relationships while simultaneously filtering by vector similarity. - Unified Search: Index codebases by embedding snippets for semantic search while keeping the function-call graph structure intact for precise dependency analysis. - Real-time Synchronization: Eliminate the need for fragile ETL pipelines; data written to the graph is immediately available for vector and BM25 queries. Many people have a thing against “yet another query language,” but we went ahead and designed a custom integration layer because it allows our Rust-based engine to perform vector-graph joins at the storage level. By bypassing the traditional application-tier join bottleneck, we achieve a 10x reduction in latency for complex agentic queries. You can see the architecture details in our documentation here: introduction. We’ve also integrated advanced research on multi-agent memory layers, which you can read about here: arxiv.2606. For those who want to see it in action, check out this video demonstration: szgQu3yq. If you’d like to try it out in a simple RAG demo, you can follow this guide: rag demo. Many thanks! Comments and feedback welcome!