Hey HN, we want to share HelixDB, a native graph-vector database designed to handle the complex relationships and semantic data inherent in professional networks. Why a hybrid? Most RAG applications rely solely on vector search, which is great for semantic similarity but fails when you need to answer relationship-dependent questions. If an AI agent needs to find a developer who worked on a specific project with a certain lead, vector search returns fragmented documents rather than an explicit connection. HelixDB solves this by natively combining a property graph engine with vector search, allowing for multi-hop reasoning alongside semantic matching. Here are some specific use cases where this architecture shines: - Talent discovery: Mapping complex organizational hierarchies and project histories to identify experts based on both explicit reporting lines and latent skill sets. - Intelligent recommendations: Traversing professional networks to suggest team compositions by analyzing shared past project nodes and similar technical expertise. - Enterprise knowledge management: Linking unstructured project documentation with explicit employee data to provide context-aware answers that standard RAG systems miss. Some might be skeptical about 'yet another database', but we implemented this natively in Rust because stitching together separate graph and vector stores creates unacceptable synchronization latency and operational overhead. Our benchmarking shows that we are on par with Pinecone and Qdrant for raw vector similarity, while our graph traversal performance is up to three orders of magnitude faster than traditional approaches like Neo4j for deep, multi-hop queries. Check out our technical deep dive here: introduction. If you’d like to try it out in a simple RAG demo, you can follow this setup guide: examples. Many thanks! Comments and feedback are more than welcome!