Hey HN, we want to share HelixDB helix db, a project our team has been working on to solve the persistent memory problem for AI agents. HelixDB is a fully native Graph-Vector Database built in Rust, designed to handle both complex relational data and semantic vector embeddings in a single, high-performance engine. Why a hybrid? While vector databases are excellent for fuzzy similarity queries, many advanced AI-driven applications require both similarity and precise, multi-hop relationship queries to maintain logical continuity. Relying on disconnected systems forces developers to deal with complex data synchronization, leading to latency and architectural fragility. HelixDB bridges this gap by providing a unified memory layer where vectors and graphs coexist natively. Here are some specific ways our architecture helps: - Complex Agent Reasoning: Use graph traversals to ensure agents respect hard logic constraints while using vector search for context retrieval. - Multi-Session State Persistence: Maintain long-term episodic memory for autonomous agents that survive restarts without losing historical context. - Enterprise Knowledge Management: Map organizational documents via knowledge graphs while enabling semantic discovery across massive document sets. Many developers are skeptical of adding 'yet another database' to their stack. We hear you, but we built HelixDB because the overhead of stitching together a separate vector store with a graph database often results in synchronization delays and data silos. By building this natively in Rust, we have streamlined the stack, allowing you to perform vector-indexed graph traversals directly within the database runtime. In terms of performance, our internal benchmarking shows that we are on par with dedicated vector databases like Pinecone and Qdrant for semantic search, while our graph traversal engine is up to three orders of magnitude faster than traditional solutions like Neo4j for specific multi-hop analytical workloads. You can see our initial architectural approach in our docs: introduction. If you'd like to try it out, you can follow this guide to build a simple GraphRAG demo: getting started. We are very eager to hear your thoughts—comments and feedback are more than welcome!