Show HN: Memory Vault – local-first memory, hybrid search, knowledge graph (github.com)

Memory Vault, a new self-hosted memory database designed for AI applications, has been released as its first stable version (v1.0). Built on PostgreSQL and utilizing the pgvector extension, it supports hybrid search combining semantic similarity and keyword matching. This innovation allows AIs like Claude or ChatGPT to maintain memory across sessions, storing relevant information such as decisions and project context. Users can interact with this memory using a local language model (LLM) without relying on cloud services, ensuring data privacy while providing real-time access during conversations. The significance of Memory Vault lies in its ability to resolve a common limitation faced by AI applications: the lack of memory persistence. This new framework enables AI models to recall past interactions and data, greatly enhancing their utility. Key features include a hybrid search engine, a REST API for integration, and a built-in knowledge graph that dynamically captures relationships between entities. With a streamlined setup process and comprehensive API endpoints, developers can easily implement custom applications that leverage consistent, contextual memory in AI interactions, marking a step forward in creating more sophisticated and useful AI systems.