RuVector is a vector database and memory substrate for AI agents built in Rust.
RuVector addresses the problem of persistent, adaptive memory for AI systems that need to retain information across sessions. It combines local semantic embeddings with persistent vector retrieval, graph relationships, and explicit feedback learning. The tool runs entirely locally by default, downloading and caching embedding models on first use, with no requirement for external database servers or API keys. Memory persists to disk under the project directory and can be recovered by reopening the same storage path in subsequent processes. The system treats learning as an explicit process driven by recorded outcomes and feedback rather than by read operations alone.
The tool suits developers building AI agents that need to remember context between invocations and want to avoid external service dependencies. It provides two typed memory layers: AgenticMemory, which unifies working, episodic, semantic, and procedural memory into a single runtime API, and AgenticDB, which persists specific agent artifacts including episodes, skills, causal edges, and policy state. The core store itself is general-purpose and supports multiple embedding providers including ONNX and Candle. Developers should note that the unified manager for AgenticMemory currently operates in-memory and its cross-type consolidation method remains incomplete.
The project receives issues from both core maintainers and external users, indicating adoption beyond the immediate team without an overwhelming support burden. Maintainers typically respond to new issues and pull requests within a few days. Work in the issue tracker centers on enhancements, bug fixes, and documentation.
