How ruvnet/ruvector works
RuVector is a Rust-centered substrate for local vector retrieval and persistent AI-agent memory, with Node.js and CLI surfaces. Its documented memory loop captures and embeds information, persists vectors and relationships, recalls relevant context, and uses outcomes for learning; memory lifecycle and shared-memory features extend that loop. The graph preserves major documented capabilities, including graph retrieval, SONA learning, and shared memory. Source excerpts are limited, so documented links cite the README; the CLI storage implementation and core storage behavior are the sampled source details.
Main entry points: storage.rs Vector store · ruvector-core Embedding providers · storage.ts Memory CLI · engine.rs Typed decisions
Components 14
Capture and encoding
- Embedding providers
crates/ruvector-core - Typed decisions
crates/ruvector-typesafe-core/src/engine.rs
Memory and retrieval
- Vector store
crates/ruvector-core/src/storage.rs - Graph relationships
crates/ruvector-graph/src/storage.rs - Similarity retrieval
crates/ruvector-core - Typed agent memory
crates/ruvector-core
Learning and lifecycle
- Outcome feedback
crates/sona/src/engine.rs - SONA learning
crates/sona/src/engine.rs - Memory lifecycle
crates/ruvector-snapshot - Shared memory
crates/mcp-brain/src/pipeline.rs
User interfaces
- Memory CLI
npm/packages/cli/src/storage.ts - Node.js API
crates/ruvector-core
Other components
- Agent or user
- Agent decision
Connections 19
- Agent or user → Memory CLI: captures
README.md - Agent or user → Node.js API: uses
README.md - Memory CLI → Embedding providers: encodes text
README.md - Node.js API → Embedding providers: encodes text
README.md - Embedding providers → Vector store: produces vectors
README.md - Memory CLI → Vector store: persists memories
README.md - Node.js API → Vector store: inserts vectors
README.md - Vector store → Similarity retrieval: searches vectors
README.md - Graph relationships → Similarity retrieval: supports graph search
README.md - Similarity retrieval → Agent decision: returns context
README.md - Agent decision → Outcome feedback: records outcomes
README.md - Outcome feedback → SONA learning: trains from feedback
README.md - SONA learning → Vector store: adapts ranking
README.md - Vector store → Memory lifecycle: snapshots or compacts
README.md - Vector store → Typed agent memory: persists typed memory
README.md - Typed agent memory → Embedding providers: uses providers
README.md - Agent or user → Typed decisions: requests decision
README.md - Typed decisions → Embedding providers: embeds inputs
README.md - Typed agent memory → Shared memory: contributes memory
README.md