Semantic Search Without the Cloud
A high-performance memory engine for Node.js, powered by Rust. Search by meaning, build knowledge graphs, and run multi-hop queries. Everything stays on your machine. No external APIs. No data leaving your environment.
Also available on crates.io for Rust projects · Apache-2.0 licensed
How It Works
Traditional vector databases send your data to external embedding APIs. Holographic Memory uses hyperdimensional computing to create high-dimensional binary vectors locally. Text becomes searchable by meaning without ever leaving your machine. The vectors are inherently lossy: even if intercepted, the original content cannot be recovered.
Capabilities
Semantic Search
Find content by meaning, not keywords. Store text as hypervectors and query by concept. Perfect for RAG pipelines, manuscript search, and knowledge bases.
Meaning Memory
Structured knowledge layer with role-filler algebra, triple stores, and multi-hop reasoning. Ask complex questions across your data and get traced, explainable answers.
Graph Engine
Explicit typed relations with multi-hop traversal, transitive and symmetric inference, and temporal filtering. Build knowledge graphs without Neo4j.
Privacy Preserving
Hyperdimensional vectors are inherently lossy. Original content cannot be reconstructed from the vector representation. Nothing leaves your machine.
Hybrid Retrieval
Three retrieval strategies in one engine: NSG proximity graphs for speed, inverted file indexes for scale, and sparse inverted indexes for precision.
Rust-Powered Performance
Core engine written in Rust with N-API bindings for Node.js. O(1) cached lookups, FxHash backend, and async worker thread execution.
Built For
Quick Start
import { HolographicMemorySystem } from "holographic-memory"; // Create a memory instance const hms = new HolographicMemorySystem(10000, "./storage"); // Store content by meaning await hms.memorizeText("sunset", "The sky blazed orange and pink"); await hms.memorizeText("storm", "Thunder cracked across the valley"); // Search by meaning const results = await hms.query("evening sky colors", 5); // → [{ id: 'sunset', similarity: 0.82 }]
Powers Scrivener MCP
Holographic Memory is the semantic search backend for Scrivener MCP, enabling writers to search their manuscripts by meaning, find thematic connections, and discover analogies across chapters.