.engram
Neural memory for AI systems. 400x faster search, spatial intelligence, graph relationships, and temporal decay — now with V2.2 map visualization.
Get Started Read WhitepaperWhat it does
Biological inspiration
Based on engrams (memory traces in neuroscience) with hierarchical organization and temporal decay algorithms.
HNSW performance
Hierarchical Navigable Small World indexing delivers O(log n) complexity with sub-millisecond search times.
Temporal intelligence
Automatic time decay and relevance scoring with hot/warm/cold/archive memory tiers.
How it works
Hierarchical trees
Organize memories in parent-child relationships with context-aware retrieval and nested associations.
Binary format
Self-contained .engram files with embedded HNSW index. Portable, efficient, no external dependencies.
Multi-modal content
Handle text, images, audio, code, and structured data with unified semantic search across all types.
Why it matters
Production ready
Real-world deployment processing 340+ session transcripts with 93.3% recall accuracy and 35+ hours uptime.
Performance breakthrough
400x improvement over traditional vector search. Sub-millisecond response times at scale.
Enterprise grade
AES-256 encryption, local-first architecture, comprehensive security features, and professional support.
Why it matters
Portability
Copy the file, copy the knowledge. No migrations, no exports, no vendor lock-in.
Privacy
Your data never leaves your machine. No cloud, no telemetry, no third-party access.
Simplicity
No Docker, no databases, no API keys required. Just a file and a Python script.