CortexGraph

Overview Link to heading

CortexGraph (formerly Mnemex) is a temporal memory system implementing human-like forgetting curves and memory consolidation patterns. Modeled after hippocampus-cortex interaction, it provides short-term to long-term memory transitions for AI systems.

Key Features Link to heading

Temporal Decay Link to heading

• Ebbinghaus forgetting curve implementation
• Spaced repetition for memory reinforcement
• Automatic garbage collection of unused memories
• Natural “Maslow effect” through conversation-based review

Memory Consolidation Link to heading

• Short-term memory (STM) with temporal decay
• Long-term memory (LTM) promotion for high-value memories
• Hippocampal bridge pattern for consolidation
• Integration with Obsidian vaults for permanent storage

Graph-Based Organization Link to heading

• Entity-relation knowledge graph
• Semantic similarity clustering
• Spreading activation for retrieval
• Explicit relation creation and tracking

Technical Implementation Link to heading

• Storage: JSON-based with optional Neo4j backend
• Embeddings: Semantic search with vector similarity
• MCP Server: Model Context Protocol integration
• Python API: Direct library usage for custom applications

Publication Link to heading

Published: PyPI package mnemex Test Coverage: 98%+ License: Apache 2.0

Links Link to heading

• GitHub Repository
• PyPI Package
• Documentation