Show HN: Self-improving agent memory system, 92% R 5 LongMemEval, PostgreSQL

Category: ai-ml

Tags: ai-agents, memory-system, postgresql, vector-search, knowledge-graph

Score: 8.0/10 (Innovation: 8, Technical: 8, Documentation: 9, Utility: 7)

MemForge is a neuroscience-inspired memory system for AI agents that actively improves stored knowledge through sleep cycles, revision, and reflection. It's interesting because it moves beyond passive storage to create self-improving memory with hybrid search, knowledge graphs, and comprehensive metrics, all built on PostgreSQL without external dependencies.

Target audience: backend devs, ai engineers, devops

Repository: https://github.com/salishforge/memforge · TypeScript · MIT · 6 stars

View on Hacker News