lightweight-memory-augmented

Reduce inference tokens by up to 38× and API calls by 30× through three-stage memory system: sensory compression, short-term consolidation, and offline long-term storage.

Independent GitHub snapshot · Automated classification · Original documentation governs