AHME

Background daemon that compresses chat conversations into structured JSON summaries using local Ollama models for persistent cross-session memory.

AHME (Adaptive Hierarchical Memory Engine) is a local sidecar daemon that automatically processes and compresses chat conversations into structured summaries. It partitions incoming text into token-accurate chunks, waits for system idle periods to avoid interfering with active AI workloads, then uses local Ollama models to compress conversations into structured JSON containing key facts, decisions, open questions, and entities. Implements hierarchical merging to recursively combine summaries into higher-level memory blocks, maintaining a persistent master memory file readable by any AI tool.