Provides a local, persistent long-term memory engine for AI coding assistants using hierarchical compression.
Sponsored
AHME is an Asynchronous Hierarchical Memory Engine designed to give AI coding assistants a robust, local long-term memory. Operating as a sidecar daemon, it efficiently compresses conversation history into a dense Master Memory Block using local Ollama models, ensuring zero cloud dependency, zero cost, and no lost context. It integrates seamlessly with any AI tool supporting the Model Context Protocol (MCP), allowing continuous context management across sessions and enabling AI assistants to start new conversations from an enriched, token-efficient checkpoint.
Key Features
01Seamless integration with MCP-compatible AI tools
021 GitHub stars
03Local, private long-term memory for AI coding assistants
04Asynchronous, CPU-idle background compression with local Ollama models
05Context-window replacement pattern for efficient token management
06Persistent SQLite queue for robust data handling and crash recovery
Use Cases
01Enhancing AI coding assistants with persistent memory across multiple conversations and sessions
02Maintaining continuous context for complex, multi-session coding projects without manual re-feeding of prior information
03Reducing token usage and context window limitations in LLM interactions by providing compressed summaries
