Show HN: Local-first long-term memory engine for AI agents·MCP/CLI· 100% local (github.com)

A new tool called **memoirs** has been introduced, providing a local-first long-term memory engine for AI agents. This innovative system allows AI agents to maintain a single, persistent memory layer across various sessions, IDEs, and models without relying on cloud services. It aggregates user transcripts and extracts vital information such as preferences and decisions using a local large language model (LLM), yielding efficient retrieval times of around 6.2 milliseconds. By utilizing SQLite and a 2GB GGUF model, memoirs runs entirely on local machines, ensuring data privacy with no external API calls or cloud dependencies. This development is significant for the AI/ML community as it enhances agents' contextual awareness and memory retention capabilities, addressing common limitations associated with memory systems. Memoirs boasts features like hybrid retrieval methods, temporal queries, auto-pruning, and explainability, all designed to enrich agent interactions. With its emphasis on privacy and user control, memoirs also supports various data formats for ingestion and is extensible with a comprehensive API and user-friendly web interface. By outperforming cloud-based solutions in retrieval metrics and speed, memoirs paves the way for more powerful and privacy-conscious AI applications.