Mnemo

About

Mnemo serves as a robust, open-source AI memory solution designed to give large language models and AI agents a long-term, persistent memory. It combines full-text search with semantic search (powered by embedded LiteLLM models), enabling sophisticated retrieval of stored information. The tool supports seamless data synchronization across various cloud storage providers like Google Drive or S3 through an embedded rclone process, making AI memory portable and consistent across different environments. It also features 'proactive memory' where tool descriptions guide AI to intelligently save relevant information such as preferences, decisions, and facts.

Key Features

• Hybrid search combining FTS5 full-text and sqlite-vec semantic capabilities.
• Embedded rclone for managing and synchronizing memory data across multiple machines or cloud storage.
• Proactive memory feature that uses tool descriptions to guide AI in saving relevant information.
• 1 GitHub stars
• Automatic detection and configuration of embedding models based on provided API keys.
• Zero-dependency mode for full-text search without requiring external API keys.

Use Cases

• Equipping AI agents or chatbots with long-term memory to remember past interactions, user preferences, and learned facts.
• Synchronizing AI's knowledge base across different computing environments or instances for consistent behavior.
• Developing AI applications that require efficient retrieval of information through both keyword-based and contextual semantic searches.