Squish – a local memory runtime for AI agents (squishplugin.dev)

Squish has launched a local memory runtime designed for AI agents, allowing developers to build applications with persistent memory without the need for a crypto token or hosted services. This innovative solution features bi-temporal memory storage that tracks validity and expiration dates, smart context capture for important information, and hybrid search capabilities combining BM25 and vector search. The architecture prioritizes user privacy by ensuring all data remains local, while also supporting collaborative settings through PostgreSQL integration. The significance of Squish lies in its promise to enhance the functionality of AI agents by providing a more intelligent and adaptive memory system. Unlike traditional methods, Squish enables seamless memory access across different AI frameworks, making it a versatile tool for developers. Its open-source nature not only fosters community collaboration but also ensures users avoid vendor lock-in, promoting ownership over their AI agents' memory. Overall, Squish presents a forward-thinking approach that enhances the capabilities of AI while preserving privacy and freedom for developers.