Pensyve

•
102
•

Empowers AI agents with a universal, durable memory runtime for learning, remembering, and knowledge sharing.

Pensyve is a robust, framework-agnostic memory runtime designed to give AI agents durable intelligence. It solves the critical problem of agents losing context between sessions by providing episodic, semantic, and procedural memory types, handling multimodal content, and employing an advanced 8-signal fusion retrieval system. Built with a Rust core, it offers offline-first operation with zero cloud dependencies, featuring FSRS-based memory decay, consolidation, and scalable storage options, ensuring agents can learn continuously and effectively.

Key Features

01FSRS-based memory decay with retrieval-induced reinforcement for dynamic forgetting and strengthening.
02Multimodal content support for text, code, images, and structured data.
03Offline-first design with a Rust core, SQLite storage, and optional Postgres backend.
041 GitHub stars
058-signal fusion retrieval, combining vector, lexical, graph, and temporal signals for precise recall.
06Three memory types: Episodic (what happened), Semantic (what is known), and Procedural (what works).

Use Cases

01Enabling agents to learn from past actions and outcomes through procedural memory.
02Equipping AI agents with long-term, intelligent memory capabilities across sessions.
03Developing AI applications requiring durable, multimodal memory that operates without cloud dependencies.