A-mem  by WujiangXu

Agentic memory system for LLM agents research paper

created 6 months ago
492 stars

Top 63.6% on sourcepulse

GitHubView on GitHub
Project Summary

This repository provides the code implementation for the paper "A-mem: Agentic Memory for LLM Agents," introducing a novel system for LLM agents to dynamically organize and interact with their memories. It targets researchers and developers building advanced LLM agents that require sophisticated memory management beyond basic storage and retrieval. The system aims to enhance agent capabilities by enabling interconnected knowledge networks and continuous memory refinement.

How It Works

The Agentic Memory system organizes memories dynamically, inspired by Zettelkasten principles. When a new memory is added, it generates structured notes with attributes, contextual descriptions, and tags. It then analyzes existing memories to establish meaningful links based on similarity, creating interconnected knowledge networks. This approach allows for continuous memory evolution and agent-driven refinement, facilitating more adaptive and intelligent agent behavior.

Quick Start & Requirements

  • Installation: Clone the repository and install dependencies using pip install -r requirements.txt within a Python virtual environment (venv or Conda).
  • Prerequisites: Python 3.9+ is recommended.
  • Running Experiments: Execute experiments using python test_advanced.py.
  • Documentation: Refer to the paper for detailed information.

Highlighted Details

  • Reproduces results from the paper "A-mem: Agentic Memory for LLM Agents."
  • Employs Zettelkasten principles for dynamic memory organization.
  • Features intelligent indexing, linking, and structured note generation.
  • Demonstrates superior performance over SOTA baselines on six foundation models.

Maintenance & Community

This repository is an implementation for a specific paper. For ongoing development and broader agentic memory system usage, refer to the official implementation at https://github.com/agiresearch/A-mem.

Licensing & Compatibility

Licensed under the MIT License. This permissive license generally allows for commercial use and integration into closed-source projects.

Limitations & Caveats

This repository is specifically designed for reproducing paper results and may not be a production-ready library for general agent development. For building agents, the official implementation at agiresearch/A-mem is recommended.

Health Check
Last commit

4 days ago

Responsiveness

1 day

Pull Requests (30d)
0
Issues (30d)
3
Star History
89 stars in the last 90 days

Explore Similar Projects

Feedback? Help us improve.