Generative agent implementation of a research paper
Top 93.3% on sourcepulse
This repository implements the "Generative Agents: Interactive Simulacra of Human Behavior" paper, enabling the creation of AI agents that exhibit human-like behaviors. It targets researchers and developers interested in agent-based simulations and interactive AI, offering local LLM support and enhanced features over the original paper's concepts.
How It Works
The project leverages Langchain for agent orchestration and Guidance for efficient LLM interaction, specifically focusing on prompt management and output control. It supports local LLMs, allowing for private and customizable agent behavior without relying on external APIs. Key features include memory retrieval, reflection, planning, and dialogue generation, aiming to closely replicate the paper's simulation capabilities.
Quick Start & Requirements
pip install guidance==0.0.63 langchain==0.0.190
Highlighted Details
Maintenance & Community
The project appears to be a personal implementation by QuangBK. No specific community channels or active maintenance signals are detailed in the README.
Licensing & Compatibility
Limitations & Caveats
Planning and dialogue generation features are noted as needing improvement. A Gradio Web UI is planned but not yet implemented. The project recommends a specific, older version of Guidance (0.0.63
) due to stability concerns with newer versions.
1 year ago
Inactive