smartgpt  by Cormanz

LLM agentic framework for complex task completion via plugins

created 2 years ago
1,760 stars

Top 24.9% on sourcepulse

GitHubView on GitHub
Project Summary

SmartGPT is an experimental Rust program designed to empower Large Language Models (LLMs) like GPT-3.5 and GPT-4 to tackle complex, multi-step tasks autonomously. It targets developers and researchers seeking advanced LLM orchestration capabilities, offering modularity and flexibility through a plugin system and configurable "Autos."

How It Works

SmartGPT employs a two-part agent architecture: a Dynamic Agent and a Static Agent. The Dynamic Agent uses a REACT-like loop for thinking, reasoning, and deciding whether to brainstorm, execute an action, or provide a final response. When an action is chosen, the Static Agent takes over, meticulously planning and executing a sequence of tool calls with precise arguments. This separation allows for complex task decomposition and execution, with memory management via a VectorDB to retain context across tasks.

Quick Start & Requirements

  • Install Rust (via rustup).
  • Clone the repository: git clone https://github.com/Cormanz/smartgpt.git && cd smartgpt
  • Run in release mode: cargo run --release
  • Configuration is managed via config.yml.
  • Documentation: https://github.com/Cormanz/smartgpt

Highlighted Details

  • First-class plugin support for modularity.
  • Dynamic and static agent architecture for task execution.
  • Memory management using VectorDB for long-term context.
  • Supports "Runner" and "Assistant" Auto types.

Maintenance & Community

  • Active development, though the project is described as experimental.
  • Discord community available for discussion and contribution.
  • Core maintainer is a high school student; development funding is sought via Patreon.

Licensing & Compatibility

  • MIT License.
  • Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The project is highly experimental, with stability and backward compatibility sacrificed for innovation. The memory system is currently simple and limited, and the ecosystem of tools and integrations is less mature compared to established projects like AutoGPT. Testing is primarily conducted with GPT-3.5 due to cost considerations.

Health Check
Last commit

1 year ago

Responsiveness

1 day

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

Explore Similar Projects

Feedback? Help us improve.