Skip to content
This repository has been archived by the owner on Jul 4, 2024. It is now read-only.

Latest commit

 

History

History
61 lines (41 loc) · 2.38 KB

README.md

File metadata and controls

61 lines (41 loc) · 2.38 KB
Raw

Turbo: An Enhanced Auto-GTP v0.4.7 Fork

Turbo is a fork of Auto-GTP v0.4.7, offering the following improvements:

Features

  • Cost and Speed: GPT-4 support that's 66% cheaper and 2X faster than the original Auto-GPT v0.4.7.
  • Improved Code Generation: The 'coder.v2' persona offers improved support for coding projects.
  • Multiple Personas: Choose from specialized configurations tailored for specific tasks.
  • Multi-Agent Support: Deploy additional agents to accomplish specific tasks.
  • Composability: Configure multi-agent setups and custom personas easily.

Compatibility

  • Optimized for: gpt-4-0314
  • Variable Performance: Results may differ when used with other GPT-4 models.
  • Contributions: Pull Requests for optimization with other models are welcome.

Installation and Setup

Git clone the Turbo repository, or use one of Auto-GPT's installation methods.

Research Interests

This fork aims to make certain aspects of AutoGTP more composable, makign it easier to test various scenarios with Auto-GPT, focusing on:

  • Multi-persona prompting
  • Multi-agent configurations
  • Memory extension techniques
  • Auto-GPT as a conductor or orchestrator of other agents

Performance Improvements

While GPT-4 offers superior performance compared to its predecessors, it comes at a higher cost and slower speed. Turbo optimizes various aspects of the interaction between the agent and the language model, delivering significant time and cost savings. Compared to Auto-GPT v0.4.7, Turbo can up up to 66% cheaper, and twice as fast.

Coding with Turbo

Turbo aims to extend Auto-GPT's coding capabilities with optimized preset prompts, and specialized agents. Coder.v2 provides further improvements in this area.

Usage:

python -m autogpt --persona='coder.v2'

Check out the examples folder for sampel code outputs.

Personas

Specialized configurations, called "personas," can be loaded to tailor Auto-GTP's behavior for specific tasks.

Usage:

python -m autogpt --persona=coder.v2

Available Personas:

  • coder.engineer
  • coder.v2
  • turbo (default)

Dolly: Multi-Agent Plugin

Turbo comes bundled with Dolly, a multi-agent plugin now upgraded with the ability to reference personas. More Details