Agile Mind

Overview

Agile Mind is an AI-powered development platform that builds software repositories from natural language descriptions. It uses a LLM-based multi-agent architecture to automate the software development process, from requirements gathering to code generation and documentation.

Features

• Multi-Agent Architecture: Specialized AI agents for different development tasks
• Code Generation: Automated creation of code from requirements or descriptions
• Collaborative Development: Agents can work together to solve complex programming challenges
• Documentation: AI-generated documentation that stays in sync with code
• Checking: Automated code review and static analysis

Online visualization

Check out the online visualization of AgileMind!

Note that the online version does not support human-agent interaction and parallel processing yet. Consider using the command line version for those features.

Usage

git clone https://github.com/wnrock/AgileMind.git
cd AgileMind

pip install -r requirements.txt

streamlit run web.py

Quick Start

Set Up Environment Variables

1. Use .env file

cp .env.template .env
# Then replace the placeholder values with actual credentials

2. Set environment variables manually

export OPENAI_API_KEY="<Your_API_key>"
export OPENAI_BASE_URL="<Your_OpenAI_API_base_url>" # Optional

3. Pass as command-line arguments

Currently only supported for Docker. Check the Docker section below.

Get Started

1. From PyPI

pip install AgileMind

agilemind "Create a 2048 game with UI" -o output

2. Docker

docker run -it                                      \
    -e OPENAI_API_KEY="<Your_API_key>"              \
    -e OPENAI_BASE_URL="<Your_OpenAI_API_base_url>" \
    -v <Your_output_dir>:/agilemind/output          \
    ghcr.io/wnrock/agilemind:latest                 \
    "Create a 2048 game with UI"                    \
    -o output

3. From source

git clone https://github.com/wnrock/AgileMind.git
cd AgileMind

pip install -r requirements.txt

python app.py "Create a 2048 game with UI" -o output