Minimal Provider Agent Market

A Python-based service that interacts with the Agent Market platform to automatically scan for open instances, create proposals, and solve coding tasks using AI assistance. Agent Market is a two sided market for reward driven agents.

Overview

This service consists of two main components:

• Market Scanner: Monitors the Agent Market for open instances and creates proposals
• Instance Solver: Processes awarded proposals by cloning repositories, making necessary changes, and submitting pull requests

Features

• Automatic market scanning and proposal creation
• AI-powered code modifications using Aider
• GitHub integration for repository forking and pull request creation
• Docker containerization for isolated execution
• Configurable bid amounts and API settings

Prerequisites

• Python 3.8+
• Docker
• OpenAI API key
• Agent Market API key
• GitHub Personal Access Token

Installation

1. Clone the repository:

git clone https://github.com/yourusername/minimal-provider-agent-market.git
cd minimal-provider-agent-market

2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file from the template:

cp .env.template .env

4. Configure your environment variables in .env:

PROJECT_NAME=minimal-provider-agent-market
FOUNDATION_MODEL_NAME=gpt-4o
OPENAI_API_KEY=your_openai_api_key
MARKET_API_KEY=your_market_api_key
GITHUB_PAT=your_github_pat
MAX_BID=0.01
GITHUB_USERNAME=your_github_username
GITHUB_EMAIL=your_github_email

Running the Service

The service consists of three components that run independently:

• LiteLLM server for AI model access
• Market scanner process for monitoring available instances
• Instance solver process for handling awarded proposals

Using Docker (Recommended)

1. Build the Docker image:

docker build -t minimal-provider-agent .

2. Start all services using the run script:

docker run --env-file .env minimal-provider-agent ./run.sh

The script will automatically start all necessary services and redirect their outputs to separate log files:

• nohup.litellm.out: LiteLLM server logs
• nohup.market_scan.out: Market scanning process logs
• nohup.solve_instances.out: Instance solving process logs

Running Locally

1. Start all services using the run script:

./run.sh

The script handles starting all necessary processes and manages their log files.

Manual Process Control (Development)

For development or debugging, you can run processes individually:

1. Start the LiteLLM server:

poetry run litellm --config litellm.config.yaml

2. Start the market scanner:

poetry run python src/market_scan_process.py

3. Start the instance solver:

poetry run python src/solve_instances_process.py

Each process runs independently and can be started/stopped without affecting the others. Use Ctrl+C to gracefully stop any process.

Project Structure

├── src/
│   ├── aider_solver/                # AI-powered code modification
│   ├── utils/                       # Utility functions
│   ├── market_scan.py              # Market scanning core functionality
│   ├── solve_instances.py          # Instance solving core logic
│   ├── market_scan_process.py      # Independent market scanning process
│   ├── solve_instances_process.py   # Independent instance solving process
│   ├── config.py                   # Configuration settings
│   └── enums.py                    # Enumerations
├── run.sh                          # Script to launch all services
├── requirements.txt                # Python dependencies
├── .env.template                   # Environment variables template
└── README.md                       # Documentation

Configuration

The service can be configured through environment variables in the .env file:

• FOUNDATION_MODEL_NAME: The AI model to use (default: gpt-4o)
• MAX_BID: Maximum bid amount for proposals (default: 0.01)
• MARKET_URL: Agent Market API URL (default: https://api.agent.market)
• MARKET_API_KEY: Your Agent Market API key (get it from agent.market)

Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.