Github Project AI Agent

An AI-powered Python application that turns a natural-language request into a GitHub project. The agent generates a structured project specification, creates a GitHub Project v2, creates the related issues in a repository, and links those issues to the project.

Features

• Generate a project specification from a user prompt.
• Create a GitHub Project v2 using the GitHub GraphQL API.
• Automatically create issues in a target GitHub repository.
• Link each created issue to the project.
• Choose between OpenAI, Anthropic Claude, or a local Ollama model.

Prerequisites

• Python installed on your machine.
• A GitHub account with a valid personal access token.
• A target GitHub repository accessible with that token.
• API keys for OpenAI or Anthropic if you plan to use those providers.

Your GitHub token must have permission to create issues and projects (Scopes : repo, project, read:org, write:discussion)

Installation

1. Clone the repository.
2. Go to the project root.
3. Run the create_env.sh script to generate the .env file:

bash create_env.sh

4. Open the generated .env file at the project root and replace the placeholder values with your real credentials.
5. Install the Python dependencies:

pip install -r requirements.txt

Configuration

The create_env.sh script creates a .env file with the following variables:

• GITHUB_TOKEN: your GitHub personal access token.
• OWNER: the GitHub username or organization that owns the target repository.
• GRAPHQL_ENDPOINT: GitHub GraphQL endpoint.
• OPENAI_API_KEY: OpenAI API key, if you use OpenAI.
• ANTHROPIC_API_KEY: Anthropic API key, if you use Claude.

The application also reads a system prompt from src/prompts/system_prompt.json.

Usage

Run the application from the src directory:

cd src
python main.py

When prompted, provide:

• the repository name, either as owner/repo or just repo if OWNER is set in .env;
• a description of the project you want to create.

LLM provider selection

By default, the local model is used. You can explicitly choose a provider with one of these options:

python main.py --local
python main.py --openai
python main.py --claude

For the local provider, make sure you have Ollama installed and you have downloaded the required model (qwen2.5-coder:14b).

You can install Ollama by running:

curl -fsSL https://ollama.com/install.sh | sh

and then download the model with:

ollama pull qwen2.5-coder:14b

Workflow

The main flow is:

1. Read the user request.
2. Generate a structured project specification with the LLM.
3. Create the GitHub Project.
4. Create the issues described in the specification.
5. Link each issue to the project.

Example

Example prompt:

Create an e-commerce project with product management, a shopping cart, online payment, and a customer account area.

The agent will generate the corresponding GitHub project and repository issues.

Troubleshooting

• Make sure the .env file exists at the project root.
• Check that GITHUB_TOKEN and OWNER are set correctly.
• Make sure the target repository exists and is accessible with your token.
• If you use OpenAI or Claude, verify the corresponding API key is present.

License

See the LICENSE file.