PR-Agent

An autonomous repo-maintainer that listens for GitHub issues and automatically generates code fixes, commits them, and opens pull requests — powered by a local LLM via Ollama.

---

How It Works

GitHub Issue Opened
        |
        v
  GitHub Webhook (POST /webhook)
        |
        v
  listener.py (FastAPI server)
        |
        v
  agent.py
    ├── scanner.py  →  Ollama identifies the relevant file
    ├── Ollama generates the code fix
    ├── git commit & push to new branch
    └── gh pr create  →  Pull Request opened

1. A new issue is opened on your GitHub repository.
2. GitHub sends a webhook to the listener.
3. The listener triggers the agent with the issue title and body.
4. The scanner uses Ollama (Llama 3.1) to identify which file the issue relates to.
5. The agent reads the file, asks Ollama to generate a fix, and writes it back.
6. The agent commits the change, pushes a new branch, and opens a PR via the GitHub CLI.

---

Prerequisites

Dependency	Purpose
Python 3.10+	Runtime
Ollama	Local LLM inference
GitHub CLI (gh)	PR creation & auth
ngrok	Expose local server to GitHub

---

Quick Start

1. Clone and install

git clone https://github.com/yashasviudayan-py/PR-Agent.git
cd PR-Agent
pip install -r requirements.txt

2. Pull the model and start Ollama

ollama pull llama3.1:8b-instruct-q8_0
ollama serve

3. Start the listener

python listener.py

You should see: INFO: Uvicorn running on http://0.0.0.0:8000

4. Expose with ngrok

In a separate terminal:

ngrok http 8000

Copy the Forwarding URL (e.g., https://abc123.ngrok-free.app).

5. Configure the GitHub webhook

1. Go to your repository on GitHub.
2. Navigate to Settings > Webhooks > Add webhook.
3. Payload URL: <your-ngrok-url>/webhook
4. Content type: application/json
5. Events: Select "Let me select individual events" and check Issues.
6. Click Add webhook.

---

Docker

Pull the pre-built image from Docker Hub:

docker pull yashasviudayan/pr-agent:latest

Or run it directly:

docker run -p 8000:8000 yashasviudayan/pr-agent:latest

Note: The container includes Python, git, and GitHub CLI. You still need Ollama running on the host (accessible at host.docker.internal:11434).

---

Usage

Open an issue on the repository. The agent will automatically:

1. Identify the relevant file.
2. Generate a code fix using Ollama.
3. Open a pull request with the changes.

---

Project Structure

.
├── .github/
│   └── workflows/
│       └── docker-publish.yml   # CI/CD: auto-push to Docker Hub
├── listener.py                  # FastAPI webhook server
├── agent.py                     # Orchestrator: scan → fix → commit → PR
├── scanner.py                   # LLM-powered file identification
├── Dockerfile                   # Container image definition
├── requirements.txt             # Python dependencies
├── LICENSE                      # MIT License
├── .env                         # Environment variables (not committed)
├── .dockerignore
└── .gitignore

---

License

This project is licensed under the MIT License.