🤖 Miniature Agent

An AI-powered coding assistant that runs locally and helps you write, edit, and run code through a simple chat interface.
Inspired by ghuntley’s coding agent workshop and enhanced with support for OpenAI APIs or free local Hugging Face models (e.g. Code Llama).

Built entirely in Python with a sleek Gradio web UI.

---

✨ Features

• 🧑‍💻 Code Generation & Explanation – ask the agent to write or explain code.
• 📂 File Tools – read, edit, and update local files.
• 🖥️ Command Execution – run shell commands (e.g. pytest).
• 🔄 Flexible Models
  • Use OpenAI GPT-3.5/4 if you have an API key.
  • Or run CodeLlama locally for free (no key required).
• 🌐 Web Interface – intuitive chat UI powered by Gradio.
• ⚡ One-Click Run – just python app.py and start coding with your AI buddy.

---

📦 Installation

1. Clone the repo

git clone https://github.com/aman-raza/Miniature-Agent-.git
cd Miniature-Agent-

2. Create a virtual environment

python3 -m venv venv
source venv/bin/activate   # (Windows: venv\Scripts\activate)

3. Install Dependencies

pip install -r requirements.txt

4. 🔑 Configuration

• Using OpenAI (recommended for accuracy):

  • Create a .env file in the project root:

    OPENAI_API_KEY=sk-xxxx

  • The agent will automatically use GPT-3.5/4.

• Using Hugging Face local model (free):

  • No .env required.

  • The app will download CodeLlama 7B

  • on first run (~13GB).

  • GPU strongly recommended.

5. 🚀 Usage

• Start the app:

  python my_agent.py

6. Gradio will launch at:

http://127.0.0.1:7860

Now you can chat with your coding agent in the browser

7. 🛠️ Example Interactions

• Writing Code

  User: Write a Python function to check if a number is prime.
  Agent: 
  def is_prime(n):
      if n < 2: return False
      for i in range(2, int(n**0.5)+1):
          if n % i == 0: return False
      return True

• Running Test

  User: Run pytest on the current project.
  Agent: {"tool": "run_cmd", "input": "pytest --maxfail=1"}
  Agent: (shows test results)

• Editing Files

  User: Add a function reverse_string(s) to utils.py
  Agent: ✅ File utils.py updated.

• Project Structure

  coding-agent/
  │── my_agent.py            # main script (run this)
  │── requirements.txt  # dependencies
  │── .env              # API keys (optional, for OpenAI)
  │── README.md         # this file

8. (Optional) Create a desktop .exe app

• Install pyinstaller and pywebview

  pip install pywebview
  pip install pyinstaller
  

• Build .exe

  pyinstaller --noconfirm --onefile --windowed main.py
  

• After building, you’ll find the .exe inside dist/main.exe. Double-click -> results in a true desktop app.

---

• ⚠️ Notes

  • Local models require significant RAM/GPU. If limited, prefer OpenAI API.
  • Always review code before executing – the agent can run shell commands.

• 📜 License

  • MIT License – feel free to fork, modify, and share.

• 🙌 Acknowledgements

  • Geoffrey Huntley – How to Build a Coding Agent
  • Hugging Face – Code Llama
  • Gradio