🐞 BugMaster: AI Code Debugging Game

A Streamlit-based game to help developers sharpen their debugging skills on AI-generated buggy Python code — powered by LLaMA3 (via Groq API).

---

🧠 Overview

BugMaster is a game where developers debug broken Python code generated by a large language model (LLM). Each round presents you with AI-generated code that almost works — your job is to fix it until it passes all hidden test cases.

---

💡 Motivation

Modern developers increasingly rely on AI to write code — but AI isn't always right. In fact, AI can generate convincing but subtly broken logic. This game helps developers build the essential skill of debugging AI-generated code, making them faster, smarter, and more resilient coders.

⚠️ Vibe coders must learn how to debug AI code — it’s a critical 2020s skill.

---

🛠 Tech Stack

• Frontend: Streamlit
• Code Editor: st.text_area (previously streamlit-code-editor)
• AI Backend: Groq API using LLaMA3
• Language: Python 3.9+
• Prompting Engine: Structured prompting to generate original, buggy, and testable code

---

🚀 How to Run the App

✅ Installation

1. Clone the repo:

   git clone https://github.com/your-username/bugmaster.git
   cd bugmaster

2. Create virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # on Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set your Groq API key:

   • Option 1: Set it in your environment

     export GROQ_API_KEY=your_api_key_here

   • Option 2: Hardcode it directly inside the app (not recommended for production)

▶️ Run the app

streamlit run app.py

---

🎮 Features

• 🧩 AI-Generated Buggy Code: Based on real-world problems and LLM-crafted challenges
• 🧠 Three Difficulty Levels: Easy, Medium, and Hard — each increasing in code complexity and logic depth
• ⌨️ Editable Code Editor: Edit and fix the broken code directly in the app
• 🧪 Hidden Test Cases: Executed silently to verify your fix — you don’t get to see them
• ⏱️ Time-Based Performance: Solve faster and celebrate with confetti
• ✅ Instant Feedback: Know immediately whether your fix passed or failed

---

📚 Difficulty Topics

Easy

• CSV cleaning
• Basic aggregations
• Datetime filtering
• Sorting and ranking
• Top-N items per group

Medium

• ETL pipelines
• Data joins
• Window functions
• Data pivoting/unpivoting
• Time-series anomaly detection

Hard

• Sessionizing logs
• Star schema generation
• Chunk-based large file processing
• Data audit reports

---

🤖 Behind the Scenes

• Prompts are sent in two stages:
  1. Generate clean & tested code with embedded tests
  2. Introduce subtle bugs (syntax or logic) into the clean code
• Only the buggy portion is shown to the user; hidden tests are stored and evaluated behind the scenes.

---

🧪 Example Flow

1. User selects "Medium"
2. AI generates a clean, tested function (e.g., reverse_string)
3. AI then introduces a bug (e.g., off-by-one error or logic flaw)
4. User sees only the buggy version and edits it
5. Upon submit, the backend runs hidden tests to validate the fix

---

📎 Requirements

• Python 3.9+
• Groq API key
• Internet access (to call LLM)

---

🤝 Contributing

Pull requests and issues are welcome! Feel free to suggest new features, difficulty levels, or bug templates.

---

🧠 Future Ideas

• Leaderboards and timed rounds
• More languages (JavaScript, Java, etc.)
• Custom test case input
• Multiplayer mode (head-to-head debugging)

---

📜 License

MIT License — feel free to use, fork, and share.

---

Happy debugging! 🐛🔧