LangChain Multi-Language Code Debugger (Powered by Gemini)

Project Overview

This is an advanced AI agent built on the LangChain framework and powered by the Gemini Large Language Model (LLM), designed for seamless integration into the developer's terminal. Its core function is to automate the process of finding and fixing errors across multiple programming languages, specifically Python, C++, and Java.

How It Works

The system operates on a closed-loop "Observe-Reason-Act" agent model:

1. Input & Detection: The user pastes buggy code. The agent first identifies the programming language (Python, C++, or Java).
2. Execution & Error Capture: It uses a set of specialized LangChain execution tools (e.g., execute_cpp_code) to run the code in an isolated environment. This action captures the precise stack trace or error output.
3. LLM Analysis: The captured error output and the original code are fed to the Gemini LLM (via the API key). Gemini analyzes the error and the code to diagnose the root cause.
4. Fix & Verification: Gemini generates the fully corrected code and an explanation. Crucially, the agent attempts to verify the fix by running the corrected code again to confirm successful execution before presenting the final answer to the user.

Motivation: Terminal-First Debugging

The core motivation is a terminal-integrated debugger which can provide personalized solutions and explanations to quickly identify code errors, acting as an on-demand tutor for absolute beginners (CS101). By automating the debugging loop with an LLM, it minimizes the tiring, time-consuming process for all developers. It can give personalized results depending on the prompt, which can significantly reduce time and effort.

Installation

1. Clone the repository:

git clone https://github.com/AbhishekU05/Hello_FOSS_25_code_debugger.git
cd Hello_FOSS_25_code_debugger

2. Install dependencies:

source foss_venv/bin/activate
pip3 install -r requirements.txt

3. Create a .env file for your environment variables

# Google Gemini API Configuration
GOOGLE_API_KEY=your-gemini-api-key-here

# Optional: Specify model (default is gemini-pro)
# GEMINI_MODEL=gemini-pro

# Optional: Set temperature (default is 0.0)
# TEMPERATURE=0.0

Getting a Gemini API Key

1. Go to Google AI Studio
2. Sign in with your Google account
3. Create a new API key
4. Copy the key to your .env file

Usage

Run the debugger:

python main.py

The debugger will:

1. Load a sample buggy code snippet
2. Identify the issues
3. Generate fixes
4. Test the corrected code
5. Display the results

Project Structure

code_debugger/
│
├── main.py              # Entry point
├── agent.py             # LangChain agent configuration with Gemini
├── tools/
│   ├── __init__.py
│   └── executor.py      # Safe Python code executor
├── examples/
│   ├── __init__.py
│   └── buggy_code.py    # Sample buggy code for testing
├── requirements.txt     # Project dependencies (updated for Gemini)
├── .env         # Environment variables template
└── README.md           # This file

Customization

Using Different Gemini Models

Edit the create_agent() call in main.py:

agent_executor = create_agent(
    model_name="gemini-pro",         # Available models: gemini-pro, gemini-pro-vision
    temperature=0.2,                 # Higher for more creative solutions
    verbose=True                     # Set to False for less output
)

Adding New Buggy Code Samples

Add new examples to examples/buggy_code.py:

BUGGY_CODE_SAMPLES["my_bug"] = '''
def my_function():
    # Your buggy code here
    pass
'''

Security

• Code execution happens in isolated subprocesses
• Dangerous operations are blocked (eval, exec, file operations)
• Execution timeout prevents infinite loops
• Output length is limited to prevent memory issues

Requirements

• Python 3.8+
• Google Gemini API key
• Dependencies listed in requirements.txt

License

MIT