🐛 CrewAI Code Debugger

A multi-agent system built with CrewAI that automatically detects and fixes errors in Python code. Two specialized agents — a Code Analyzer and a Code Corrector — work sequentially under a Manager agent to deliver clean, corrected code.

---

📚 Table of Contents

• Overview
• Workflow Diagram
• Project Structure
• Agents
• Tasks
• Example
• Setup & Installation
• Usage
• Configuration
• License

---

🧠 Overview

This project demonstrates a sequential CrewAI pipeline with planning enabled. Given a snippet of buggy Python code, the system:

1. Runs the code through a Code Analyzer agent (powered by CodeInterpreterTool) to detect all syntax and logical errors.
2. Passes the error report to a Code Corrector agent that returns a fully fixed, PEP-8 compliant version.
3. A Manager agent oversees the entire workflow, coordinating delegation and ensuring quality output.

---

🔀 Workflow Diagram

flowchart TD
    A([🚀 Start]) --> B[📥 Input: Buggy Python Code]
    B --> C{🧠 Manager Agent\nOversees Workflow}

    C --> D[🔍 Code Analyzer Agent\nCodeInterpreterTool]
    D --> E[📋 Analysis Task\nIdentify Syntax & Logic Errors]
    E --> F[📄 Error Report\nList of Issues Found]

    F --> G[🔧 Code Corrector Agent]
    G --> H[✏️ Correction Task\nFix Identified Errors]
    H --> I[✅ Corrected Python Code]

    I --> J([🏁 End: Output Delivered])

    style A fill:#4CAF50,color:#fff
    style J fill:#4CAF50,color:#fff
    style C fill:#FF9800,color:#fff
    style D fill:#2196F3,color:#fff
    style G fill:#9C27B0,color:#fff
    style F fill:#f5f5f5,stroke:#999
    style I fill:#E8F5E9,stroke:#4CAF50

Loading

The .mmd source file is located at Flow/workflow.mmd.

---

📁 Project Structure

crewai-code-debugger/
├── Flow/
│   └── workflow.mmd            # Mermaid workflow diagram source
├── crewai_debugger.ipynb       # Main notebook — agents, tasks, and crew
├── requirements.txt            # Python dependencies
├── .env.example                # Environment variable template
├── .gitignore
├── LICENSE
└── README.md

---

🤖 Agents

Agent	Role	Tool	Delegation
Code Analyzer	Identifies syntax & logical errors	CodeInterpreterTool	❌
Code Corrector	Fixes all identified errors	—	❌
Manager	Oversees and coordinates the pipeline	—	✅

---

📋 Tasks

1. Analysis Task

• Agent: Code Analyzer
• Tool: CodeInterpreterTool — executes the code inside a Docker container to surface runtime and syntax errors
• Output: A numbered list of all errors with type, location, and description

2. Correction Task

• Agent: Code Corrector
• Context: Receives the error list from the Analysis Task via context=[analysis_task]
• Output: Complete, corrected Python code ready to run

---

💡 Example

Input (Buggy Code)

def fibonacci_iterative(n):
if n < 0:
return []
elif n == 1:
return [0]
elif n == 2:
return [0, 1]
fib_sequence = [0, 1]
for i in range(2, n):
next_fib = fib_sequence[-1] + fib_sequence[-2]
fib_sequence.append(next_fib)
return fib_sequence

Expected Output (Corrected Code)

def fibonacci_iterative(n):
    if n < 0:
        return []
    elif n == 1:
        return [0]
    elif n == 2:
        return [0, 1]
    fib_sequence = [0, 1]
    for i in range(2, n):
        next_fib = fib_sequence[-1] + fib_sequence[-2]
        fib_sequence.append(next_fib)
    return fib_sequence

---

⚙️ Setup & Installation

Prerequisites

• Python 3.10+
• Docker Desktop (required by CodeInterpreterTool to execute code safely)
• An OpenAI API key with an active billing plan

Steps

# 1. Clone the repository
git clone https://github.com/SANJAI-s0/crewai-code-debugger.git
cd crewai-code-debugger

# 2. Create and activate a virtual environment
python -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. Configure environment variables
cp .env.example .env
# Edit .env and add your OPENAI_API_KEY

---

🚀 Usage

Open and run the notebook:

jupyter notebook crewai_debugger.ipynb

Run all cells top to bottom. The crew will execute sequentially:

1. Code Analyzer evaluates the buggy code and lists all errors.
2. Code Corrector applies fixes based on the error report.
3. The final corrected code is displayed in the Run the Pipeline cell.

To debug different code, update the BUGGY_CODE variable in Cell 3.

---

🔧 Configuration

Variable	Description	Required
OPENAI_API_KEY	Your OpenAI API key (paid plan required)	✅
OPENAI_MODEL_NAME	Model to use (default: gpt-4o)	❌

Set these in your .env file (use .env.example as a template).

CodeInterpreterTool requires Docker Desktop to be running. Start it before executing the pipeline cell.

---

📄 License

This project is licensed under the MIT License.