Python Code Interpreter & CSV Agent Project

This project is a multi-agent system built with LangChain, OpenAI GPT-4 Turbo, and experimental utilities.
It includes two powerful agents: 1. Python Code Agent -- Executes Python code dynamically using a REPL. 2. CSV Agent -- Analyzes data from a CSV file using Pandas with natural language queries.

Both agents are combined into a Grand Agent, which decides which tool to use based on the input prompt.

---

Features

• Python Agent
  • Converts natural language into Python code.
  • Executes code in a REPL environment.
  • Handles errors by retrying execution.
  • Example: Generate QR Codes, perform calculations, automate Python scripts.
• CSV Agent
  • Reads and queries CSV files.
  • Uses Pandas for data analysis.
  • Example: Answer questions like "How many guests appeared how many times across all the seasons?" from the friends_complete_episodes_database.csv dataset.
• Grand Agent
  • Orchestrates between Python Agent and CSV Agent.
  • Routes user prompts to the right agent depending on the task.

---

Installation

1. Clone the repository:

   git clone https://github.com/your-username/python-code-interpreter.git
   cd python-code-interpreter

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate   # Linux / macOS
   venv\Scripts\activate    # Windows

3. Install dependencies:

   pip install -r requirements.txt

4. Add your OpenAI API key in a .env file:

   OPENAI_API_KEY=your_api_key_here

---

Usage

Run the project using:

python main.py

Example Queries

• CSV Agent Example

  How many guests have made how many times appearances across all the seasons?
  

• Python Agent Example

  Generate 15 QR Codes that point to my LinkedIn profile https://linkedin.com/in/mouyse/ and save them in the 'LinkedIn_QRCodes' directory.
  

• Grand Agent Example The grand agent can decide whether to use the Python Agent or CSV Agent automatically:

  "How many guests have made how many times appearances across all the seasons?"
  "Generate 15 QR Codes that point to my LinkedIn profile."
  

---

Project Structure

.
├── main.py                        # Main script containing agents
├── friends_complete_episodes_database.csv  # Dataset used by the CSV Agent
├── .env                           # API key storage
├── requirements.txt               
└── README.md

---

Next Steps

• Add support for multiple CSV files with dynamic selection.
• Enhance the Python agent with plotting/visualization support.
• Build a web-based UI to interact with both agents.

---

Tech Stack

• Python 3.9+
• LangChain
• OpenAI GPT-4 Turbo
• Pandas
• dotenv

---

License

This project is licensed under the MIT License.