🌟 AI Code Hacker 🚀

This repository showcases an implementation of an AI-powered code hacker 🤖 designed to solve problems from the Meta Hacker Cup 2024 🏆. With advanced Large Language Models (LLMs) like Google's Gemini or OpenAI's GPT-4o, this project highlights the cutting-edge capabilities of AI in competitive programming and algorithmic problem-solving. 💡💻

🔎 Overview

The following flowchart illustrates the problem-solving pipeline:

🚀 Getting Started

1. Install Dependencies

Ensure you have Python installed, and then run:

pip install -r requirements.txt  

2. Prepare the Data

Download and extract the problem dataset as follows:
1.Get the Meta Hacker Cup 2024 Round 2 Data.
2.Extract the data into your preferred directory:

mkdir path_to_save_data  
tar xvf contestData_2024_r2.tar -C path_to_save_data  

🔧 Running the Solution

Step 1: Set Up Your API Key

Set your API key for Gemini using the GOOGLE_API_KEY environment variable:

export GOOGLE_API_KEY=your_api_key  

Step 2: Choose a Mode of Operation

AutoGen Mode

Uses the AutoGen framework to generate a solution based on problem input.

python run_autogen.py \
    <problem_id> \
    <problem_name> \
    <data_path>  

LangChain Mode

Uses the LangChain framework to generate a solution to the problem input.

python run_langchain.py \
    <problem_id> \
    <problem_name> \
    <data_path>  

Argument Description

• problem_id: A unique identifier for the problem.
• problem_name: The name of the problem (typically the directory name).
• data_path: The directory where the problem dataset is stored.

Result

After execution, Python code that solves the given problem will be generated in the data_path/problem_name directory.

📊 Benchmarking

Performance evaluation and benchmarking tools are under development and will be released in future updates.