DSA Problem Solver - a system that outperforms ChatGPT in solving unseen leetcode medium and hard problems

Overview

The DSA Problem Solver project aims to simplify the process of solving Data Structures and Algorithms (DSA) problems by leveraging advanced language models and retrieval-augmented generation (RAG) techniques. This tool is designed to assist users in tackling challenging DSA problems, making the process more intuitive and less daunting.

Current Approach:

Open to suggestions:

flowchart TD
    A[Start] --> B[Generate problem approach]
    B --> C[Generate pseudo code]
    C --> D[Generate Python code with test cases]
    D --> E[Initialize variables]
    E --> F{Tries <= 5 and\nnot solved?}
    F -->|Yes| G[Attempt to execute code]
    G --> H{Execution successful\nand all tests passed?}
    H -->|No| I[Handle Failure]
    I --> J[Update Python code]
    J --> F
    H -->|Yes| K[Mark as solved]
    K --> F
    F -->|No| L[Return results]
    L --> M[End]

    style F decision
    style H decision

Loading

flowchart TD
    A[Start handleFailure] --> C[Trying to correct the code]
    C --> D[Get code corrector prompts with test case failure and error messages]
    D --> E[Initialize LLM with OpenAI GPT-4]
    E --> F[Generate corrected code with test cases using LLM]
    F --> G[Extract and process corrected code with tests]
    G --> I[Return corrected code]
    I --> J[Return corrected code]
    J --> K[End handleFailure]

Loading

Features

• Agentic LLM Integration: Utilizes language models to understand and generate solutions for various DSA problems by breaking it down into simpler steps
• Retrieval-Augmented Generation (RAG): Enhances problem-solving capabilities by combining retrieval of latest information on the internet to support generation of tailored solutions.
• Code Execution: Uses native python engine to eecute code and test cases to enable iterative improvements based on error messages / test case failure

Challenges

• Getting an LLM to understand logic
• To break the barrier - the system is as good as the LLM used

Usage

1. Input Problem Statement: Provide the DSA problem statement either via the command line interface or through the web interface.

2. Receive Solution: The tool will analyze the problem using the Agentic LLM and RAG techniques, then provide a detailed solution with explanations.

3. Explore Solutions: Users can review the solutions and explanations, use the playground to test the solution and interact with the tool for further clarification or alternative approaches.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For any questions or support, feel free to reach out.