Competitive Programming Code Generation 🚀

Overview

This project introduces a robust, multi-agent system with Hybrid RAG approach designed to analyze and solve competitive programming (CP) problems while generating comprehensive and edge-case test cases. With a curated database of over 300+ competitive programming problems, the framework captures detailed problem descriptions, constraints, examples, and solution strategies across various complexity levels, including brute force, intermediate, and optimal approaches.

By constructing a dynamic knowledge graph with FAISS based RAG (Retrieval-Augmented Generation), the system retrieves and synthesizes problem insights to develop efficient solution pipelines. Through agent collaboration, it automates the problem-solving process from analysis to code generation.

System Architecture 🧩

Problem Database

• Extensive Dataset: Over 300 CP problems with well-structured information, including solution approaches and executable code.
• Detailed Insights: Comprehensive problem-solving steps enhance solution retrieval and validation.

Hybrid RAG : Knowledge Graph & FAISS 🌐

• Dynamic Knowledge Graph: Captures relationships between problem attributes and solution patterns for intelligent retrieval.
• FAISS-Based Retrieval: Enables fast and scalable semantic search for relevant problems and strategies.
• Structured and Faster Reasoning: Seamlessly integrates structured knowledge from the graph with high-speed semantic retrieval from FAISS, enabling deep reasoning, multi-hop problem-solving, and context-aware solution generation.

Agent Framework 🤖

• Analysis Agent: Breaks down user-provided CP problems, identifying key constraints and objectives.
• Planning Agent: Develops structured solution strategies, algorithms, and pseudocode.
• Coding Agent: Translates pseudocode into fully executable and optimized code solutions.
• Test Case Agent: Automatically generates diverse test cases, including edge cases, for solution validation.
• Optimizing Agent: Evaluates and optimizes time and space complexity to produce the most efficient solution.

Technical Workflow ⚙️

1. Problem Analysis: The Analysis Agent dissects the problem statement to extract critical constraints.
2. Knowledge Retrieval: Graph-RAG is leveraged to retrieve analogous problem insights.
3. Solution Planning: The Planning Agent formulates algorithmic strategies and detailed pseudocode.
4. Code Generation: The Coding Agent converts pseudocode into an optimized, executable solution.
5. Test Case Creation: The Test Case Agent produces diverse and comprehensive test scenarios.
6. Complexity Optimization: The Optimizing Agent evaluates and fine-tunes the code for optimal performance.
7. Solution Deployment: Outputs a validated and efficient solution ready for testing.

Key Features 🔑

• Scalable Knowledge Integration: Continuously learns and adapts as new problems are added.
• Agent Collaboration: Distributed architecture ensures task specialization and parallel processing.
• Comprehensive Test Generation: Robust testing guarantees solution reliability.
• Knowledge-Driven Problem Solving: Intelligent retrieval of patterns for algorithm development.
• Complexity Optimization: Ensures the most efficient code in terms of time and space.
• Modular and Extensible: Supports future integration with other platforms.

Tech Stack 🛠️

• Agent Framework: CrewAI for agent-based architecture
• Database and Knowledge Graph: Neo4j
• Retrieval and Processing: LangChain, FAISS
• Frontend and Backend: FastAPI, HTML, CSS

Repo Structure

• backend/

  • agents.py - Implements the Knowlege Graph and FAISS based RAG approach with agent definitions.
  • main.py - Main backend script handling API requests and execution.

• frontend/

  • index.html - Main HTML structure for the web interface.
  • script.js - JavaScript logic for interactivity.
  • styles.css - Styling for the frontend.

• requirements.txt - Lists dependencies required for the backend.

Future Scope 🌟

• Integrated Compiler: Build a compiler linked with a debugger cycle to generate the perfect code.
• Dataset Expansion: Ongoing enrichment of problem and solution datasets.
• Enhanced Retrieval Algorithms: Optimized similarity detection for improved knowledge graph queries.
• Platform Integration: Compatibility with external CP environments.
• Advanced Debugging Features: Agents for real-time code analysis and performance optimization.

Mentors & Contributors 👥

• Mentors: Param Thakkar, Pranav Janjani, Anish Singh
• Contributors: Ninad Shegokar, Khush Agrawal, Sujal Sakpal and Rohit Bhargav

Acknowledgements ❤️

• Team Works-On-My-Machine
• Community of Coders (COC) at VJTI, Mumbai.