LeetCode-RAG-Tutor 🤖

Built a small local RAG engine over a weekend to make LeetCode revision less painful. Instead of scrolling through long notes and old solutions, I wanted a way to search my personal DSA knowledge base and quickly retrieve patterns, dry-runs, code snippets, and mnemonics from my markdown notes.

What it does

• Indexes structured Markdown/Notion exports
• Retrieves relevant problem-solving patterns using semantic search
• Preserves code blocks and note structure for better context
• Answers queries using an LLM over the retrieved notes

Tech Stack

• Chunking: MarkdownHeaderTextSplitter
• Embeddings: all-MiniLM-L6-v2
• Vector Store: ChromaDB
• LLM: Llama 3.3 70B via Groq
• Framework: LangChain

Some things I added

• MD5-based incremental sync so only new/updated notes get re-indexed
• Header-aware chunking to avoid breaking code blocks
• Persistent local vector database
• Fast inference through Groq (usually around ~1 second response time)

Quick Start

1. Install dependencies:

pip install langchain langchain-chroma langchain-huggingface langchain-groq sentence-transformers python-dotenv

2. Create a /notes/ folder and drop your markdown files inside.
3. Add your API key in a local .env file:

GROQ_API_KEY=your_key_here

4. Run the interactive tool:

python app.py

Still a small project, but it's already become my go-to tool for interview prep and revising old problem-solving patterns.