RAG Pipeline — Chat with Your Notes

A Retrieval-Augmented Generation (RAG) pipeline that answers questions based on your PDF notes using ChromaDB and Groq LLM.

What it does

Upload your PDF notes, ask any question, and get accurate answers with citations showing exactly which page the answer came from.

Tech Stack

• Document Loading — PyMuPDF
• Text Splitting — LangChain RecursiveCharacterTextSplitter
• Embeddings — SentenceTransformers (all-MiniLM-L6-v2)
• Vector Store — ChromaDB
• LLM — Groq (LLaMA 3.3 70b)

Setup

1. Clone the repo

git clone https://github.com/AkshitaSharma211/RAG.git
cd RAG

2. Install dependencies

pip install -r requirements.txt

3. Add your Groq API key

echo "GROQ_API_KEY=your_key_here" > .env

4. Run the notebook Open pdf_loader.ipynb and run all cells top to bottom.

How it works

1. PDFs are loaded and split into chunks
2. Each chunk is converted to embeddings (384-dim vectors)
3. Embeddings stored in ChromaDB
4. User question is embedded and matched against stored chunks
5. Top matching chunks sent to Groq LLM for answer generation

Example

result = adv_rag.query("What is video compression?", top_k=3, min_score=0.1)
print(result['answer'])