📄 PDF Retrieval-Augmented Generation (RAG) System

This project is a PDF-based RAG pipeline built with LangChain, Ollama, ChromaDB, and ElevenLabs. It lets you load PDFs, split them into chunks, attach metadata, store them in a vector database, and query them using a multi-query retrieval strategy — with optional text-to-speech output.

---

🚀 Features

• PDF ingestion via PDFPlumberLoader
• Text chunking with overlap using RecursiveCharacterTextSplitter
• Metadata tagging (title, author, date) for filtering and better retrieval
• Vector storage in ChromaDB
• Multi-query retriever to improve recall
• LLM querying using Ollama (default: llama3.2)
• Text-to-speech output via ElevenLabs

---

📦 Installation

# Clone the repository
git clone https://github.com/Sourav01112/voice-rag-pdf-assistant.git
cd voice-rag-pdf-assistant

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create .env file for ElevenLabs API
echo "ELEVENLABS_API_KEY=your_api_key_here" > .env

---

📂 Project Structure

voice-rag-pdf-assistant/
│── data/                  # PDF files go here
│── db/vector_db/          # ChromaDB persistence
│── rag_system.py          # Main RAG pipeline
│── requirements.txt
│── README.md

---

🛠 Usage

python rag_system.py

• Place your .pdf files inside the data/ folder.

• The system will:

  1. Load PDFs
  2. Split into text chunks
  3. Add metadata
  4. Store embeddings in ChromaDB
  5. Run a multi-query retrieval chain
  6. Answer your question
  7. Optionally convert the answer to speech

---

📌 Example Output

Loading PDF files...
Processing PDF file: report.pdf
Pages loaded: 12
Splitting text into chunks...
Created 45 text chunks
Adding metadata...
Setting up vector database...
Vector database created and populated
Setting up retrieval chain...
Retrieval chain setup complete

Query: Does the document mention any specific technologies?
Response: Yes, it discusses blockchain and AI applications in the financial sector.

Converting to speech...

---

🔗 Requirements

• Python 3.9+
• Ollama installed locally
• ElevenLabs API key (optional for TTS)

---