PDF Question Answering Assistant

An end-to-end Retrieval-Augmented Generation (RAG) application for querying PDF documents using semantic search and local LLM inference.

---

Features

• Upload and process PDF documents
• Extract and chunk document text
• Generate vector embeddings using Sentence Transformers
• Store embeddings using FAISS vector database
• Perform semantic similarity retrieval
• Generate context-aware answers using Llama 3 via Ollama
• Display source page references

---

Tech Stack

• Python
• Streamlit
• FAISS
• Sentence Transformers
• Ollama
• Llama 3
• LangChain Text Splitters

---

Pipeline

PDF Upload → Text Extraction → Chunking → Embedding Generation → Vector Search → Context Retrieval → LLM Response Generation

---

Demo

Application UI

Question Answering Example

---

Run Locally

git clone https://github.com/asthasinghcs/pdf-rag-assistant.git

cd pdf-rag-assistant

python -m venv venv

venv\Scripts\activate

pip install -r requirements.txt

streamlit run app/main.py

---

Local LLM Setup

Install Ollama:

https://ollama.com/download

Then run:

ollama run llama3

---

Project Structure

app/
├── main.py
├── pdf_processor.py
├── chunking.py
├── embeddings.py
├── vector_store.py
├── retriever.py
├── prompts.py
└── llm.py

---

Architecture

PDF Upload
↓
Text Extraction
↓
Chunking
↓
Embedding Generation
↓
FAISS Vector Storage
↓
Semantic Retrieval
↓
LLM Response Generation

---

Future Improvements

• Hybrid search (BM25 + vector retrieval)
• Chat memory
• Multi-PDF support
• Metadata filtering
• Better reranking pipeline
• Deployment support