An AI system using Retrieval-Augmented Generation (RAG) to answer questions from your PDFs. It retrieves relevant document chunks using embeddings and cites sources, enabling factual and explainable responses.
Highlights
- Upload PDFs and chat with their content.
- Local embeddings via
sentence-transformers(all-MiniLM-L6-v2). - Fast vector search using FAISS.
- LLM: LLaMA 3 (
meta-llama/Meta-Llama-3-70B-Instruct) served via Hugging Face Inference API. - Built with Streamlit + LangChain
ConversationalRetrievalChain.
Tech Stack
- UI: Streamlit
- Orchestration: LangChain (
ConversationalRetrievalChain, memory viaConversationBufferMemory) - Embeddings:
sentence-transformers(all-MiniLM-L6-v2) - Vector Store: FAISS (CPU)
- LLM: Hugging Face Endpoint for
Meta-Llama-3-70B-Instruct - PDF Parsing: PyPDF2
- Config:
python-dotenvwith.env
I built an AI system that uses Retrieval-Augmented Generation to answer questions from PDFs. Instead of relying purely on the LLM, it retrieves the most relevant document chunks using vector embeddings, ensuring factual and explainable responses.
RAG-based AI PDF Question Answering system using LangChain, FAISS, and LLMs to enable accurate, source-grounded question answering over unstructured documents.
A Retrieval-Augmented Generation (RAG) system that allows users to:
- Upload PDF documents
- Ask natural language questions
- Receive accurate, source-grounded answers extracted from the PDFs
