PDF Conversation

Ask questions about a PDF and get answers grounded in its content.

A Streamlit app that lets you upload one or more PDFs and query them in natural language. Text is extracted, chunked, embedded, and indexed with FAISS; a retrieval-augmented chain answers questions using only the retrieved passages, with conversation memory across turns.

How it works

1. Extract — text is pulled from every page of the uploaded PDFs (PyPDF2).
2. Chunk — text is split into overlapping chunks for retrieval (RecursiveCharacterTextSplitter).
3. Embed & index — chunks are embedded (HuggingFaceInstructEmbeddings) and stored in a FAISS vector store.
4. Answer — a conversational retrieval chain (LangChain + Google PaLM) answers questions using the most relevant chunks, remembering prior turns in the session.

Setup

pip install -r requirements.txt

Create a .env file with your Google API key:

GOOGLE_API_KEY=your-api-key-here

Run

streamlit run multipdf.py

Stack

• Streamlit
• LangChain + Google PaLM (langchain_community.llms.GooglePalm)
• FAISS (faiss-cpu)
• PyPDF2
• HuggingFace Instructor embeddings

License

MIT