ContextIQ

ContextIQ is a professional "Chat with PDF" Retrieval-Augmented Generation (RAG) app. Upload a PDF, index it in Pinecone using Gemini embeddings, and ask questions with answers grounded in retrieved context.

Architecture Overview

1. Upload a PDF in the Streamlit UI
2. Extract text and split into chunks (RecursiveCharacterTextSplitter)
3. Generate embeddings with Gemini
4. Store vectors in Pinecone
5. Embed the query and retrieve relevant chunks
6. Generate an answer strictly from the retrieved context

Tech Stack

• Streamlit (UI)
• Google Gemini (LLM + embeddings)
• Pinecone (vector database)
• LangChain (chunking)
• PyPDF (PDF parsing)

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Create an environment file:

cp .env.example .env

3. Fill in the required values in .env:

• GOOGLE_API_KEY
• PINECONE_API_KEY
• PINECONE_INDEX_NAME

4. Run the app:

streamlit run app.py

Notes

• The app keeps only the latest uploaded PDF (vectors are stored in a single Pinecone namespace and overwritten on each upload).
• If no relevant context is found, the app will say it does not know.