PageSage: PDF RAG Web Application

PageSage is a web application that uses Retrieval Augmented Generation (RAG) to allow users to communicate with a PDF document.
It works by vectorizing uploaded pdfs and then using a pre-trained model to generate text.

How It's Built

• The frontend is built with ReactJS and TypeScript.
• The frontend is built with FastAPI.
• Embeddings are generated using Langchain Sentence Transformers.
• The vector storage is using ChromaDB.
• The prompts are processed in context by Google Gemini API.

How It Works

• Upload your PDF file using drag and drop.
• The PDF is being uploaded to the backend, split into chunks, and stored in the vector database.
• The text area becomes enabled, and you can send queries to the backend which will be answered in the context of the uploaded document.

How To Run

• Clone the repo
• cd into the frontend directory, and run the command npm i.
• cd into the backend directory, and run pipenv install, pipenv shell and then fastapi dev.
• If you don't have pipenv installed, run pip install pipenv
• You should have a .env file, with the api key set as shown in the .env.example file.