DeepReader

A semantic document Q&A API - upload a PDF and ask questions about it. The API retrieves relevant context from your document and generates grounded answers using an LLM.

What it does

• Upload any PDF document via /upload
• Ask questions about the document via /ask
• Answers are generated using retrieved context, not hallucination

Stack

• Python, FastAPI
• LangChain, ChromaDB
• FastEmbed Embeddings (BAAI/bge-small-en-v1.5)
• OpenRouter API (LLM)
• Docker, Hugging Face Spaces

Endpoints

• POST /upload - accepts a PDF, chunks and embeds it into ChromaDB
• POST /ask - accepts a question, retrieves relevant chunks, returns LLM answer

Live

DeepReader on Hugging Face Spaces

No frontend - interact with the API directly via Swagger UI at: https://abubakker66-deepreader.hf.space/docs

Run locally

git clone https://github.com/abubakkersiddiqq/deep-reader
cd deep-reader
cp .env.example .env  # add your OpenRouter API key
docker build -t deep-reader .
docker run -p 8000:8000 deep-reader

Local Swagger UI available at: http://localhost:8000/docs