🧠 Clinical PDF QA + Summarizer — Fullstack RAG App

An end-to-end PDF understanding system built with:

• 🧠 Backend: RAG (Retrieval-Augmented Generation) w/ LangChain, Transformers, FAISS
• ⚛️ Frontend: Modern React (TypeScript) + Tailwind + Vite/Next.js (client-side)
• ⚙️ Infrastructure: Celery for async ingestion, Redis for queueing, Docker-ready

---

📦 Features

✍️ Upload + Process PDFs

• Upload .pdf via drag/drop or button
• Validate + monitor upload progress
• Summary auto-generated after ingestion

💬 Chat with your PDF

• Ask questions like: "What’s the treatment plan?"
• Backend uses retrieval from FAISS vectorstore
• Model: deepset/roberta-base-squad2 via HuggingFace pipeline

🧾 Summarization

• Extracts & chunks text from PDF
• Uses Falconsai/text_summarization for chunk-wise summaries

---

🧱 Project Structure

🖥 Backend (Python)

.
├── query.py               # Answer questions from vectorstore
├── ingest.py              # Chunk text + embed into FAISS
├── summarizer.py          # Summarize entire PDF
├── pdf_reader.py          # PDF → text extraction (PyMuPDF)
├── background_tasks.py    # Celery task for ingestion
├── Dockerfile             # Container setup

💻 Frontend (React/TS)

/components
├── UploadArea.tsx         # Drag-and-drop upload UI
├── SummaryBox.tsx         # Beautiful summary display + copy
├── ChatBox.tsx            # Q&A chat interface
├── ChatInput.tsx          # Send questions
├── PDFPreview.tsx         # Show uploaded PDF metadata
├── ProgressBar.tsx        # Upload progress visual
├── StatusIndicator.tsx    # Status badge (Idle, Uploading, etc.)
├── LoadingSkeleton.tsx    # Skeleton UIs
├── TypingIndicator.tsx    # Bot is typing...

---

🚀 Quickstart

Backend

# 1. Install dependencies
pip install -r requirements.txt

# 2. Start Redis
redis-server

# 3. Start Celery
celery -A background_tasks worker --loglevel=info

# 4. Start FastAPI or Flask server (not included here)
uvicorn app:app --reload

Frontend

npm install
npm run dev

---

🌐 API Endpoints

Endpoint	Method	Description
/Uploads	POST	Upload a PDF file
/summarize	GET	Get summary of uploaded file
/ask (optional)	POST	Ask a question about uploaded PDF

---

⚙️ Technologies

• 🧠 Transformers: deepset/roberta-base-squad2, Falconsai/text_summarization
• 🗃️ FAISS vectorstore + LangChain retrievers
• 🐇 Celery + Redis for background ingestion
• 🧩 Tailwind CSS + shadcn/ui for frontend polish

---

📌 Notes

• All summaries and vectorstores are cached locally (uploads/ + vectorstores/)
• PDFs are chunked via RecursiveCharacterTextSplitter
• Chat and summarization both rely on existing .txt if available for speed

---

🙌 Credits

Built with ❤️ by Ritankar.