📄 PaperRAG

PaperRAG is an AI-powered document ingestion and question-answering system.
It allows users to upload PDFs (research papers, resumes, reports, etc.), processes them into chunks, summarizes their content (text, tables, images), and stores embeddings in a vector database. Users can then chat with their documents through a simple web interface.

---

🚀 Live Demo

• Frontend (Next.js + tRPC) → https://paperrag.onrender.com/
• Backend (FastAPI) → https://paperrag-backend-296287667274.us-central1.run.app/

⚠️ Note: Backend is currently deployed on Google Cloud Run (ephemeral storage). Uploads may fail or reset. For persistence, see Future Scope.

---

✨ Features

• 📂 Upload PDFs directly in the browser.
• 🧩 Extracts text, tables, and images from PDFs.
• 📝 Summarizes extracted content with LLMs.
• 🔍 Embeds & stores summaries into Chroma DB.
• 💬 Chat interface to query documents using semantic search + LLM reasoning.
• 🔄 Session management (multiple chats, persistent state).
• 🐳 Docker Compose for local development.
• ☁️ Deployment ready for Render (frontend) and Cloud Run (backend).

---

🛠️ Tech Stack

Frontend

• Next.js (React framework)
• tRPC (typesafe APIs)
• React Query
• TailwindCSS + ShadCN UI
• Deployed on Render

Backend

• FastAPI (Python web framework)
• Unstructured (PDF parsing)
• Chroma (vector database)
• Prisma (ORM, SQLite)
• LangChain (summarization & query)
• Deployed on Google Cloud Run

Infrastructure

• Docker & Docker Compose
• Persistent volumes (local dev)
• Future support for managed DBs (Postgres, Pinecone)

---

⚙️ Local Development

Clone the repo and start everything with Docker Compose:

git clone https://github.com/Exynos21/PaperRAG.git
cd PaperRAG

# Copy env files
cp backend-python/.env.example backend-python/.env
cp frontend/.env.local.example frontend/.env.local

# Start services
docker-compose up --build

Services

• Frontend → http://localhost:3000
• Backend (Swagger UI) → http://localhost:8000/docs

---

🌍 Deployment

Frontend (Render)

• Deploy frontend/ as a Static Site on Render.
• Set environment variables:

NEXT_PUBLIC_BACKEND_URL=https://paperrag-backend-296287667274.us-central1.run.app

Backend (Cloud Run)

Build and push the backend Docker image to GCP Artifact Registry.

Deploy with Cloud Run:

gcloud run deploy paperrag-backend \
  --source ./backend-python \
  --platform managed \
  --allow-unauthenticated \
  --region us-central1

⚠️ Note: On Cloud Run, file storage (uploads/, chroma_db/, prisma/dev.db) is ephemeral.
For production, consider using a VM or Render with a persistent disk.

---

📌 Future Scope

• ✅ Migrate from SQLite → Postgres (persistent, scalable)
• ✅ Replace local Chroma DB → Pinecone / Weaviate / Milvus
• 🔐 Add authentication (login, API keys)
• 📊 Add dashboards to view ingested files & queries
• 📱 Mobile-friendly UI