📄 AskPDF

Chat with your PDFs. Get grounded answers. Jump straight to the source.

An AI-powered document intelligence application that combines RAG, semantic retrieval, Gemini, and page-aware citations to turn static PDFs into interactive, conversational knowledge.

---

✨ What is AskPDF?

AskPDF turns a PDF from a file you read into a document you can talk to.

Upload a document, ask a question in natural language, and AskPDF retrieves the most relevant context before asking Gemini to generate a grounded answer. When the answer comes from a specific part of the document, the application can surface the page reference directly in the conversation and let you jump to that page in the PDF viewer.

The project combines:

• 📄 PDF text extraction
• 🧩 document chunking
• 🔎 embedding-based semantic retrieval
• 🧠 Retrieval-Augmented Generation (RAG)
• 📑 page-aware retrieval
• 🎯 source/page citations
• 🔗 clickable citation navigation
• 💬 conversational follow-up questions

The current implementation uses a React + TypeScript frontend and a Node.js + Express + TypeScript backend, with MongoDB for persistence and Gemini for AI capabilities. fileciteturn13file0L42-L64

---

🎥 Live Demo

👉 Open AskPDF

Upload a PDF → ask a question → inspect the source.

---

🌟 Why AskPDF?

Most document chat applications stop at:

“Here is an answer generated from your PDF.”

AskPDF is designed around a more useful workflow:

“Here is the answer — and here is where it came from.”

That means the system is not only focused on answering questions, but also on making those answers easier to verify, navigate, and trust.

---

🚀 Features

📄 Document Intelligence

• PDF upload and text extraction
• Page offset calculation
• Automatic document chunking
• Temporary file cleanup after processing
• Support for document-grounded conversational queries

🔍 Semantic Retrieval

• Embedding-based retrieval
• Cosine-similarity ranking
• Meaning-aware search instead of simple keyword matching
• Relevant context selection before generation

📑 Page-Aware Retrieval

Ask questions such as:

What is mentioned on page 20?

AskPDF can use the requested page context instead of treating the entire document as an undifferentiated text corpus.

🎯 Grounded Answers & Citations

• Answers are generated from retrieved document context
• Source pages are surfaced with responses
• Citations can be clicked to navigate to the relevant PDF page
• Designed to make answers easier to fact-check

💬 Conversational UX

Ask follow-up questions naturally:

What is the main argument?

Can you explain that in simpler terms?

What evidence supports that?

The application is designed to preserve the conversational context needed for follow-up interactions.

🛡️ Reliability & Production Considerations

• Gemini API error/rate-limit handling
• Required environment-variable validation
• CORS configuration
• Temporary upload cleanup
• Production frontend/backend deployment

---

🧠 How AskPDF Works

AskPDF follows a Retrieval-Augmented Generation (RAG) pipeline with additional page-aware logic.

                         ┌─────────────────────┐
                         │     Upload PDF      │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │   Extract Text      │
                         │ + Page Information  │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │   Chunk Document    │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │ Generate Embeddings │
                         │      via Gemini     │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │   User Question     │
                         └──────────┬──────────┘
                                    │
                         ┌──────────┴──────────┐
                         ▼                     ▼
                Semantic Retrieval      Page Retrieval
                         │                     │
                         └──────────┬──────────┘
                                    ▼
                         ┌─────────────────────┐
                         │ Relevant Context    │
                         │   Selected/Ranked   │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │   Gemini Generation │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │ Grounded Answer +   │
                         │ Page Citations      │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │ Click Citation →    │
                         │ Jump to PDF Page    │
                         └─────────────────────┘

The repository describes this flow as text extraction with page offsets, chunking, embedding generation, semantic or page-aware retrieval, context injection, grounded Gemini generation, and clickable PDF navigation. fileciteturn13file0L107-L131

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         ASKPDF PLATFORM                         │
├──────────────────────────────┬──────────────────────────────────┤
│          FRONTEND            │             BACKEND               │
│                              │                                  │
│  React + TypeScript          │  Node.js + Express + TypeScript │
│  Vite                        │                                  │
│  Tailwind CSS                │  ┌────────────────────────────┐  │
│  TanStack Query              │  │ Controllers / Routes       │  │
│  Axios                       │  └─────────────┬──────────────┘  │
│                              │                │                  │
│  PDF Viewer                  │                ▼                  │
│  Chat Interface              │  ┌────────────────────────────┐  │
│  Citation Navigation         │  │ RAG / PDF / Chat Services │  │
│                              │  └─────────────┬──────────────┘  │
└──────────────┬───────────────┘                │                 │
               │                                │                 │
               └────────────── HTTP ────────────┘                 │
                                                │                 │
                         ┌──────────────────────┼───────────────┐ │
                         │                      │               │ │
                         ▼                      ▼               ▼ │
                    MongoDB                 Gemini          pdf-parse
                         │                      │               │ │
                         └──────────────────────┴───────────────┘ │
└─────────────────────────────────────────────────────────────────┘

The repository is organized around a React/Vite client and a Node/Express server with configuration, controllers, middleware, models, routes, services, utilities, and workers. fileciteturn13file0L139-L164

---

🛠️ Tech Stack

Layer	Technology	Purpose
Frontend	React	Interactive document/chat UI
Language	TypeScript	Type-safe application code
Build Tool	Vite	Fast frontend development/builds
Styling	Tailwind CSS	UI styling
Data Fetching	TanStack Query	Server-state management
HTTP	Axios	API communication
Backend	Node.js + Express	API and application server
Database	MongoDB + Mongoose	Persistence
AI	Gemini API	Embeddings and generation
PDF	pdf-parse	PDF text extraction
Deployment	Vercel + Render	Frontend/backend hosting
Database Hosting	MongoDB Atlas	Managed MongoDB

This stack reflects the repository's documented frontend, backend, AI/RAG, and deployment technologies. fileciteturn13file0L69-L103

---

📁 Project Structure

AskPDF/
│
├── client/                         # React + Vite frontend
│   ├── public/                     # Static assets
│   └── src/
│       ├── assets/                 # Images and global styles
│       ├── components/             # Reusable UI components
│       ├── lib/                    # Utilities and API clients
│       └── pages/                  # Main views/routes
│
├── server/                         # Node + Express backend
│   ├── src/
│   │   ├── config/                 # Configuration/database setup
│   │   ├── constants/              # Shared constants
│   │   ├── controllers/            # Request handlers
│   │   ├── middleware/             # Express middleware
│   │   ├── models/                 # Mongoose models
│   │   ├── routes/                 # API routes
│   │   ├── services/               # RAG, chat, PDF logic
│   │   ├── utils/                  # Helper utilities
│   │   └── workers/                # Background processing
│   └── uploads/                    # Temporary PDF storage
│
├── AskPDF.postman_collection.json  # API testing collection
├── DEVELOPMENT.md                  # Development notes/log
└── README.md                       # Project documentation

The repository also includes a Postman collection for manual API testing. fileciteturn13file0L24-L26

---

🚀 Getting Started

Prerequisites

• Node.js: v18+
• MongoDB: local MongoDB or MongoDB Atlas
• Gemini API Key
• npm

These prerequisites match the project's documented setup. fileciteturn13file0L173-L183

1. Clone

git clone https://github.com/Jyatin/AskPDF.git
cd AskPDF

2. Backend

cd server
npm install

Create server/.env, then start the backend:

npm run dev

Backend development server:

http://localhost:5000

3. Frontend

Open another terminal:

cd client
npm install
npm run dev

Frontend development server:

http://localhost:5173

The repository documents these backend/frontend development flows and ports. fileciteturn13file0L173-L233

---

🔑 Environment Variables

Create server/.env:

# Required
MONGO_URI=mongodb+srv://<user>:<password>@cluster.mongodb.net/askpdf
GEMINI_API_KEY=your_gemini_api_key
PORT=5000

# Required in production
CORS_ORIGIN=https://ask-pdf-vert.vercel.app

🔒 Never commit real credentials or API keys to Git.

The repository currently documents these environment variables, including the production CORS origin. fileciteturn13file0L261-L275

---

📦 Production Build

Backend

cd server
npm run build

Frontend

cd client
npm run build

These production build commands are part of the documented workflow. fileciteturn13file0L235-L257

---

☁️ Deployment

AskPDF is currently documented as deployed using:

Frontend       → Vercel
Backend        → Render
Database       → MongoDB Atlas
AI             → Gemini API

Production

Frontend

https://ask-pdf-vert.vercel.app/

Backend API

https://askpdf-backend-xt83.onrender.com

Health Check

https://askpdf-backend-xt83.onrender.com/health

The repository documents the Vercel frontend, Render backend, MongoDB Atlas database, and Gemini-based AI deployment. fileciteturn13file0L281-L296

---

🧪 Testing

AskPDF currently does not have a formal automated test suite.

For API testing and manual verification, the repository includes:

AskPDF.postman_collection.json

This matches the current project documentation. fileciteturn13file0L300-L304

---

⚠️ Current Limitations

AskPDF is actively evolving. Current documented limitations include:

• Conversation history is currently stored in frontend React state and is lost after a page reload.
• Semantic retrieval currently calculates cosine similarity in application memory rather than using a dedicated vector database.
• Retrieval is not optimized for very large document collections.
• Uploaded PDFs use temporary backend local storage during processing.

These are explicitly documented in the current project README. fileciteturn13file0L308-L315

---

🗺️ Roadmap — V2

[x] PDF upload + extraction
[x] Document chunking
[x] Semantic retrieval
[x] RAG-based QA
[x] Page-aware retrieval
[x] Clickable page citations
[ ] Persistent conversations
[ ] Multi-document conversations
[ ] Improved retrieval/ranking
[ ] Dedicated vector database
[ ] Streaming responses
[ ] User authentication
[ ] Persistent cloud document storage
[ ] Large-document background processing
[ ] Advanced table/image understanding

The planned V2 direction currently includes persistent conversations, multi-document support, improved ranking, scalable vector search, streaming responses, authentication, cloud storage, background processing, security improvements, and richer document understanding. fileciteturn13file0L319-L334

---

🤝 Contributing

Contributions, ideas, bug reports, issues, and pull requests are welcome.

git checkout -b feature/your-feature

# make your changes

git add .
git commit -m "feat: describe your change"
git push origin feature/your-feature

Then open a Pull Request with:

• what changed
• why it changed
• how it was tested
• screenshots or recordings for UI changes

---

📌 Project Status

🟢 Active Development

AskPDF is an evolving project focused on making document interaction more grounded, navigable, and useful through retrieval-augmented AI.

---

👨‍💻 Author

Jyatin Singh

Full-Stack Developer · AI Builder · Open Source Contributor

---

⭐ Support AskPDF

If AskPDF is useful or interesting, consider giving the repository a ⭐.

It helps the project get discovered and encourages continued development.

Upload. Ask. Retrieve. Verify.

📄 AskPDF

---

📄 License

No license is currently specified in the repository.

Add a LICENSE file when you are ready to define how the project may be used, modified, and distributed.