# PaperMind AI

PaperMind AI is a web application designed to make reading and understanding research papers easier. Research papers can be lengthy and difficult to follow, especially when trying to identify their main findings, methodologies, and related work.

With PaperMind AI, users can upload a research-paper PDF and explore its content through AI-powered analysis, structured insights, and an interactive chat interface.

# Screenshots

## 📸 Screenshots

### Dashboard
![PaperMind AI Dashboard](screenshots/dashboard.png)

## Signup
![PaperMind AI Dashboard](screenshots/signup.png)

### Research Paper Analysis
![Research Paper Analysis](screenshots/summary.png)

### Chat with Paper
![Chat with Paper](screenshots/chat.png)

### Technical Stack
![Technical Stack](screenshots/TechStack.png)

### Presentation
![Presentation](screenshots/presentation.png)


## Project structure

```text
papermind-final/
├── src/                 # React frontend
├── server/              # Node.js/Express backend
│   ├── src/
│   ├── .env.example
│   └── package.json
├── package.json
└── vite.config.js
```

## 1. Install frontend

From the project root:

```bash
npm install
```

## 2. Configure backend

Open a terminal in `server`:

```bash
cd server
npm install
```

Copy `.env.example` to `.env` and add your own Gemini API key:

```env
PORT=5000
MONGODB_URI=mongodb://127.0.0.1:27017/papermind
JWT_SECRET=replace_with_a_long_random_secret
GEMINI_API_KEY=your_actual_gemini_api_key_here
GEMINI_MODEL=gemini-3.8-flash
MAX_ANALYSIS_CHARS=80000
ANALYSIS_CHUNK_CHARS=18000
```

Never commit or share `.env`.

## 3. Start MongoDB

Make sure local MongoDB is running at:

```text
mongodb://127.0.0.1:27017
```

## 4. Start backend

From `server`:

```bash
npm run dev
```

The backend runs at `http://localhost:5000`.

## 5. Start frontend

Open a second terminal at the project root:

```bash
npm run dev
```

Open the Vite URL shown in the terminal, normally `http://localhost:5173`.

## Paper analysis flow

```text
PDF upload
  -> PDF validation
  -> pdf-parse 2.x text extraction
  -> readable-text validation
  -> text cleaning and chunking
  -> Gemini paper-grounded analysis
  -> structured JSON validation
  -> MongoDB storage
  -> React analysis tabs
```

The application does not intentionally generate sample research-paper results. Missing information is represented as `Not mentioned in the paper.`. Scanned/image-only PDFs are reported as requiring OCR rather than being analyzed as if text had been extracted.

## Notes
The Gemini API key stays on the Node.js backend.
The current PDF limit is 25 MB.
The backend uses pdf-parse 2.x and its PDFParse API.
The Chat with Paper endpoint is restricted to the selected paper's stored extracted text.