# Smart Document Search

Upload PDFs and text files, then ask questions in natural language. AI finds answers from your documents with source citations.

## Features

- **Upload PDF/TXT** - Drag-and-drop document upload
- **Auto-summarization** - AI summarizes each uploaded document
- **Natural language Q&A** - Ask anything about your documents
- **Source citations** - Every answer references which document it came from
- **Conversation memory** - Follow-up questions understand context
- **Multi-document** - Search across all uploaded documents simultaneously

## Quick Start

```bash
# Backend
cd backend
cp .env.example .env
npm install
npm start            # Port 3004

# Frontend
cd frontend
npm install
npm run dev          # Port 5177
```

### AI Setup (choose one - both are FREE)

| Provider | Best For | Setup |
|----------|----------|-------|
| **Groq** (cloud) | Deployment, sharing | Get free key at [console.groq.com/keys](https://console.groq.com/keys), add `GROQ_API_KEY` to `.env` |
| **Ollama** (local) | Development, offline | Install from [ollama.com](https://ollama.com), run `ollama pull llama3.2:3b` |

If both are configured, Groq is used first with Ollama as fallback.

## Tech Stack

- **Frontend**: React 19 + Vite + react-dropzone
- **Backend**: Express.js + pdf-parse
- **AI**: Groq (cloud) / Ollama (local) - both free, no paid API needed
- **Search**: Text chunking with keyword-based retrieval (RAG pattern)