---
title: Card & Letterhead OCR Extractor
emoji: 🪪
colorFrom: red
colorTo: green
sdk: gradio
app_file: app.py
app_port: 7860
pinned: false
---

# Visiting Card & Letterhead OCR Extractor

Extracts structured contact data from visiting cards and letter headers/footers using a two-step NVIDIA AI pipeline:

**nemoretriever-ocr-v1** → raw text → **nvidia-nemotron-nano-9b-v2** → structured JSON

## Extracted Fields

| Field | Description |
|---|---|
| Company Name | Firm / organisation name |
| Contact Person | Individual's name |
| Designation | Job title / role |
| Mobile | Mobile number(s) |
| Phone | Landline / office number(s) |
| Email | Email address(es) |
| Address | Full postal address |
| PIN | PIN / ZIP / postal code |
| City / State / Country | Location breakdown |
| GST Number | 15-char GSTIN |
| Website | URL |
| Fax | Fax number |

## Files in this Space

| File | Purpose |
|---|---|
| `app.py` | FastAPI app — OCR pipeline + serves `index.html` at `/` |
| `index.html` | Frontend UI (drop / paste / browse images) |
| `requirements.txt` | Python dependencies |
| `README.md` | This file |

> **No Dockerfile needed** — HF Spaces handles the container automatically with `sdk: gradio`.

## Setup on Hugging Face Spaces

1. Create a new Space → SDK: **Gradio**
2. Upload: `app.py`, `index.html`, `requirements.txt`, `README.md`
3. Go to **Settings → Variables and secrets** → add secret:
   - Name: `NVIDIA_API_KEY`
   - Value: your `nvapi-...` key from [build.nvidia.com](https://build.nvidia.com)
4. Space builds automatically. The UI appears at your Space URL.

## Local Usage

```bash
pip install fastapi uvicorn requests python-multipart pydantic
export NVIDIA_API_KEY=nvapi-YOUR_KEY_HERE
python app.py
# Open http://localhost:7860
```

## How to Use the UI

- **Click** the drop zone to browse files
- **Drag & drop** an image onto the drop zone
- **Paste** (`Ctrl+V` / `⌘V`) a screenshot or copied image — no file needed!
- Switch to **Batch mode** to process up to 10 cards at once
- Use **Copy JSON** or **Copy CSV row** to export results