# PDF Form Filler

A web application that allows uploading PDF forms, converting them to HTML via an LLM, and filling out the original PDFs with form data.

## Prerequisites

- **Go 1.21+**
- **pdftk** (for extracting form fields and filling PDFs)
- **pdftoppm** (from poppler-utils, for converting PDF pages to images for the LLM)
- **OpenAI API key** (or compatible LLM API)

### Installing dependencies

```bash
# Ubuntu/Debian
sudo apt-get install pdftk poppler-utils

# macOS
brew install pdftk-java poppler

# Windows
# pdftk: https://www.pdftk.com/
# pdftoppm: install poppler via winget or vcpkg
```

## Configuration

Set these environment variables:

| Variable | Required | Description | Default |
|----------|----------|-------------|---------|
| `OPENAI_API_KEY` | Yes | Your OpenAI API key | - |
| `OPENAI_API_URL` | No | Custom LLM API endpoint | `https://api.openai.com/v1/chat/completions` |
| `OPENAI_MODEL` | No | LLM model to use | `gpt-4o` |

## Building & Running

```bash
# Build
go build -o pdf-form-filler .

# Run
./pdf-form-filler

# Or run directly
OPENAI_API_KEY=your-key go run main.go
```

The server starts on `:8080` by default.

## Usage

1. Go to `http://localhost:8080/admin`
2. Upload a PDF that contains form fields
3. Click "Convert to HTML" to generate a web form using the LLM
4. Click "Open Form" to view the public form at `/form/{id}`
5. Fill out the form and submit — the filled PDF will download automatically

## Routes

| Route | Method | Description |
|-------|--------|-------------|
| `/admin` | GET | Admin dashboard — list and manage forms |
| `/admin/upload` | POST | Upload a new PDF form |
| `/admin/convert` | POST | Convert a PDF to HTML via LLM |
| `/admin/delete` | POST | Delete a form |
| `/form/{id}` | GET | Public form page (HTML) |
| `/form/{id}` | POST | Submit form — downloads filled PDF |

## How It Works

1. **Upload**: PDF is saved to the `pdfs/` directory, record created in SQLite
2. **Extract**: `pdftk dump_data_fields` extracts form field names and types
3. **Convert**: LLM receives field metadata and generates matching HTML form
4. **Fill**: On submission, form data is written to an FDF file, `pdftk fill_form` merges it with the original PDF

## Database

SQLite database at `db/forms.db` — auto-created on first run.