# Architecture

This page gives a one-diagram overview of pdf2image: dependency injection on load, the public `pdf2image` class, the internal progress overlay, and the CDN libraries they rely on.

## System Overview

```mermaid
graph TB
    subgraph On load
        Boot[Bootstrap IIFE] --> Inject[Inject preconnect / preload / script / stylesheet]
    end
    subgraph CDN
        PDFJS[pdf.js 3.4.120<br/>+ worker + CMap]
        JSZip[JSZip 3.7.1]
        CSS[pdf2image.css @latest]
    end
    Inject --> PDFJS
    Inject --> JSZip
    Inject --> CSS
    App[Caller] --> Class[pdf2image]
    Class -->|convert| PDFJS
    PDFJS --> Canvas[canvas.toDataURL]
    Canvas --> Images[images array]
    Class -->|download| JSZip
    Images --> JSZip
    JSZip --> Anchor[a download click]
    Class --> Loading[pdf2imageLoading overlay]
    CSS --> Loading
```

## Layers

| Layer | Component | Responsibility |
|---|---|---|
| Load | Bootstrap IIFE | Attaches pdf.js, JSZip, and the stylesheet to `<head>` |
| Public API | `pdf2image` class | Normalizes options; `convert()`, `download()`, `images` |
| Internal UI | `pdf2imageLoading` | Creates and updates the progress overlay, switching labels by locale |
| External dependencies | pdf.js, JSZip, browser APIs | Parse and render the PDF, build the ZIP, trigger the download |

## Cross-Cutting Principles

- **No bundled dependencies**: all third-party code is fetched from CDNs at runtime, so the package itself is one small file
- **Client-side only**: the PDF never leaves the browser, and there is no server component
- **Single entry point**: `pdf2image` is the only public class; `pdf2imageLoading` is not exposed

## Further Reading

- [Full architecture document (module diagrams, data flow, state machine)](https://github.com/pardnio/pdf2image/blob/main/doc/architecture.md)
- [Core Concepts](/core-concepts)