AI Content Detector

Architecture Overview

Built for zero-ops, privacy-first operation, and a predictable click-to-scan user experience.

Core flow

  1. User triggers scan from popup/context menu/shortcut.
  2. Background service worker routes the request.
  3. Content script collects visible images or selected text.
  4. Offscreen document performs local inference (or BYO provider call).
  5. Results return to content script and render badges.

Design constraints

  • No maintainer-operated backend (0 ops).
  • No auto-scan as user scrolls.
  • Local mode keeps data on-device.
  • BYO key mode sends data only to user-selected provider.

System map

Architecture diagram for extension message and inference flow

Popup / Options

React UIs for user actions and settings management.

Background + Offscreen

Message orchestration and long-lived inference context.

Content script

In-page scan targets, overlay rendering, and result display.

Deep technical docs

This page is the high-level summary. The full implementation detail lives in the repository docs: