Skip to content

Sonra0/Collar_AI

Repository files navigation

CollarAI

Real-time body language feedback for video calls using AI vision analysis. Works with Google Meet, Zoom, Microsoft Teams, Slack, Discord, and Webex.

Features

  • Real-time monitoring every 30 seconds during active meetings
  • Smart alerts for posture, facial expression, hand gesture, and appearance issues
  • Post-meeting summary with scores, trend chart, and top action items
  • Privacy-first behavior using your own API key, optional ephemeral mode, and local session storage
  • Configurable sensitivity, notifications, and retention settings

Tech Stack

  • Chrome Extension Manifest V3
  • Vanilla JavaScript (ES6+)
  • Webpack 5 + Babel
  • Anthropic Claude API and OpenAI API support
  • Chart.js for summary visualizations

Installation & Setup

Prerequisites

  • Node.js 18+
  • npm
  • Google Chrome
  • Anthropic or OpenAI API key

Development Setup

  1. Install dependencies:
npm install
  1. Build extension:
npm run build
  1. Load in Chrome:
  • Open chrome://extensions/
  • Enable Developer mode
  • Click Load unpacked
  • Select dist/
  1. Configure API key:
  • Open extension popup
  • Pick provider and paste key
  • Set sensitivity/privacy preferences (alerts, retention, ephemeral mode)
  • Click Validate key
  • Save settings

Usage

  1. Join a video call on any supported platform (Google Meet, Zoom, Teams, Slack, Discord, or Webex) with your camera enabled.
  2. Extension auto-starts monitoring when self video is detected.
  3. You receive live notifications based on configured sensitivity.
  4. When meeting ends, summary page opens automatically.

Available Scripts

npm run build
npm run dev
npm run frame-recorder
npm run lint
npm run format

Project Structure

src/
  background/   service worker: analysis + notifications + sessions
  content/      video capture + platform detection
  popup/        settings/status UI
  summary/      post-meeting report UI + chart
  utils/        constants, API client, storage wrapper

Troubleshooting

Save captured frames to project folder

  • Run npm run frame-recorder from the project root.
  • This starts a local server at http://127.0.0.1:3131 and saves frames to pictures/.
  • Keep it running while the extension is active.

Extension not detecting meeting video

  • Confirm camera is enabled in your video call platform.
  • If the extension was installed while a meeting was already open, wait a few seconds or open the popup once; the service worker now auto-attaches the content script to existing tabs.
  • If detection still does not start, refresh the meeting tab.
  • Check content script console logs in DevTools.

API analysis fails

  • Re-check provider + key in popup.
  • Confirm API credits/rate limits.
  • Ensure extension has host permissions (included in manifest).

Notifications not appearing

  • Confirm Chrome notifications are enabled.
  • Confirm notifications toggle is on in popup.
  • Check cooldown behavior (2-minute anti-spam window).

Cost Notes

  • Anthropic/OpenAI image analysis incurs provider-side API costs.
  • Approximate cost depends on model and meeting duration.

Demo Assets

  • Pitch outline: docs/PITCH.md
  • Screenshot folder: docs/screenshots/

License

MIT

About

Collar_AI for Dev Season of Code

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors