Bouncer Feed Filter
Skill by ara.so — Daily 2026 Skills collection.
Bouncer is a browser extension (Chrome/Edge/iOS) that uses AI to filter unwanted posts from Twitter/X feeds in real time. Users define filters in plain language ("crypto", "engagement bait", "rage politics"), and Bouncer classifies and hides matching posts using AI models — local (WebGPU via WebLLM) or cloud (OpenAI, Gemini, Anthropic, OpenRouter, Imbue).
Repository Structure
Installation & Build
From Source (Chrome/Edge)
Load in Chrome:
- Go to
chrome://extensions - Enable Developer mode
- Click Load unpacked → select
Bouncer/folder - Navigate to
twitter.comorx.com
Development Build (watch mode)
Production Build
AI Backend Configuration
Bouncer supports multiple providers. Configure via the extension popup Settings panel.
Provider / Model Matrix
API keys are stored in Chrome's chrome.storage.local — never hardcoded.
Core Architecture
1. MutationObserver — Content Script
The content script watches the Twitter feed for new posts:
2. Post Extraction — Twitter Adapter
3. Classification Request
4. Hiding Filtered Posts
Adding a New AI Provider
Register it in the provider registry:
Result Caching
Bouncer caches classification results so repeated posts don't trigger new inference calls:
Filter Management
Filters are stored and retrieved via chrome.storage.sync:
Local WebGPU Models (WebLLM)
Local models run entirely in-browser via WebGPU — zero data sent externally:
Chrome Extension Manifest Key Points
Troubleshooting
Common Patterns
Check if a post should be processed (before API call):
Storing API key securely (popup UI):


