About Sift

Sift is a DOM-first, privacy-conscious browser extension designed to help users filter content on supported feeds. It operates without a backend, accounts, analytics, or hidden scraping, ensuring a privacy-focused experience. Sift runs locally in your browser, identifying visible posts or articles on supported websites. It then classifies this content against your interests using your own model provider, adds visual quality indicators, and can optionally hide low-value content. Currently, Sift supports X/Twitter, Reddit, and Substack.

Key Features:

• DOM-First Approach: Inspects only content visible in the current tab's DOM, avoiding background crawling or server-side feed indexing.
• Privacy-Conscious: Operates locally with no backend, no data collection, and no selling of user data.
• BYOK (Bring Your Own Key): Your API key for model providers stays within trusted extension contexts and is never injected into webpage JavaScript.
• Session-Only Data: Preferences can be stored locally, but post text, browsing history, and classification logs are not persisted.
• Supported Sites: X/Twitter, Reddit, Substack.
• Architecture: Built with Manifest V3 and TypeScript, featuring a background service worker, content scripts, popup UI, options UI, and modular site adapters.

Sift aims to provide a cleaner, more focused browsing experience by empowering users to control the quality of the content they see.