Summary

Siftly is a self-hosted, local organizer for Twitter/X bookmarks designed to transform them into a searchable, categorized, and visual knowledge base. It targets power users and researchers seeking to manage and leverage their saved content without relying on cloud services or browser extensions. The primary benefit is enhanced knowledge discovery through advanced AI-powered search and interactive visualizations, all while keeping data private on the user's machine.

How It Works

Siftly employs a four-stage AI pipeline for bookmark processing: entity extraction (identifying hashtags, URLs, mentions, and tools), vision analysis (tagging images/videos), semantic tagging (generating descriptive tags), and categorization. This multi-modal approach enriches raw tweet data, enabling natural language search by meaning and creating an interactive mindmap visualization of bookmark relationships. The pipeline runs locally, with only configured AI API calls being external.

Quick Start & Requirements

• Primary install / run command: git clone https://github.com/viperrcrypto/Siftly.git && cd Siftly && ./start.sh (recommended, handles dependencies, DB setup, and opens the app). Manual setup involves npm install, npx prisma generate, npx prisma db push, and npx next dev.
• Non-default prerequisites and dependencies: Node.js 18+, npm. AI features require an Anthropic API key or the Claude Code CLI (which requires macOS).
• Estimated setup time or resource footprint: start.sh provides a streamlined setup. Uses a local SQLite database.
• Links: Local app runs at http://localhost:3000.

Highlighted Details

• AI search allows natural language queries (e.g., "funny meme about crypto crashing") across tweet text, image OCR, and tags.
• Interactive mindmap visualizes bookmarks organized by category, allowing exploration of the knowledge graph.
• Supports multiple import methods (bookmarklet, console script) directly from X, eliminating the need for browser extensions.
• Customizable categories with AI-driven classification based on user-defined descriptions.
• Fully local architecture, with data stored in SQLite and external AI calls being the only exceptions.

Maintenance & Community

Developed by @viperr. The README does not specify community channels (like Discord/Slack) or notable contributors/sponsorships.

Licensing & Compatibility

MIT License. This permissive license allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The Claude Code CLI for seamless AI authentication is currently limited to macOS. Users on Linux or Windows must configure an Anthropic API key manually. Vision analysis is performed on image/GIF/video thumbnails.