zenitlab/ai-hot-radar

AI Hot Radar is an open-source news aggregator specifically for AI industry news. It pulls from more than eight sources continuously, including Twitter/X (via a paid API), Bing search, Hacker News, Bilibili, IT House, and the Google DeepMind blog, then uses AI scoring to filter out the noise and surface only the items worth reading. The README and interface are in Chinese, and the project is aimed at people working in or following the AI industry. The filtering works in two stages. A cheaper AI model does a first pass to remove content that is clearly not about AI. A more thorough scoring pass then evaluates the remaining items on five dimensions and assigns a tier. Only items that score above a threshold appear in the Curated view. Duplicate items from multiple sources about the same event are merged, with more authoritative sources (like an official company blog) taking priority over secondary coverage. The application has four main sections. The Radar view shows all incoming items sorted by publication time. The Curated view shows only the high-scoring items. The AI Daily Digest is generated automatically each day at 8am Beijing time and covers six categories: key events, model news, domestic and international updates, AI products, community discussion, and research papers. The Keywords section lets you subscribe to specific topics, with AI expanding your search terms to include synonyms and related phrases, matching items are pushed to you in real time via WebSocket and can trigger email alerts. The system also exposes everything it aggregates through three integration channels: an Agent Skill that works with Claude Code and Cursor, RSS feeds for standard feed readers like Feedly or Inoreader, and a small REST API with five unauthenticated JSON endpoints suitable for connecting to workplace bots or automation workflows. To run it locally, you need Node.js 18 or later and an API key for any OpenAI-compatible model (the README examples use Alibaba Cloud, DeepSeek, and similar options). The backend runs on NestJS with SQLite via Prisma, the frontend is React 19 with Vite and Tailwind CSS.