Technology News Search

• 🌍 International: TechCrunch, The Verge, Wired, Ars Technica, MIT Technology Review, VentureBeat, Hacker News
• 🇨🇳 China: 36氪, 虎嗅, IT之家, 机器之心, 量子位, 钛媒体, 掘金, InfoQ中文站, 开源中国, 博客园, SegmentFault, V2EX, 极客公园, 爱范儿, PingWest品玩, 少数派, 雷锋网, 阿里云开发者

Smart Source Routing

Enhanced with Domain Aliases + Network Adaptation: The skill automatically detects technical domains from your keyword and searches only relevant sources. Network accessibility is detected silently in the background - if global sources are not accessible, the system automatically uses China sources only. Detection result is cached for 5 minutes.

How it works:

1. Silently detect network environment (cached for 5 minutes)
2. Filter sources by network accessibility (global vs China-only)
3. Analyze keyword (supports English and Chinese)
4. Detect technical domains (e.g., "Electron" → frontend, "ChatGPT" → AI, "web" → frontend via alias)
5. Search relevant sources + general sources
6. Rank by heat score and return results

Examples (assuming global network accessible):

• "Electron 技术资讯" → Detects: {general, frontend} → Searches: ~37 sources
• "web development" → Detects: {general, frontend} → Searches: ~37 sources (alias: "web" → frontend)
• "前端框架" → Detects: {general, frontend} → Searches: ~37 sources
• "ChatGPT 最新消息" → Detects: {general, ai} → Searches: ~30 sources
• "ML models" → Detects: {general, ai} → Searches: ~30 sources (alias: "ML" → ai)
• "机器学习" → Detects: {general, ai} → Searches: ~30 sources
• "Docker 安全漏洞" → Detects: {general, devops, security} → Searches: ~32 sources
• "云计算" → Detects: {general, devops} → Searches: ~28 sources (alias: "云" → devops)
• "database optimization" → Detects: {general, backend} → Searches: ~35 sources (merged: database → backend)
• "运维自动化" → Detects: {general, devops} → Searches: ~28 sources
• "树莓派 IoT" → Detects: {general, hardware} → Searches: ~23 sources
• "OpenAI ChatGPT" → Detects: {general, ai} → Searches: ~30 sources (company + product keywords)
• "技术新闻" (generic) → Detects: {general} → Searches: ~25 general sources only

Network Adaptation (completely transparent to user):

• 🌍 Global accessible: Uses all 75 sources
• 🇨🇳 China-only network: Automatically switches to 18 China sources (掘金, InfoQ中文站, 开源中国, 博客园, SegmentFault, V2EX, 36氪, 虎嗅, IT之家, 机器之心, 量子位, 钛媒体, 极客公园, 爱范儿, PingWest品玩, 少数派, 雷锋网, 阿里云开发者)
• ⚡ Fast detection: 3-second timeout with 5-minute cache
• 🔇 Silent operation: No user notification, completely seamless

To disable smart routing and search all available sources (respects network availability):

python3 search_news.py "keyword" --all-sources

Supported Technical Domains

The smart routing system recognizes keywords in these 9 core domains (English and Chinese):

1. Frontend/Web: React, Vue, Angular, Electron, JavaScript, TypeScript, Webpack, Vite, Vercel, Netlify, JAMstack, MERN, 前端, 网页, 界面
2. Backend: Python, Go, Java, Rust, Node.js, Django, Flask, Spring, MySQL, PostgreSQL, MongoDB, Redis (databases merged), LAMP, LEMP, 后端, 服务器, API, 数据库, 存储
3. Mobile: Android, iOS, Flutter, React Native, Swift, Kotlin, 移动开发, 手机, App
4. AI/ML: AI, ChatGPT, LLM, OpenAI, Anthropic, Google AI, DeepMind, Machine Learning, PyTorch, TensorFlow, Copilot, 人工智能, 机器学习, 大模型, 百度, 阿里
5. DevOps: Docker, Kubernetes, CI/CD, Jenkins, Terraform, Ansible, AWS, Azure, GCP, Cloud Computing (cloud merged), HashiCorp, 运维, 部署, 云计算, 云服务, 阿里云, 腾讯云
6. Hardware: Arduino, Raspberry Pi, IoT, ESP32, 硬件, 物联网, 嵌入式
7. Security: Security, Vulnerability, CVE, Exploit, Encryption, 安全, 漏洞, 网络安全, 信息安全
8. OS: Linux, Windows, macOS, Kernel, Ubuntu, 操作系统, 内核, 系统
9. Blockchain: Ethereum, Bitcoin, Web3, Smart Contract, Solana, DeFi, NFT, 区块链, 加密货币, 比特币, 以太坊

Domain Aliases (auto-resolved):

• "web", "网站" → Frontend
• "database", "db", "数据库", "存储" → Backend
• "cloud", "云", "云服务" → DevOps
• "ML", "machine-learning", "数据科学" → AI
• "ops", "SRE", "infrastructure" → DevOps
• "fe" → Frontend
• "be", "server", "服务端" → Backend
• "IoT", "embedded" → Hardware
• "infosec", "cybersecurity", "信息安全" → Security

Quick Start

When user asks: "Search for Electron tech news" or "搜索 Electron 技术资讯"

Execute:

# macOS/Linux
cd scripts
python3 search_news.py "Electron" --limit 15

# Windows
cd scripts
python search_news.py "Electron" --limit 15

The script will:

1. Detect domains: {general, frontend}
2. Search 26 relevant sources (13 general + 13 frontend)
3. Output JSON with articles ranked by heat score

Read the JSON and present results in Markdown format with translations.

Workflow

1. Extract keyword from user query

   • English examples: "Search for OpenAI news", "Find articles about ChatGPT"
   • Chinese examples: "搜索 苹果 的科技新闻", "查找 AI 相关资讯"

2. Run search script

   # macOS/Linux (with smart routing)
   cd scripts
   python3 search_news.py "[keyword]" --limit 15 --max-per-source 5
   
   # To search all sources (disable smart routing)
   python3 search_news.py "[keyword]" --limit 15 --all-sources
   
   # Windows
   cd scripts
   python search_news.py "[keyword]" --limit 15 --max-per-source 5
   

   Parameters:

   • --limit 15: Fetch up to 15 articles from each source
   • --max-per-source 5: Display max 5 articles per source (ensures diversity)
   • --no-balance: Disable balancing (show all results sorted by heat)
   • --all-sources: Search all 63 sources (disable smart routing)

3. Read JSON output

   • Script outputs to stdout
   • Contains: keyword, total_found, search_time, results array
   • Each result has: title, summary, url, published_at, source, language, heat_score, duplicate_sources

4. Translate English content

   • For articles with language: "en", translate title and summary to Chinese
   • Keep technical terms in English (AI, GPT, API, SDK, etc.)
   • Format: English Title / 中文翻译

5. Format results by heat tier

   • 🔥 Hot News (90+): Top stories appearing on multiple sources or very recent
   • 📈 Trending (60-89): Moderately popular or recent stories
   • 📰 Related (<60): Other relevant matches

6. Present as Markdown

   • See Output Format section below

Source Balancing

Default behavior: The script limits each source to 5 articles to ensure diversity across different news sources.

Why balancing matters:

• Prevents single sources (like Hacker News) from dominating results
• Ensures exposure to different editorial perspectives
• Provides better coverage across international and Chinese sources

Customization:

# Show more articles per source
python3 search_news.py "AI" --max-per-source 10

# Disable balancing (show all by heat score only)
python3 search_news.py "AI" --no-balance

How it works:

1. Fetch articles from all sources (up to --limit per source)
2. Calculate heat scores for all articles
3. Sort by heat score (highest first)
4. Apply diversity filter: keep top --max-per-source from each source
5. Result: Balanced mix of high-quality articles from diverse sources

Output Format

Present search results in this Markdown format:

# 🔍 "[Keyword]" Technology News

> 📊 Found 12 articles from 7 sources
> 🕐 Search time: 2026-02-18 14:30

---

## 🔥 Hot News (Heat 90+)

### 1. OpenAI Announces GPT-5 Release Date / OpenAI 宣布 GPT-5 发布日期
**Source**: TechCrunch | **Published**: 2h ago | **Heat**: ⭐⭐⭐⭐⭐ (95)

**Summary**: OpenAI CEO Sam Altman revealed that GPT-5 will launch in Q2 2026... / OpenAI CEO Sam Altman 透露 GPT-5 将在 2026 年第二季度推出...

🔗 [Read more](https://techcrunch.com/2026/02/18/openai-gpt5)

*💡 Also on: The Verge, Wired, MIT Tech Review*

---

### 2. GPT-5 性能测试曝光：超越人类专家水平
**Source**: 机器之心 | **Published**: 5h ago | **Heat**: ⭐⭐⭐⭐ (88)

**Summary**: 据可靠消息源透露，GPT-5 在多项基准测试中已超越人类专家水平...

🔗 [Read more](https://jiqizhixin.com/articles/2026-02-18-gpt5)

---

## 📈 Trending News (Heat 60-89)

### 3. [Title] / [Translation]
...

## 📰 Related News (Heat <60)

### 10. [Title]
...

Format Notes:

• Group by heat score tiers (Hot 90+, Trending 60-89, Related <60)
• Always translate English content to Chinese
• Include source attribution and publish time
• Show duplicate sources with "Also on: ..." line
• Use star ratings for heat visualization (5 stars = 90-100, 4 stars = 70-89, etc.)
• Provide clickable links to original articles

Source Configuration

Sources are configured in references/sources.json.

To enable/disable sources: Edit the JSON file and set "enabled": true/false.

To add new sources: Add entry to sources array with required fields (id, name, url, type, language, category).

The heat score (0-100) combines multiple factors:

• Multi-source bonus: +20 per duplicate source (same story on multiple sites)
• Time decay:
  • 24h or less: +40 points (very fresh)
  • 24-48h: +28 points (recent)
  • 48-72h: +16 points (somewhat recent)
  • 72h+: +8 points (older)
• Keyword match quality:
  • Exact title match: +30
  • Partial title match: +15
  • Summary match: +5
• HN engagement: +1 per 10 points (max +20)
• Reddit engagement: +1 per 10 upvotes (max +20, NEW)
• Official source bonus: +10 (for official blogs, NEW)
• Base score: 20

Articles appearing on multiple sources rank higher, indicating broader industry interest. Official blog posts from React, Vue, Docker, etc. get priority. Reddit posts with high upvotes get boosted.

Notes

• Always translate: English titles and summaries should have Chinese translations
• Preserve technical terms: Keep AI, GPT, API, SDK, LLM, etc. in English
• Clean presentation: Use Markdown only, no complex HTML artifacts
• Cross-reference duplicates: Show which sources covered the same story
• Keyword matching: Script searches in both titles and summaries across all languages

Troubleshooting

No results found:

• Try broader keywords (e.g., "AI" instead of "GPT-5 benchmarks")
• Check if sources are accessible (firewall/network issues)
• Verify sources are enabled in references/sources.json

Script errors:

• Ensure Python 3.6+ is installed (python3 --version on macOS/Linux, python --version on Windows)
• Check network connectivity
• Review stderr output for specific error messages
• Some sources may be temporarily unavailable

Windows-specific notes:

• Use python instead of python3 in commands
• Some Chinese sources may have encoding issues if system locale is not UTF-8 (already handled in code)

Slow performance:

• Reduce --limit parameter (default is 15)
• Disable some sources in references/sources.json
• Network speed affects RSS fetching time