News
Fetch and present the latest headlines using the fetch_web_page tool.
Categories and Sources
| Category | Primary |
|---|---|
| 综合 / General | 新华网 https://www.xinhuanet.com/ |
| 政治 / Politics | 人民网·中共新闻网 https://cpc.people.com.cn/ |
| 财经 / Finance | 中国经济网 http://www.ce.cn/ |
| 社会 / Society | 中国新闻网·社会 https://www.chinanews.com/society/ |
| 国际 / World | CGTN https://www.cgtn.com/ |
| 科技 / Tech | 科技日报 https://www.stdaily.com/ |
| 体育 / Sports | CCTV体育 https://sports.cctv.com/ |
| 娱乐 / Entertainment | 新浪娱乐 https://ent.sina.com.cn/ |
Workflow
1. Identify what the user wants
- Single category: fetch that category's URL.
- Multiple categories: fetch all relevant URLs in parallel (call
fetch_web_pagemultiple times in the same response). - No category specified: default to General (新华网) or ask if ambiguous.
2. Fetch the page
Call fetch_web_page with a generous max_chars to capture enough headlines:
{
"url": "https://36kr.com/",
"max_chars": 10000
}
Use max_chars between 8000–12000 for news homepages — smaller values cut off too many headlines.
3. Extract headlines
From the returned text, look for repeating patterns: headline titles, brief descriptions, timestamps. Most Chinese news sites return plain-text content with clear headline structures after HTML is stripped.
Focus on:
- The headline (title of the story)
- A one or two sentence summary (from the teaser or first paragraph)
- The date/time if visible in the text
- The source name
4. Present results
Format your reply like this (adapt to Chinese or English based on the user's language):
📰 [Category] 最新资讯 — 来源:[Source Name]
[Headline] [One or two sentence summary.] (date/time if available)
[Headline] [Summary.]
(more headlines...)
🔗 查看完整报道:[Source URL]
- List 5–10 headlines per category. More is fine if the user asked for an overview.
- If the user asked for multiple categories, present each as a separate section.
- Match the user's language: reply in Chinese if they asked in Chinese, English if in English.
Error handling
| Situation | What to do |
|---|---|
| Site times out or returns an error | Try the fallback URL from the table |
| Content looks like JS placeholders (no readable text) | Use the fallback; mention the site may need a browser |
| Both primary and fallback fail | Tell the user and suggest they open the URL directly |
| Content is mostly ads/navigation with few headlines | Increase max_chars to 15000 and retry |
Tips
- Fetch multiple categories in parallel (same response, multiple tool calls) to respond faster.
- Don't repeat the same headline across categories if sources overlap.
- Always include the source URL in your reply so the user can read the full story.
- If the user specifies a city or region (e.g., "北京新闻", "上海财经"), note this in your reply but use the same national sources — regional editions are not in this list.