Travel Information and News
Aggregates travel news, destination info, and reviews from multiple sources.
⚠️ CRITICAL RULE: Output language MUST match the user's query language. If the user writes in Chinese, ALL output (titles, content, summaries) must be in Chinese. If in Japanese, output in Japanese. Never return raw English results when the query is in another language.
Quick Start
# Required: TAVILY_API_KEY in env or ~/.openclaw/.env
# Optional: BRAVE_API_KEY for Brave Search fallback
python scripts/search.py --query "Tokyo travel March 2026" --format text
python scripts/search.py --query "京都賞櫻推薦" --format pdf --output result.pdf
python scripts/search.py --query "Bali hotel reviews" --format docx --output result.docx
Search Sources
| Source | Required | When Used |
|---|---|---|
| Tavily | ✅ Yes | Primary search for all queries |
| Brave Search | ❌ No | Fallback when Tavily results insufficient |
| Browser (Xvfb+Chromium+Puppeteer) | ❌ No | Sites Tavily/Brave can't reach (Xiaohongshu, X/Twitter, etc.) |
Installing Optional Sources
Brave Search: Set BRAVE_API_KEY env var.
Browser suite (三件套): Requires three components working together:
- Xvfb — Virtual framebuffer (provides a fake display for Chromium, default: 1200x720x24)
- Chromium — Browser engine
- Puppeteer (Node.js) — Controls Chromium programmatically
Why not headless mode? Some websites block headless browsers. Running Chromium on a virtual display (Xvfb) makes it appear as a real browser, bypassing most detection.
Note: For advanced interactions requiring simulated clicking (e.g., navigating paginated results, clicking "load more" buttons), install the desktop-control skill from ClawHub:
npx clawhub install desktop-control
Installation:
# Xvfb + Chromium
apt-get install -y xvfb chromium
# Puppeteer
npm install puppeteer
Parameters
All parameters can be customized. Defaults are provided.
| Parameter | Default | Description |
|---|---|---|
query |
required | Search query in any language |
time_range |
pm |
Time filter: pd(24h), pw(week), pm(month), py(year), or custom |
region |
ALL |
Region code: ALL, US, CN, JP, TW, etc. |
max_results |
unlimited | Limit number of results (0 = unlimited) |
format |
text |
Output format: text, docx, pdf |
output |
stdout | Output file path |
use_brave |
false |
Enable Brave Search fallback |
use_browser |
false |
Enable browser scraping for blocked sites |
no_reviews |
false |
Disable review aggregation (reviews ON by default) |
language |
auto | Output language (auto = follow query language) |
Output Formats
- text — Plain text summary to stdout
- docx — Word document with structured sections
- pdf — PDF with Noto Sans SC font (supports CJK)
Review Sources
When reviews are requested, the skill attempts to aggregate ratings from:
- TripAdvisor (via browser scraping)
- Google Maps / Google Reviews (via browser scraping)
Workflow
- Receive user query
- Detect query language — output MUST match this language unless user specifies otherwise
- Search Tavily with query + parameters
- If results insufficient → optionally try Brave Search
- If specific sites blocked → optionally use browser suite
- If reviews requested → aggregate ratings
- Compile and deduplicate results
- Translate all results to detected/specified language (critical: query language ≠ result language)
- Format output (text / docx / pdf)
Limits
- Browser suite requires ~512MB RAM for Xvfb + Chromium
- Review scraping depends on site structure (may break if sites change)
- Rate limits apply to Tavily and Brave APIs
