yuuzu-search

name

yuuzu-search

description

Use Yuuzu hosted search gateway (search.yuuzu.net / openclaw-searxng-plus) for web query tasks when the user asks to use "oracle-my websearch", "search.yuuzu.net", or this custom search API. Handles request building, token-authenticated calls to /search, /crawl, /map, and /crawl/site endpoints, parameter tuning (q, topK, freshness, needContent, needCrawl, needMap, needSiteCrawl, format, lang), and quick health/auth diagnostics.

Yuuzu Search Gateway

Use this skill to query Yuuzu's custom search + crawl gateway instead of generic web search tools.

Quick Start

Use the Bash tool with inline curl to call the gateway directly. Do NOT pre-check whether the token exists — just run the command. If it fails, show the error output to the user and stop.

Token resolution (inline, single line):

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//")

Base URL: https://search.yuuzu.net

Default call pattern (search example):

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//") && curl -sS https://search.yuuzu.net/search \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"q":"your query here","topK":5,"freshness":"any"}' | python3 -m json.tool 2>/dev/null || echo "FAILED"

If the curl call returns an error (401, 403, 5xx, connection refused), report the raw error to the user and stop. Do not retry or attempt token diagnostics.

Endpoints Overview

Method	Path	Purpose
GET	/health	Health check
POST	/search	Search + optional crawl/map/site-crawl enrichment
POST	/crawl	Crawl a single URL
POST	/map	Discover URLs from a site (sitemap + link extraction)
POST	/crawl/site	Start async site crawler (returns job ID)
GET	/crawl/site/{job_id}	Poll site crawl progress + paginated results
DELETE	/crawl/site/{job_id}	Cancel a running site crawl

POST /search

Core search endpoint with optional enrichment flags.

Body fields:

Field	Type	Default	Description
q	string (1..500)	required	Query string
topK	int (1..50)	10	Max results
freshness	any|day|week|month	any	Time filter
lang	string	auto-detect	Force language (en, zh-TW)
needContent	bool	false	Fetch page body via lightweight fetcher
needCrawl	bool	false	Deep crawl via Crawl4AI (richer than needContent)
needMap	bool	false	Discover related URLs per result domain
needSiteCrawl	bool	false	Launch async site crawl per result domain
format	markdown|text|html	markdown	Output format (when needCrawl/needSiteCrawl)
stripLinks	bool	false	Remove markdown links from content
crawlMaxDepth	int (1..5)	1	Crawl depth
crawlMaxPages	int (1..20)	1	Crawl page limit
crawlTimeoutMs	int (1000..120000)	15000	Crawl timeout
crawlConcurrency	int (1..10)	3	Parallel crawl workers
crawlOnlyMainContent	bool	true	Extract main content only
crawlBypassCache	bool	false	Skip cached content

Response fields per result:

Field	Type	Description
title	string	Page title
url	string	Page URL
snippet	string	Sanitized snippet
content	string?	Full page content (when needContent/needCrawl)
score	float	Hybrid ranking score
source	string	Search engine source
content_partial	bool	Whether content was truncated
relatedUrls	string[]?	Discovered URLs (when needMap)
siteCrawlJobId	string?	Async job ID (when needSiteCrawl)

Example — basic search:

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//") && curl -sS https://search.yuuzu.net/search \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"q":"OpenClaw ACP runtime","topK":5,"freshness":"week"}'

Example — with crawl enrichment:

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//") && curl -sS https://search.yuuzu.net/search \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"q":"FastAPI best practices","topK":5,"needCrawl":true,"format":"markdown"}'

POST /crawl

Crawl a single URL and return cleaned content.

Body:

Field	Type	Default	Description
url	string	required	HTTP/HTTPS URL
maxDepth	int (1..5)	1	Crawl depth
maxPages	int (1..20)	1	Page limit
timeoutMs	int (1000..120000)	15000	Timeout
onlyMainContent	bool	true	Main content extraction
bypassCache	bool	false	Skip cache
format	markdown|text|html	markdown	Output format
respectRobots	bool	true	Respect robots.txt
stripLinks	bool	false	Remove links

Response: { result: { url, content, title, markdown, success, partial, source }, timing_ms }

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//") && curl -sS https://search.yuuzu.net/crawl \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/article","format":"markdown"}'

POST /map

Discover URLs from a website via sitemap.xml + HTML link extraction.

Body:

Field	Type	Default	Description
url	string	required	Target site URL
includePatterns	string[]	[]	URL glob filters
respectRobots	bool	true	Filter by robots.txt
useSitemap	bool	true	Try sitemap.xml

Response: { urls: string[], total: int, source: { sitemap: int, links: int }, timing_ms }

TOKEN=$(grep 'GATEWAY_AUTH_TOKEN' ~/.bashrc ~/.zshrc 2>/dev/null | head -1 | sed "s/^.*GATEWAY_AUTH_TOKEN=['\"]*//" | sed "s/['\"].*$//") && curl -sS https://search.yuuzu.net/map \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://docs.example.com","includePatterns":["*/api/*"]}'

POST /crawl/site (Async)

Start a background site crawl job. Returns immediately with a job ID.

Body:

Field	Type	Default	Description
url	string	required	Start URL
maxDepth	int (1..10)	3	BFS depth
maxPages	int (1..500)	100	Page limit
format	markdown|text|html	markdown	Output format
respectRobots	bool	true	Respect robots.txt
includePatterns	string[]	[]	URL include globs
excludePatterns	string[]	[]	URL exclude globs
concurrency	int (1..10)	3	Parallel workers
timeoutMs	int (1000..600000)	30000	Overall timeout

Response: { jobId, status, startedAt }

Poll progress: GET /crawl/site/{jobId}?offset=0&limit=50

• Returns { jobId, status, progress: { discovered, crawled, failed }, results: [...], resultTotal, offset, limit, timing_ms }
• Status: running | completed | failed | cancelled

Cancel: DELETE /crawl/site/{jobId}

• Returns { jobId, status: "cancelled" } or 409 if already finished.

Parameter Guidance

• freshness=week for fast-moving topics; freshness=any for broad research.
• topK between 3-10 for agent workflows.
• needContent=true for lightweight body text (fast, cached).
• needCrawl=true for deep content extraction with Readability + Markdown cleaning (slower, higher quality).
• needMap=true to discover sibling pages on result domains.
• needSiteCrawl=true to launch background crawl jobs per domain (check siteCrawlJobId in results to poll).
• format=markdown (default) for LLM consumption; format=text for plain text; format=html for raw HTML.
• lang=zh-TW to force Traditional Chinese ranking boost; omit for auto-detection.
• stripLinks=true when downstream doesn't need hyperlinks in content.

Troubleshooting

1. 403 with Cloudflare challenge => check Cloudflare WAF rules for search.yuuzu.net.
2. 401 => token is missing/invalid.
3. 429 => rate limit (60 req/min) or max concurrent site crawl jobs reached.
4. 5xx => verify gateway/searxng/redis containers on oracle-my.

For detailed ops checklist: references/runbook.md

Error Handling

If the curl call fails, show the raw output to the user and stop. Common errors:

• 401 — token missing or invalid
• 403 — Cloudflare WAF block
• 429 — rate limit (60 req/min)
• 5xx — gateway/backend down

Do NOT attempt to diagnose or fix token issues. Just report the error.

Resources

scripts/

• scripts/search_request.sh — Search query helper (fallback, prefer inline curl)
• scripts/crawl_request.sh — Single URL crawl helper (fallback)
• scripts/map_request.sh — URL discovery helper (fallback)

references/

• references/runbook.md — Full API contract, deployment, diagnostics, and Cloudflare notes