Jusou Search
用聚搜GO(Jusou)给 Agent 联网搜索:一个 Key 调 8 个搜索引擎(Exa / Tavily / Serper / SerpApi / Brave / 博查 / 百度 / AnySearch),按积分计费、失败不扣分。用户要求联网搜索、查最新信息、找资料、核实事实、多引擎对比,或提到 Tavily...
Jusou Search — Web Search for AI Agents
When to use
- You need real-time / latest information (news, prices, versions, schedules, policies) or facts beyond training data
- Finding sources, similar pages, documentation and official sites
- Comparing results across multiple search engines (global + China coverage complement each other)
Setup
Set the environment variable JUSOU_API_KEY to the user's key from the console at https://jusouai.cn (keys look like jsk_...).
New users get 100 free points on sign-up (valid 7 days) — enough to try every engine. If there is no key, ask the user to sign up or redeem a card at jusouai.cn and put the key in the environment variable. Never ask the user to paste the key into the chat.
How to search
Prefer the bundled script (zero-dependency Python 3). It lives in scripts/ next to this SKILL.md; $SKILL_DIR below means the directory containing this file — locate it first, then build the path:
python3 "$SKILL_DIR/scripts/jusou_search.py" "Hangzhou weather"
python3 "$SKILL_DIR/scripts/jusou_search.py" "GPT-5 review" --engine exa --num 5
python3 "$SKILL_DIR/scripts/jusou_search.py" "新能源汽车销量" --engines baidu,bocha,exa --merge
python3 "$SKILL_DIR/scripts/jusou_search.py" "earnings report" --engine tavily --depth advanced
You can also call the unified API directly: POST https://jusouai.cn/v1/search (Bearer auth), with the same parameters as the script.
Examples
# User: "What's the weather in Hangzhou today?"
$ python3 "$SKILL_DIR/scripts/jusou_search.py" "杭州天气" --engine bocha --num 3
1. 2026年11月19日杭州天气预报
https://bocha.cn/share/7c867faf-48ae-49a6-9243-9cee3166cc61
2026年11月19日杭州天气预报: 11/19 (周四): 天气:晴,温度:14/6°C,风向风力:西南风<3级 …
-- 本次:博查 -1 积分,剩余 129 (charged 1 point, balance 129 — printed on stderr)
# User: "Verify the latest official height of Mount Everest for me."
$ python3 "$SKILL_DIR/scripts/jusou_search.py" "珠穆朗玛峰最新高程" --engines exa,bocha --merge --num 3
1. 习近平同尼泊尔总统班达里互致信函 共同宣布珠穆朗玛峰高程_中国政府网 (gov.cn)
3. 8848.86米!珠穆朗玛峰最新高程公布_央视网
https://news.cctv.com/2020/12/09/ARTIf96rCeRLlKsVWjJ9i9G1201209.shtml
…中尼两国8日联合对外宣布,珠穆朗玛峰的最新高程为8848.86米…
-- 本次:Exa -1 积分、博查 -1 积分,共 -2 积分,剩余 127 (charged 2 points, balance 127)
Choosing an engine
| Scenario | Engine |
|---|---|
| English / semantic research, finding similar pages | exa |
| Structured results to feed a model | tavily |
| Google web / news / scholar / shopping | serper |
| Google results in the official SerpApi shape | serpapi |
| Independent index | brave |
| Chinese web pages (China-site coverage) | bocha, baidu (advanced depth = Baidu AI Search, generates an answer directly) |
| "Find me a WeChat Official Account article about X" | bocha, baidu — append site:mp.weixin.qq.com to the query |
| "See what Zhihu says about X" | bocha, baidu — append site:zhihu.com to the query |
| "How do I fix this error?" (Stack Overflow) | serper, exa — append site:stackoverflow.com to the query |
| "Check the official docs for X" | exa, serper — append site: + the official domain to the query |
| Default engine, multi-source aggregation fallback | anysearch |
Omit engine to use the default engine (anysearch). To compare, pass --engines a,b,c (each engine is charged separately). Fact-checking: for a claim that matters, don't trust a single source — run 2–3 engines at once (--engines exa,bocha,serper --merge) and compare: if independent sources agree, it's likely solid; if they conflict, dig deeper before answering. This is just the server-side multi-engine search used deliberately, not extra logic in this skill.
Pricing (keep it in mind, don't burn points)
- Most engines cost 1 point per basic search (serpapi 4, brave 2). Higher tiers: exa advanced 2 / deep 4, tavily advanced 2, baidu advanced (AI Search) 3. serper, serpapi, anysearch and bocha only have a basic tier — passing a higher tier returns 400.
- Failed calls are never charged. Every response reports the points charged and the remaining balance; when the balance is insufficient the API says so explicitly — stop and tell the user to top up instead of retrying.
- Exa can return up to 100 results; beyond 10 results, each additional block of 10 costs extra points. Test with a small sample before large crawls.
Troubleshooting
- 401: key invalid or not set — ask the user to check the environment variable.
- 402: insufficient points — suggest redeeming a card to top up (the same key accumulates automatically).
- 429: engine busy or rate-limited — retry once with a different engine, or wait and retry.
- 400: bad parameters (e.g. empty query) — fix the parameters; don't retry as-is.
聚搜GO 联网搜索
什么时候用
- 需要实时/最新信息(新闻、价格、版本、赛程、政策)或训练数据之外的事实
- 找资料、找同类网页、查文档与官网
- 想让多家搜索引擎对比着搜(国内外覆盖互补)
准备
环境变量 JUSOU_API_KEY = 用户在 https://jusouai.cn 控制台的 Key(形如 jsk_...)。
新用户注册赠送 100 积分(7 天有效),够把每个引擎都试一遍。没有 Key 时,提示用户去 jusouai.cn 注册/兑换后把 Key 配到环境变量,不要让用户把 Key 贴在聊天里。
怎么搜
优先用随附脚本(零依赖 Python 3)。脚本在本 SKILL.md 同目录的 scripts/ 下,下例 $SKILL_DIR 指本文件所在目录,先定位它再拼路径:
python3 "$SKILL_DIR/scripts/jusou_search.py" "杭州天气"
python3 "$SKILL_DIR/scripts/jusou_search.py" "GPT-5 review" --engine exa --num 5
python3 "$SKILL_DIR/scripts/jusou_search.py" "新能源汽车销量" --engines baidu,bocha,exa --merge
python3 "$SKILL_DIR/scripts/jusou_search.py" "财报" --engine tavily --depth advanced
也可以直接调统一接口 POST https://jusouai.cn/v1/search(Bearer 鉴权),参数同脚本。
示例
# 用户:「杭州今天天气怎么样?」
$ python3 "$SKILL_DIR/scripts/jusou_search.py" "杭州天气" --engine bocha --num 3
1. 2026年11月19日杭州天气预报
https://bocha.cn/share/7c867faf-48ae-49a6-9243-9cee3166cc61
2026年11月19日杭州天气预报: 11/19 (周四): 天气:晴,温度:14/6°C,风向风力:西南风<3级 …
-- 本次:博查 -1 积分,剩余 129
# 用户:「珠穆朗玛峰现在到底多高?帮我核实一下。」
$ python3 "$SKILL_DIR/scripts/jusou_search.py" "珠穆朗玛峰最新高程" --engines exa,bocha --merge --num 3
1. 习近平同尼泊尔总统班达里互致信函 共同宣布珠穆朗玛峰高程_中国政府网 (gov.cn)
3. 8848.86米!珠穆朗玛峰最新高程公布_央视网
https://news.cctv.com/2020/12/09/ARTIf96rCeRLlKsVWjJ9i9G1201209.shtml
…中尼两国8日联合对外宣布,珠穆朗玛峰的最新高程为8848.86米…
-- 本次:Exa -1 积分、博查 -1 积分,共 -2 积分,剩余 127
选引擎
| 场景 | 引擎 |
|---|---|
| 英文/语义找资料、找相似网页 | exa |
| 给模型喂结构化结果 | tavily |
| Google 网页/新闻/学术/购物 | serper |
| 与 SerpApi 官方同形的 Google 结果 | serpapi |
| 独立索引 | brave |
| 中文网页(国内站覆盖) | bocha、baidu(advanced 深度 = 百度 AI 搜索,直接生成答案) |
| 「找一篇讲 X 的公众号文章」 | bocha、baidu,查询词后加 site:mp.weixin.qq.com |
| 「看看知乎怎么说 X」 | bocha、baidu,查询词后加 site:zhihu.com |
| 「这个报错怎么修」(Stack Overflow) | serper、exa,查询词后加 site:stackoverflow.com |
| 「查一下 X 的官方文档」 | exa、serper,查询词后加 site: + 官网域名 |
| 默认引擎、多源聚合兜底 | anysearch |
不指定 engine 用默认引擎 anysearch;要对比就 --engines a,b,c(各自扣分)。事实核查: 重要的说法别只信一家——用 --engines exa,bocha,serper --merge 一次搜 2–3 家对照:独立来源说法一致,可信度就高;互相打架就先深挖再回答。这只是把服务端现成的多引擎搜索用在核查上,Skill 里不另做逻辑。
计费(心里有数,别乱烧)
- basic 档多为 1 积分/次(serpapi 4、brave 2);高级档:exa advanced 2 / deep 4、tavily advanced 2、baidu advanced(AI 搜索)3。serper、serpapi、anysearch、bocha 只有 basic 档,传高级档会报 400。
- 失败不扣分;每次返回都带本次扣分与剩余积分,余额不足会明确报错——这时停下来告诉用户去续费,不要反复重试。
- Exa 最多可返回 100 条结果,超过 10 条后每多 10 条加收积分,大批量抓取前先小样试。
出错怎么办
- 401:Key 无效或没配,让用户检查环境变量。
- 402:积分不足,提示兑换卡密续费(同一 Key 自动叠加)。
- 429:引擎繁忙或限流,换一家引擎重试一次,或稍后再试。
- 400:参数问题(如 query 为空),修正参数,不要原样重试。
- Downloads
- 162All time
- Security audit
- Review
- Last updated
- 3d ago
- Current version
- v1.0.3
- License
- MIT-0
