Valyu Search
Search across the world's knowledge.
When to Use
Trigger this skill when the user asks for:
- - "search the web", "web search", "look up", "find online", "find papers on..."
- "current news about...", "latest updates on..."
- "research [topic]", "what's happening with...", "deep research on..."
- "extract content from [URL]", "scrape this page", "get the text from..."
- "answer this with sources", "what does the research say about..."
- Fact-checking with citations needed
- Academic, medical, financial, or patent research
- Structured data extraction from web pages
Prerequisites
- - Get an API key at valyu.ai
- Set
VALYU_API_KEY in the Gateway environment (recommended) or in ~/.openclaw/.env.
Commands
Run a search across the web:
CODEBLOCK0
Search across news, academic papers, financial data, patents and more
CODEBLOCK1
Parameters
| Parameter | Type | Description |
|---|
| INLINECODE2 | string | Search query (required) |
| INLINECODE3 |
string |
"web",
"proprietary",
"news", or
"all" (default:
"all") |
|
maxNumResults | number | 1-20 (default: 10) |
|
includedSources | string[] | Limit to specific sources (e.g.,
["valyu/valyu-arxiv"]) |
|
excludedSources | string[] | Exclude specific sources |
|
startDate | string | Filter from date (YYYY-MM-DD) |
|
endDate | string | Filter to date (YYYY-MM-DD) |
|
countryCode | string | ISO 3166-1 alpha-2 (e.g.,
"US",
"GB") |
|
responseLength | string |
"short",
"medium",
"large",
"max" |
|
fastMode | boolean | Reduced latency mode |
|
category | string | Natural language category (e.g.,
"academic research") |
|
relevanceThreshold | number | 0.0-1.0 (default: 0.5) |
Available Proprietary Sources
| Source | Description |
|---|
| INLINECODE27 | Academic papers from arXiv |
| INLINECODE28 |
Medical and life science literature |
|
valyu/valyu-stocks | Global stock market data |
Additional sources: BioRxiv, MedRxiv, clinical trials, FDA drug labels, WHO health data, SEC filings, USPTO patents, Wikipedia, UK Parliament, UK National Rail, maritime vessel tracking, and more.
2. Contents API
Extract clean, structured content from any URL. Converts web pages to markdown or structured data.
Usage
CODEBLOCK2
Parameters
| Parameter | Type | Description |
|---|
| INLINECODE30 | string[] | Array of URLs to extract (required) |
| INLINECODE31 |
string | Output length:
"short",
"medium",
"large",
"max" |
|
extractEffort | string |
"auto",
"lightweight",
"moderate",
"heavy" |
|
jsonSchema | object | JSON Schema for structured extraction |
3. Answer API
Get AI-generated answers grounded in real-time search results with citations.
Usage
CODEBLOCK3
Parameters
| Parameter | Type | Description |
|---|
| INLINECODE42 | string | Question or query (required) |
| INLINECODE43 |
string | Search scope:
"web",
"proprietary",
"news",
"all" |
|
outputSchema | object | JSON Schema for structured responses |
4. DeepResearch API
Launch async, multi-step research tasks that produce detailed reports with citations.
Modes
| Mode | Duration | Use Case |
|---|
| INLINECODE49 | ~5 min | Quick answers, lightweight research |
| INLINECODE50 |
~10-20 min | Balanced research with deeper insights |
|
heavy | ~90 min | In-depth, complex analysis |
Usage
CODEBLOCK4
Parameters
| Parameter | Type | Description |
|---|
| INLINECODE52 | string | Research query (required) |
| INLINECODE53 |
string |
"fast",
"standard",
"heavy" |
|
outputFormats | array |
["markdown"],
["pdf"], or JSON Schema object |
|
files | array | File attachments (base64 encoded, max 10) |
|
urls | string[] | URLs to analyze (max 10) |
|
webhookUrl | string | HTTPS URL for completion notification |
Choosing the Right API
| Need | API |
|---|
| Quick facts, current events, citations | Search |
| Read/parse a specific URL |
Contents |
| AI-synthesized answer with sources |
Answer |
| In-depth analysis or report |
DeepResearch |
References
Valyu 搜索
搜索全球知识。
使用时机
当用户提出以下请求时触发此技能:
- - 搜索网络、网络搜索、查找、在线搜索、查找关于...的论文
- 关于...的最新新闻、关于...的最新动态
- 研究[主题]、...发生了什么、深度研究...
- 从[URL]提取内容、抓取此页面、获取...的文本
- 用来源回答这个问题、研究对此有何说法
- 需要引用来源的事实核查
- 学术、医学、金融或专利研究
- 从网页中提取结构化数据
前提条件
- - 在 valyu.ai 获取 API 密钥
- 在 Gateway 环境(推荐)或 ~/.openclaw/.env 中设置 VALYUAPI_KEY
命令
在网络中执行搜索:
bash
node {baseDir}/scripts/valyu.mjs search web <查询词>
搜索新闻、学术论文、金融数据、专利等
bash
node {baseDir}/scripts/valyu.mjs search news <查询词>
参数
| 参数 | 类型 | 描述 |
|---|
| query | string | 搜索查询词(必填) |
| searchType |
string | web、proprietary、news 或 all(默认:all) |
| maxNumResults | number | 1-20(默认:10) |
| includedSources | string[] | 限制特定来源(例如:[valyu/valyu-arxiv]) |
| excludedSources | string[] | 排除特定来源 |
| startDate | string | 按日期筛选起始(YYYY-MM-DD) |
| endDate | string | 按日期筛选截止(YYYY-MM-DD) |
| countryCode | string | ISO 3166-1 alpha-2(例如:US、GB) |
| responseLength | string | short、medium、large、max |
| fastMode | boolean | 低延迟模式 |
| category | string | 自然语言类别(例如:academic research) |
| relevanceThreshold | number | 0.0-1.0(默认:0.5) |
可用的专有来源
| 来源 | 描述 |
|---|
| valyu/valyu-arxiv | arXiv 的学术论文 |
| valyu/valyu-pubmed |
医学和生命科学文献 |
| valyu/valyu-stocks | 全球股票市场数据 |
其他来源:BioRxiv、MedRxiv、临床试验、FDA 药品标签、WHO 健康数据、SEC 文件、USPTO 专利、维基百科、英国议会、英国国家铁路、海事船舶追踪等。
2. 内容 API
从任何 URL 提取干净、结构化的内容。将网页转换为 Markdown 或结构化数据。
使用方法
bash
node {baseDir}/scripts/valyu.mjs contents https://example.com --summary
参数
| 参数 | 类型 | 描述 |
|---|
| urls | string[] | 要提取的 URL 数组(必填) |
| responseLength |
string | 输出长度:short、medium、large、max |
| extractEffort | string | auto、lightweight、moderate、heavy |
| jsonSchema | object | 用于结构化提取的 JSON Schema |
3. 答案 API
获取基于实时搜索结果并附有引用来源的 AI 生成答案。
使用方法
bash
node {baseDir}/scripts/valyu.mjs answer 什么是量子计算? --fast
参数
| 参数 | 类型 | 描述 |
|---|
| query | string | 问题或查询词(必填) |
| searchType |
string | 搜索范围:web、proprietary、news、all |
| outputSchema | object | 用于结构化响应的 JSON Schema |
4. 深度研究 API
启动异步、多步骤的研究任务,生成带有引用来源的详细报告。
模式
| 模式 | 时长 | 使用场景 |
|---|
| fast | 约 5 分钟 | 快速答案、轻量级研究 |
| standard |
约 10-20 分钟 | 平衡研究,提供更深入的见解 |
| heavy | 约 90 分钟 | 深入、复杂的分析 |
使用方法
bash
node {baseDir}/scripts/valyu.mjs deepresearch create AI 市场趋势 --mode heavy --pdf
参数
| 参数 | 类型 | 描述 |
|---|
| query | string | 研究查询词(必填) |
| mode |
string | fast、standard、heavy |
| outputFormats | array | [markdown]、[pdf] 或 JSON Schema 对象 |
| files | array | 文件附件(Base64 编码,最多 10 个) |
| urls | string[] | 要分析的 URL(最多 10 个) |
| webhookUrl | string | 用于完成通知的 HTTPS URL |
选择合适的 API
| 需求 | API |
|---|
| 快速事实、当前事件、引用来源 | 搜索 |
| 读取/解析特定 URL |
内容 |
| 带来源的 AI 合成答案 |
答案 |
| 深度分析或报告 |
深度研究 |
参考