tavily-search · diff
git:20260311.7d39363 to git:20260801.3cb01ce
43 added, 38 removed. Audit A to A.
---
name: tavily-search
description: >
- 使用 Tavily Search API 进行联网检索。当用户需要查找最新资讯、官方文档、新闻动态、教程入口、政策变化或任何实时事实时，优先触发此技能。
- 出现“查一下”、“最新”、“现在”、“官网”、“文献/论文/课程入口”、“有没有更好的方案”等信息检索请求时，务必加载此技能。
- 支持搜索网页、获取摘要与原始链接。本技能应优先于纯模型猜测，为用户提供可核验的信息来源。
+ Use the Tavily Search API to perform web searches. When users need to find the latest information, official documentation, news updates, tutorial links, policy changes, or any real-time facts, prioritize triggering this skill.
+ Be sure to load this skill whenever information retrieval requests appear, such as "look this up," "latest," "now," "official website," "literature/paper/course entry point," or "is there a better solution?"
+ It supports searching web pages and retrieving summaries and original links. This skill should take precedence over pure model guessing to provide users with verifiable sources of information.
---
- # Tavily Search 联网检索
+ # Tavily Search Web Search
- 本技能提供从互联网实时检索信息的能力，通过 Tavily API 获取高质量的网页结果、摘要及链接。
+ This skill provides the ability to search the internet for information in real time, using the Tavily API to obtain high-quality web results, summaries, and links.
- ## 适用场景
+ ## Use Cases
- - **实时资讯**：查找最新消息、版本更新、公司公告、社会活动、科技新闻。
- - **资源定位**：检索官方文档、开发手册、论文原文入口、在线课程页面、工具官网。
- - **事实核验**：回答包含“最近/当前/最新/官网/今天”等时间敏感词的问题。
- - **深度背景**：在回答复杂问题前，先通过搜索补充行业背景或多方观点。
+ - **Real-Time Information**: Find the latest news, version updates, company announcements, social events, and technology news.
+ - **Resource Location**: Search for official documentation, developer manuals, entry points to original papers, online course pages, and official tool websites.
+ - **Fact-Checking**: Answer questions containing time-sensitive terms such as "recent," "current," "latest," "official website," or "today."
+ - **In-Depth Background**: Before answering complex questions, use search to gather industry background or multiple perspectives.
- ## 依赖与配置
+ ## Dependencies and Configuration
- - **依赖**：`python3`, Python 包 `tavily-python`。
- - **环境变量**：必须设置 `TAVILY_API_KEY`。
+ - **Dependencies**: `python3`, Python package `tavily-python`.
+ - **Environment Variable**: `TAVILY_API_KEY` must be set.
- ### 环境变量配置
- 若未配置环境变量，可通过以下链接快速设置：
- - [设置 TAVILY_API_KEY](minis://settings/environments?create_key=TAVILY_API_KEY&create_value=)
+ ### Environment Variable Configuration
- 若环境缺失依赖，可执行：
+ If the environment variable is not configured, you can quickly set it using the following link:
+
+ - [Set TAVILY_API_KEY](minis://settings/environments?create_key=TAVILY_API_KEY&create_value=)
+
+ If a dependency is missing from the environment, run:
+
```bash
python3 -m pip install tavily-python
```
- ## 工具与脚本
+ ## Tools and Scripts
- 本技能包含一个核心 Python 脚本，支持文本和 JSON 格式输出。
+ This skill includes one core Python script that supports text and JSON output.
- ### 脚本位置
+ ### Script Location
+
`/var/minis/skills/tavily-search/scripts/search.py`
- ### 命令示例
+ ### Command Examples
```bash
- # 基础检索（默认文本输出，含标题/URL/摘要）
- python3 scripts/search.py "Python 3.13 新特性"
+ # Basic search (default text output, including title/URL/summary)
+ python3 scripts/search.py "Python 3.13 New Features"
- # 指定返回条数
- python3 scripts/search.py "上海 本周活动" --max 8
+ # Specify the number of results to return
+ python3 scripts/search.py "Shanghai Events This Week" --max 8
- # 获取 JSON 格式（适合模型进一步处理）
- python3 scripts/search.py "Rust Web 框架" --format json
+ # Get JSON format (suitable for further processing by the model)
+ python3 scripts/search.py "Rust Web Framework" --format json
- # 包含 API 生成的参考答案（若有）
- python3 scripts/search.py "什么是 RAG 技术" --include-answer
+ # Include the reference answer generated by the API (if any)
+ python3 scripts/search.py "What is RAG Technology" --include-answer
```
- ### 参数说明
+ ### Parameter Description
- - `query` (位置参数): 搜索关键词。
- - `--max`: 返回结果数量，默认 5。
- - `--format`: `text` (默认) 或 `json`。
- - `--include-answer`: 布尔开关，在文本模式下显示 API 聚合生成的 `answer` 字段。
+ - `query` (positional argument): Search keyword.
+ - `--max`: Number of results to return. Default is 5.
+ - `--format`: `text` (default) or `json`.
+ - `--include-answer`: Boolean switch. In text mode, displays the API-aggregated `answer` field.
- ## 工作流建议
+ ## Workflow Recommendations
- 1. **先搜索后回答**：对于非通用常识（尤其是近期事件），先调用 `search.py` 获取上下文，再整合回答。
- 2. **带来源引用**：在回答结尾或段落中，务必附上搜索结果中的 URL 作为来源，增强可信度。
- 3. **结构化处理**：若需处理大量信息，优先使用 `--format json` 并在内部解析结果。
+ 1. **Search First, Then Answer**: For topics that are not general common knowledge, especially recent events, first call `search.py` to obtain context, then synthesize the answer.
+ 2. **Include Source Citations**: At the end of the answer or within paragraphs, be sure to include URLs from the search results as sources to improve credibility.
+ 3. **Structured Processing**: If you need to process a large amount of information, prioritize using `--format json` and parse the results internally.
minis_url: minis://skills/tavily-search/SKILL.md
+