feat(search_toolkit): Add Alibaba Tongxiao Search API Support #2127
Conversation
There was a problem hiding this comment.
thanks @RonaldJEN for the contribution! there's some conflicts, could you resolve this?
添加新的 search_ali 方法，调用阿里巴巴通晓搜索 API，支持以下特性： - 支持时间范围和行业过滤 - 支持分页查询 - 自动提取摘要信息 (优先使用 summary，回退到 mainText) - 可选返回网页正文和 Markdown 格式内容 - 支持搜索结果重排序优化 - 返回结构与其他搜索方法保持一致 - fix(search_toolkit): resolve linting issues in search_ali method
添加新的 search_ali 方法，调用阿里巴巴通晓搜索 API
Add example for search_ali method usage
将 search_ali_response 示例输出字符串中的全角冒号 ： 修改为半角冒号 :
Fix line length (E501)
Adds a trailing comma to the parameters of the search_ali function
|
@Wendong-Fan Thanks for pointing out the conflicts! I've resolved them now. |
There was a problem hiding this comment.
thanks @RonaldJEN 's contribution, left some comments below, could we also add unit test code?
| effective for Chinese language queries. | ||
|
|
||
| Args: | ||
| query (str): The search query string (length >= 1 and <= 100). |
There was a problem hiding this comment.
add validation for length limit?
| timeRange (str): Time frame filter for search results. Default | ||
| is "NoLimit". Options include: | ||
| - 'OneDay': Past day. | ||
| - 'OneWeek': Past week. | ||
| - 'OneMonth': Past month. | ||
| - 'OneYear': Past year. | ||
| - 'NoLimit': No time limit (default). |
There was a problem hiding this comment.
use literal instead of str here? for the variable naming, following current camel style use time_range instead of timeRange
| industry (Optional[str]): Industry-specific search filter. When | ||
| specified, only returns results from sites in the specified | ||
| industries. Multiple industries can be comma-separated. | ||
| Options include: | ||
| - 'finance': Financial industry. | ||
| - 'law': Legal industry. | ||
| - 'medical': Medical industry. | ||
| - 'internet': Internet (curated). | ||
| - 'tax': Tax industry. | ||
| - 'news_province': Provincial news. | ||
| - 'news_center': Central news. |
There was a problem hiding this comment.
use literal instead of str here?
| - 'tax': Tax industry. | ||
| - 'news_province': Provincial news. | ||
| - 'news_center': Central news. | ||
| page (int): Page number for results pagination. Default is 1. |
There was a problem hiding this comment.
for default value, use format (default: :obj:1), same as others
| Dict[str, Any]: A dictionary containing search results or an error | ||
| message. The structure includes: | ||
| - 'requestId': A unique identifier for the request. | ||
| - 'results': A list of dictionaries, each representing a | ||
| search result with the following keys: | ||
| - 'result_id': The index of the result. | ||
| - 'title': The title of the webpage. | ||
| - 'snippet': A dynamic summary of relevant content matching | ||
| the query keywords. | ||
| - 'mainText': The main content of the webpage (if | ||
| returnMainText is True). | ||
| - 'markdownText': Markdown formatted content (if | ||
| returnMarkdownText is True). | ||
| - 'hostname': The name of the website. | ||
| - 'url': The URL of the webpage. | ||
| - 'publishTime': Publication timestamp in milliseconds. | ||
| - 'score': Relevance score. | ||
| - 'searchInformation': Additional metadata about the search | ||
| operation. | ||
| - or 'error': An error message if something went wrong. |
There was a problem hiding this comment.
could we simpify the returns to make it more tidy? since the toolkit would be used by agent, it would be more LLM friendly to make it simple
| @@ -394,3 +394,20 @@ class PersonInfo(BaseModel): | |||
| on improving efficiency, fault tolerance, and minimizing resource overheads. | |||
| =============================================================================== | |||
| """ | |||
|
|
|||
| search_ali_response = SearchToolkit().search_ali( | |||
| query="阿里巴巴2025年的芯片投入", | |||
There was a problem hiding this comment.
use english query as example would be better
|
@RonaldJEN you can fix all things mentioned and its gtg |
Add Alibaba Tongxiao Search API Support
Feature Description
This PR introduces the
search_alimethod to theSearchToolkitclass, integrating the Alibaba Tongxiao Search API. Tongxiao Search is a powerful real-time search API providing structured data from various search engines and knowledge bases, with specific optimizations for Chinese content search.API Reference: Standard Search API - GenericSearch
Implementation Details
Retrieves API credentials via the
TONGXIAO_API_KEYenvironment variable.Supports the following search parameters:
timeRange: Time frame filter (OneDay/OneWeek/OneMonth/OneYear/NoLimit)industry: Industry filter (finance, law, medical, etc.)page: Result paginationreturnMainText: Whether to return webpage main textreturnMarkdownText: Whether to return Markdown formatted contentenableRerank: Whether to enable result reranking (can reduce response time)Return Structure Formatting:
summaryfield; usesmainTextas a fallback summary ifsummaryis unavailable.search_bing).Documentation
The method includes a detailed docstring covering:
This PR enhances CAMEL's search capabilities, particularly for Chinese content, offering users more diverse search options.