Baidu AI Search Plugin

Overview

Baidu AI Search Plugin provides powerful search capabilities for the Dify platform, supporting both basic search and intelligent search generation modes.

Quick Start

Get API Key

1. Visit Baidu Qianfan Platform
2. Register or log in to your account
3. Create an API Key
4. Add your API Key in Dify tool configuration

Function Description

Baidu Search

Function Overview
Searches real-time information across the web based on user queries and returns summaries, URLs, and other information.

Parameter Description

Parameter	Type	Required	Description
query	string	Yes	Search keywords or phrases used to specify the content to be searched. Supports natural language queries and can contain multiple keywords. Agents should construct appropriate query statements based on user intent to ensure the accuracy and relevance of search results
resource_type_filter	array	No	Specifies the types of search resources and the number of results returned for each type. Default configuration is [{"type": "web","top_k": 10}], indicating that only web-based search results are returned, with a maximum of 10 results

Example Code

# Basic search
result = tools.baidu_web_search(query="Tourist attractions in Beijing")

# Custom resource filtering
result = tools.baidu_web_search(
    query="Tourist attractions in Beijing",
    resource_type_filter=[{"type": "web", "top_k": 5}]
)

Smart Search Generation

Function Overview
Searches real-time information across the web based on user queries and provides intelligent summarized responses.

Parameter Description

Parameter	Type	Required	Description
query	string	Yes	Search keywords or phrases used to specify the content to be searched. Supports natural language queries and can contain multiple keywords. Agents should construct appropriate query statements based on user intent to ensure the accuracy and relevance of search results
model	string	No	Specifies the LLM model for search result summarization. When empty string, it uses Baidu basic search and returns raw results; when specified as "ERNIE-3.5-8K", it uses the corresponding large model for intelligent summarization. Model documentation: Baidu Qianfan Model List
instruction	string	No	Instruction parameters used to control the style and format of search result output. Such as: result sorting method, content filtering conditions, output format, etc. When not specified, the default search result display method will be used
temperature	float	No	Sampling parameter that controls the randomness of model output, range is (0, 1]. Default value is 1e-10. This parameter only takes effect when the model parameter specifies an LLM model
top_p	float	No	Nucleus sampling parameter that controls the diversity of model output, default value is 1e-10. This parameter only takes effect when the model parameter specifies an LLM model
resource_type_filter	array	No	Specifies the types of search resources and the number of results returned for each type. Default configuration is [{"type": "web","top_k": 10}], indicating that only web-based search results are returned, with a maximum of 10 results

Model Support

• ERNIE-3.5-8K
• Other models supported by Baidu Qianfan platform

Example Code

# Basic smart search
result = tools.smart_search(query="Tourist attractions in Beijing")

# Advanced smart search
result = tools.smart_search(
    query="Tourist attractions in Beijing",
    model="ERNIE-3.5-8K",
    instruction="Summarize key points in list format, sorted by time in descending order",
    temperature=0.7,
    top_p=0.9,
    resource_type_filter=[{"type": "web", "top_k": 10}]
)

Support & Feedback

For any questions or suggestions, please contact the developer or refer to Baidu Qianfan Documentation.