Search Testing
Before officially launching the LLM, use the search test preview feature to verify the relevance and content of the data retrieved by the LLM
Why Do You Need Search Testing?
Ensure Response Accuracy
The quality of AI assistant responses directly affects customer satisfaction
Incorrect or irrelevant answers damage brand trust
Avoid Information Gaps
Prevent important product details or safety instructions from being overlooked
Optimize the efficiency of knowledge base usage
Improve Q&A Consistency
Ensure the same question receives a consistent answer at different times
Avoid response instability caused by knowledge base updates
Through search testing and segment preview features, you can instantly review the quality of data segments used by the AI assistant to ensure your Q&A standards.
How to Get Started?
Go to the Knowledge Base page, select the knowledge base you want to view, and enter its settings
Click on Search Test to enter the page
Type the content you want to search in the dialog box: "What should I pay attention to when choosing a campsite?", then click the search button. The AI assistant will search for the most relevant documents based on your input.

View search results
The right side will display the segments that the AI assistant retrieves when answering. As shown in the figure, there are 12 segments (to configure the number of retrieved segments, refer to: How to Create a Knowledge Base: Basic Settings)

The right side shows the retrieved content used to answer this question. Each segment displays:
Word count: The number of words covered by the segment
Hit count: The number of times the AI assistant referenced this segment when answering (only official Q&A sessions are counted; test Q&A sessions are not included in hit count calculations)
A higher hit count indicates that users ask more questions related to this segment.
Search test records past test Q&A sessions. You can quickly click on history records to retest, ensuring that the same topics maintain or improve data reference quality.

Check Knowledge Segment Quality
Through search testing, you can determine which segments the AI assistant uses during reasoning, and then check whether the content of these segments needs optimization, such as:

If you find many blank segments in the search results, and these blank segments all belong to the file "Beginner Camping.pdf", you can then check whether the content of "Beginner Camping.pdf" is correct and update it immediately before going live.
Search Test by Metadata
What Is Metadata Search?
The Metadata search feature allows you to perform precise queries based on document attribute information, rather than being limited to the document content itself. This feature is particularly suitable for:
Document classification queries: Search by document type, tags, categories, and other information
Attribute filtering: Filter by document creation time, author, source, and other attributes
Precise targeting: Quickly find document collections that meet specific criteria
How to Use Metadata Search?
On the search test page, you can:
Select search mode: Switch to "Metadata Search" mode
Specify search criteria: Enter the metadata field and value you want to search
View search results: The system will display document segments that match the metadata criteria
Search examples:
Search for documents in a specific category:
category:Product DescriptionSearch for documents within a specific time range:
date:2025-01Search for documents by a specific author:
author:Customer Service Department
Advantages of Metadata Search
1. More Precise Results
Query directly against document attributes, reducing irrelevant results
Suitable for scenarios that require precise filtering
2. Faster Targeting
No need for full-text search; quickly locate target documents
Especially suitable for knowledge bases with large volumes of documents
3. Combined with Content Search
Can be used in combination with regular content search
First narrow down the scope with metadata, then pinpoint with content search
Practical Use Cases
Scenario 1: Product Document Management
Search for all documents in a specific product line
Example:
product_line:Camping Equipment
Scenario 2: Time Range Query
Find recently updated documents
Example:
updated_date:2025-11
Scenario 3: Document Type Filtering
View only specific types of documents (e.g., operation manuals, FAQs, etc.)
Example:
doc_type:FAQ
Search Testing Helps You
Test Important Questions Regularly
Test core business-related questions frequently
Retest after knowledge base updates
Record and compare improvements in test results
Identify Content Gaps
Discover knowledge gaps: Test common questions and find queries with no matching results
Example: Testing "tent waterproof rating explanation" reveals no relevant segments → waterproof specification documents need to be added
Evaluate Content Completeness
Check answer depth: Confirm whether the AI answer covers all aspects of the question
Example: Query "camping safety precautions" → Check whether it covers weather, wildlife, fire safety, and other aspects
Question Testing Tips
You can:
Ask the same question in different ways
Test complex scenarios and multi-condition queries
Verify the retrieval effectiveness of technical terms and product names
Use these testing tips to comprehensively evaluate the accuracy and completeness of responses.
Last updated
Was this helpful?
