Querying a Knowledge Base

After you create a knowledge base for your AI project, you'll want to do some quick queries to ensure you have the expected data.

This topic includes:

• List Columns Simulation Examples
• Search Schema Simulation Examples
• Query Data Directly Simulation Examples

You can check the setup of your Knowledge Bases by selecting the Details tab

To simulate a Knowledge Base, select the Test tab.

List Columns

Enter table names

Enter keywords to search

Enter a SQL statement to run

Database knowledge bases support two result formats: Markdown-KV and JSON. The format you selected when creating the knowledge base determines how query results are displayed in the Test tab.

When you select Markdown-KV as the Format Type, query results are returned in a structured, human-readable format with headings and key-value pairs.

List Columns example:

The result shows database and table information with column details in a structured format:

# Database "retail_demo"
## Table "customer_profiles"
### Column 1
name: customer_id
type: varchar
comment:
### Column 2
name: first_name
type: varchar
comment:

Query data directly example:

For example, running this query:

SELECT customer_id, first_name, last_name, city, state 
FROM customer_profiles 
LIMIT 1

Returns results in this format:

# Records
## Record 1
customer_id: CUST-00613
first_name: Aiden
last_name: Hernandez
city: San Diego
state: CA

Truncation warning:

When query results exceed the Max rows limit, a warning message is automatically appended to help the AI agent recognize that results are incomplete:

For example, when Max rows is set to 2 but the query attempts to retrieve more records, the following warning appears:

# Warning
[TRUNCATED] Results limited to 2 rows. To retrieve more data, add pagination 
to your SQL query using OFFSET N LIMIT M (OFFSET must come before LIMIT), 
or add WHERE conditions to narrow the results.

This warning prompts the AI agent to refine the query with pagination or additional filtering conditions.

When you select JSON as the Format Type, query results are returned in traditional JSON format, similar to previous versions of the knowledge base feature.

Next Step

Creating an Agent