What Is Data-Agent

Data-Agent is a data intelligence suite in PolarDB for AI that covers the full workflow from querying data to generating insights.

Features

Ask data questions in natural language. The system uses natural language to SQL (NL2SQL) to generate reusable query statements.
After you run a query and retrieve results, you can summarize them or generate charts.
1. Use the natural language to chart (NL2Chart) model. Data-Agent selects the best visualization for your data and renders an interactive chart in real time.
2. Use the one-click intelligent summary (NL2SQL_SUMMARY) model. Data-Agent synthesizes data, extracts key insights, and provides business recommendations.

PolarDB for AI provides a Data-Agent best practice example that walks through these steps and helps you understand Data-Agent in depth. By combining these three capabilities, Data-Agent delivers closed-loop analysis—from data queries to business insights—through conversation.

Billing

Data-Agent offers two billing modes: Add an AI node and Serverless. Choose the mode that fits your workload and budget.

Note

The Serverless billing mode is in a canary release. To use it or for questions, submit a ticket.

Add an AI node: Add a dedicated AI node to your PolarDB for MySQL cluster. You are billed based on the AI node specifications you select. Ideal for stable, high-frequency workloads where dedicated resources ensure consistent performance.

Serverless: This mode does not require a dedicated AI node, making PolarDB for AI more accessible. Enabling is free. You are billed based on the actual number of SQL calls. Ideal for fluctuating workloads, infrequent use, or first-time users.

Specifications	Price (CNY/call)	Concurrency limit (QPS)
`small`	0.075	1
`medium`	0.15	5
`large-t1`	0.2	10
`large-t2`	0.3	20
`xlarge-t1`	0.45	50

Note

QPS is the maximum number of requests processed per second. If the rate exceeds the concurrency limit of your specification, the system returns an error. Implement rate limiting or backoff retry in your client application.

A single Data-Agent call is one question-and-answer interaction using an SQL statement. For example:

/polar4ai/SELECT * FROM PREDICT (MODEL _polar4ai_nl2sql, SELECT 'Find the two students with the most absences. Sort the results in descending order by the number of absences and show the student names and their absence counts.') WITH (basic_index_name='schema_index');