DataWorks:DataWorks Agent

Step 1: Switch to Agent mode

Click the icon in the upper-right corner of Data Studio to open Copilot Chat, then switch to Agent mode in the lower-left corner.

Step 2: Select an agent

Enter / in the input box to open the agent menu and select the agent that matches your task:

In each DataWorks product module, the agent is selected automatically based on your current context.

Step 3: Add context (optional)

Enter @ in the dialog box or click @ in the lower-right corner to attach context. Adding context helps the agent understand your data model and produce more accurate results.

Step 4: Switch the LLM (optional)

By default, Copilot uses the DataWorks default model, which intelligently allocates models based on task requirements. To use a different model, click the icon at the bottom of the dialog box and select a model.

Step 5: Refine with multi-turn conversation

Enter your request in the dialog box. Ask follow-up questions or provide more details to refine the result across multiple turns until the agent produces what you need.

Data Integration Agent

Describe a data synchronization requirement in natural language (Chinese or English). The agent parses your intent and automatically generates the full task configuration — source and target data source types, table structure mappings, column filtering conditions, partitioning strategies, and scheduling parameters.

What the agent does:

• Parses your description to extract sync requirements and populate source, target, and mapping configuration automatically.

• Creates a data synchronization node that you can review and modify before saving.

Steps:

1. Enter / and select Data Integration Agent.

2. Describe your sync requirements, including source, target, table names, and sync method.

3. The agent creates a data synchronization node. Click the node to view or modify the configuration.

Example prompt:

Create an offline synchronization task to sync the ods_user_info_d table from MySQL to the ods_user_info_d table in MaxCompute.

Data Studio Agent

Build ETL workflows through natural language. The agent covers the full development cycle: requirements analysis, code generation, dependency configuration, and release.

What the agent does:

• Breaks the task into sub-steps (create nodes, generate code, configure dependencies) and shows a To-do list with real-time status.

• Generates node code that you can review, keep, or discard before it's applied.

• Compiles an execution summary when the workflow is complete.

Steps:

1. Describe your data development requirement in natural language and attach context with @ as needed.

2. The agent breaks the task into steps and executes them. Review generated code at each step.

3. Choose to keep or discard each change before the agent moves to the next step.

Example prompt:

Build a user profile analysis workflow.

Storage management: The agent creates nodes and files in project or personal directories. To set the default path, configure The default storage path for generating code files in the Copilot settings center. For details, see Personal settings. If the node type conflicts with the target directory rules (for example, creating a data integration node in a personal directory), the agent prompts for confirmation before proceeding.

Data Governance Agent

Manage data governance through natural language commands instead of manual form-based configuration. The agent analyzes your data, recommends actions, and executes governance operations automatically.

What the agent does:

• Quality rule configuration: Analyzes column types, business semantics, and table importance to recommend and configure monitoring rules — including primary key uniqueness, non-null constraints, and enumeration range checks.

• Quality issue resolution: Identifies governance issues (such as frequently accessed tables without quality rules) and performs remediation automatically.

Steps:

1. Enter / and select Data Governance Agent.

2. Describe your governance requirement. The agent analyzes the relevant tables and proposes an action plan.

3. Confirm the plan. The agent configures and executes the governance operations.

Example prompts:

Automatically generate quality rules for the core user dimension table dim_user_info.

Configure quality rules related to table row counts for all tables starting with ods_.

Find frequently accessed tables that don't have quality rules, then recommend and configure rules for them.

Help me resolve all open issues in the data quality dimension.

Data Map Agent

Search and explore metadata across your data assets using natural language — no precise keywords needed.

What the agent does:

• Natural language search: Finds tables based on your business intent rather than exact names.

• Scope-aware filtering: Understands when you specify a project or domain and narrows results accordingly.

• Deep data exploration: Answers follow-up questions about data lineage, ownership, and column definitions for a specific table.

Example prompts:

Find summary tables related to user activity.

In the adm_bi project, find tables related to business operations.

@dws_bi_metric_di What are the direct downstream dependencies of this table? Who are the owners that would be affected by changes?

Data O&M Agent

Get a comprehensive health assessment and root-cause diagnosis for task instances. The agent integrates analysis across dependency chains, resource levels, historical run trends, change impacts, log anomalies, and data quality — then outputs a structured diagnostic report.

For details, see AI-powered O&M.