All Products
Search
Document Center

E-MapReduce:EMR AI Assistant quick start

Last Updated:Aug 11, 2026

EMR AI Assistant connects directly to your cluster to analyze real-time operational data. Unlike generic large models, it provides actionable recommendations based on your cluster's actual performance and extensive big data operations expertise, eliminating the need for complex troubleshooting and manual data collection. This guide explains how to get started with EMR AI Assistant.

Permissions

  • Alibaba Cloud accounts: You can use EMR AI Assistant without additional configuration.

  • RAM users: The Alibaba Cloud account must grant the following permissions before RAM users can use EMR AI Assistant.

    • Basic session permissions (required): These permissions enable RAM users to use intelligent Q&A, view chat history, and submit feedback. Attach the following emr-agent custom policy to the RAM user.

      # This policy grants a RAM user access to EMR AI Assistant features such as intelligent chat, anomaly diagnosis, and operations consultation.
      {
        "Version": "1",
        "Statement": [
          {
            "Effect": "Allow",
            "Action": [
              "emr-agent:ChatCompletion",
              "emr-agent:CreateOrder",
              "emr-agent:DeleteChannel",
              "emr-agent:DeleteCustomReportTask",
              "emr-agent:DeleteSession",
              "emr-agent:ExecuteTaskNow",
              "emr-agent:FeedbackOnMessage",
              "emr-agent:GetChannel",
              "emr-agent:GetCustomReportTask",
              "emr-agent:GetCustomerAccessStatus",
              "emr-agent:GetDailyReport",
              "emr-agent:GetExecution",
              "emr-agent:GetPresignedUrl",
              "emr-agent:GetSubscriptionStatus",
              "emr-agent:GetTask",
              "emr-agent:GetUsageStatistics",
              "emr-agent:ListChannels",
              "emr-agent:ListCustomReportTasks",
              "emr-agent:ListDailyReports",
              "emr-agent:ListExecutions",
              "emr-agent:ListMessage",
              "emr-agent:ListRegions",
              "emr-agent:ListSession",
              "emr-agent:ListTasks",
              "emr-agent:ListUsageRecords",
              "emr-agent:ToggleChannel",
              "emr-agent:UpdateCard",
              "emr-agent:UpdateSubscriptionOverage",
              "emr-agent:UpdateUsageLimit",
              "emr-agent:UpsertChannel",
              "emr-agent:UpsertCustomReportTask",
              "emr-agent:UpsertTask"
            ],
            "Resource": "*"
          }
        ]
      }

      Alternatively, attach the AliyunEMRAssistantFullAccess system policy.

    • Operations permissions (optional): To perform actual operations through the assistant, such as viewing instance details or diagnosing issues, the RAM user also needs the corresponding instance-level permissions. For more information, see Grant permissions to a RAM user.

Step 1: Activate EMR AI Assistant

  1. Log on to the Alibaba Cloud EMR console.

  2. In the left-side navigation pane, choose EMR AI Assistant.

  3. Each user receives a free quota of 1,000,000 tokens. You can start asking questions directly on the chat page.

  4. After you use up your free quota, go to Assistant Management, click Activate, and select a subscription plan.

Step 2: Interact in the console

After activation, you can interact directly with EMR AI Assistant in the console.

  1. Enter your questions in the chat box. Examples:

    • "Run a full health check."

    • "What are the recent slow queries?"

    • "Is the number of buckets for this table appropriate?"

  2. EMR AI Assistant connects to your instance, performs an analysis, and returns diagnostic results and recommendations.

  3. You can also select specific expert skills and choose the target instance as prompted.

  4. You can ask follow-up questions in the chat for a more in-depth, multi-turn analysis.

Step 3: Configure scheduled tasks

Configure EMR AI Assistant to run analyses on a schedule and send you the results.

  1. On the EMR AI Assistant page, click Task Center > Task Configuration.

  2. Select a task type:

    • health check: Set the push time (09:00 by default) and the instance to monitor.

    • slow SQL alert: Set the slow query threshold (greater than 10s by default).

  3. Click Save to enable the task.

After a task runs, you can view the complete results of each run in Task Center > execution history.

Note

The image above shows an example of a daily health check report.

Step 4: Configure an IM channel

Integrate EMR AI Assistant with your DingTalk group so team members receive notifications directly in the group chat.

  1. On the EMR AI Assistant page, click IM configuration.

  2. Select DingTalk and follow the instructions to enter the webhook URL and signature key.

After the configuration is complete, team members will receive notifications from EMR AI Assistant in the DingTalk group.

(Optional) Step 5: Use API access

You can integrate EMR AI Assistant into your automated operations workflows by using APIs. For more information, see Call EMR AI Assistant by using APIs.

Recommendations

  1. Be specific in your questions: Include instance names, time ranges, and specific symptoms to help the assistant pinpoint the issue accurately.

  2. Use scheduled tasks: Configure daily reports and weekly health checks to proactively identify potential risks without manual review.

  3. Integrate with DingTalk groups: Connect the assistant to your team's operations group to give everyone direct access and help new members get up to speed quickly.

  4. Use with alerts: Combine the assistant's alert notifications with CloudMonitor to enable automatic diagnosis after an alert is triggered.

  5. Review execution history: Regularly review historical health checks and daily reports in the Task Center to track the health trends of your cluster.