Big Data 15 min read

MaxAgent Tutorial: Activate, Ask, and Verify Your First Query

This step-by-step guide walks new users through enabling MaxAgent standard edition, confirming account permissions, asking a structured first question about MaxCompute external tables, validating the AI response against official documentation, reviewing conversation history, and understanding the shared Credits quota model for sustainable usage.

Alibaba Cloud Big Data AI Platform
Alibaba Cloud Big Data AI Platform
Alibaba Cloud Big Data AI Platform
MaxAgent Tutorial: Activate, Ask, and Verify Your First Query

Introduction

MaxAgent is the AI intelligent assistant on the MaxCompute platform. It lets users ask product-knowledge questions in natural language or combine specific job and resource information for diagnostic analysis. This first lesson targets data developers, operators, platform administrators, and technical leads who are new to MaxAgent. The example only performs product-knowledge Q&A — no tables are created, no SQL is executed, and no configurations are changed, so no business data preparation is required.

Why Use MaxAgent

Daily MaxCompute work spans product documentation, job logs, resource monitoring, cost, and permission portals. Many tasks are not complex, but time is spent finding entry points, supplying context, and cross-checking. MaxAgent places these scenarios into natural-language dialogue so users describe the goal first, then gradually fill in project, job, or time-range details.

Lower the Product Learning Barrier

When unfamiliar with a concept or feature, users can ask in everyday language without mastering every command or page location. MaxAgent answers questions about MaxCompute features, use cases, and best practices, helping new users find the right direction faster.

Take Business Questions Straight into Data Analysis

DataAgent is MaxAgent's data-analysis role, exposed in the console as "Query Data." Within the current account's accessible data scope, it can locate projects, schemas, tables, and partitions; run bounded queries; analyze metrics; diagnose anomalies; check data quality; and return results as tables, charts, dashboards, or analysis reports. Users do not need to know full table names or write SQL fluently, but must still define time ranges, metric definitions, analysis goals, and verify query evidence and conclusions.

Shorten Diagnosis and Analysis Chains

For SQL failures, slow runs, or data skew, users can initiate diagnosis with concrete job information. Platform administrators can also analyze cost, Quota, storage, permissions, and metadata.

From Analysis Results to Management Closed Loop

MaxAgent not only analyzes problems but already supports project management, Quota management, and permission management. Users with appropriate permissions can trigger certain platform management tasks in natural language, reducing context-switching across pages and commands.

Keep Periodic Work Running Continuously

For repetitive tasks like abnormal-job patrol, scheduled tasks can be configured and results delivered via DingTalk bots. This shifts work from ad-hoc questioning to periodic execution, with follow-up drill-down based on reports.

Permission Inheritance, Sensitive Operation Control, Auditable Process

DataAgent reads only data the current account can access; any table or data operation requires both account permission verification and a second confirmation.

Project, Quota, and permission management actions can only be initiated by users with corresponding rights and must pass the actual page's confirmation or approval; before execution, check target objects, change content, and impact scope.

Historical conversations are retained and reviewable; management operations, scheduled-task execution records, and Credits details provide additional audit trails. Important conclusions and change results still require manual verification.

MaxAgent's value lies not in making all decisions for users, but in connecting knowledge query, DataAgent analysis, job diagnosis, controlled management, and periodic patrol. For example, periodic patrol can surface job anomalies, then drill down with diagnostic capabilities; for platform issues, controlled management actions can be triggered and execution results verified. This reduces repetitive searching, organizing, and cross-page operations, letting users focus on judgment, confirmation, and handling. Different capabilities have different rollout scopes; this article starts with product-knowledge Q&A suitable for new users.

Enable Standard Edition

Log in to the MaxCompute console ( https://maxcompute.console.aliyun.com/) and find MaxAgent in the left navigation to enter the standalone page.

Enable Standard Edition . If the page indicates it is not yet enabled, follow the on-screen guide; otherwise, start chatting immediately.

Credits are MaxAgent's usage quota. Standard Edition opens at 0 CNY and provides 300 Credits per month , allowing free-tier exploration before purchasing Credit packs. The purchase link is

https://common-buy.aliyun.com/?commodityCode=odps_MaxAgent_public_cn

.

[1] https://maxcompute.console.aliyun.com/ [2] https://common-buy.aliyun.com/?commodityCode=odps_MaxAgent_public_cn
img1_console
img1_console

Confirm Account Permissions

Enabling the service and accessing project resources are separate. When MaxAgent reads project basics, job logs, or table metadata, it acts as the requester. The requester identity must have permission to view the corresponding project, logs, or table information. If the requester is a RAM user, contact the primary account or administrator to grant required authorizations per actual task. Note: being able to enter MaxAgent does not imply access to all projects. This article only asks general product knowledge, so no project-specific log or table access is needed.

Complete the First Question

For the first experience, pick a question that does not depend on production data. Example: before using external tables, understand their use cases, prerequisites, and common limits. In the session input box, send:

1. 请介绍 MaxCompute 外部表的适用场景、创建前提和常见限制。
2. 请区分通用说明与不同数据源的特殊要求,并附上相关官方文档。
3. 如果某项限制需要结合数据源或格式判断,请明确说明。

This phrasing clarifies the desired content, required evidence, and operational boundaries. The expected answer should explain suitable scenarios for external tables, what to confirm before use, and which limits depend on specific data sources or formats.

If the answer remains too broad, add context in the same session:

1. 我的数据存放在 OSS,暂时只希望了解读取方案。
2. 请收窄到 OSS 外部表,说明需要确认的授权和数据格式。

"Is the Answer Correct?" — Verify One Fact

After receiving the answer, pick one conclusion that affects subsequent decisions and cross-check it with the official documentation. For instance, verify: "Does using an external table copy source data into MaxCompute?"

The MaxCompute official document "External Table Overview" states that external table data resides in the external system; using an external table does not copy data into MaxCompute storage.

MaxCompute Official Document "External Table Overview" : https://help.aliyun.com/zh/maxcompute/overview-11
img2_doc
img2_doc
img3_doc
img3_doc

Review Historical Conversations

On the left side of the MaxAgent standalone page, historical sessions can be found and revisited. Important answers and verification sources can be copied as text or saved as screenshots.

Understand Credits: Shared Quota, Personal Limits, and Usage

Credits are MaxAgent's usage unit, mainly covering model tokens and compute resources launched by MaxAgent instances. Input questions, context, and generated answers all consume Credits; fees for tasks submitted to the MaxCompute execution engine are not included in Credits. This article only does knowledge Q&A when DataAgent runs SQL queries or later scenarios actually submit SQL, separate MaxCompute compute costs apply.

A tenant (one primary account and its related identities) shares the tenant's available Credits. Administrators can set a tenant total limit and configure default personal limits or individual limits for RAM Users, RAM Roles, etc. A personal limit is merely a usage ceiling within the shared pool — it does not reserve an independent Credit allocation for that identity. Any of the following constraints being unmet will block new MaxAgent requests: tenant available balance, tenant total limit, or personal limit.

Credit limit management is accessed via MaxCompute Console > Intelligent Assistant > MaxAgent Management > Quota Management.

img4_credits
img4_credits

To view quotas and usage, go to MaxCompute Console > Intelligent Assistant > Credits Details, which shows total Credits, used Credits, remaining Credits, and usage breakdown.

img5_credits
img5_credits

Troubleshooting: Identify Which Layer Is Blocking You

If the page prompts that Standard Edition is not enabled, confirm the enablement status first.

If a permission error appears, verify the current login identity and required grants; for specific projects, check resource-level permissions — do not bypass by over-granting.

If Credits are insufficient, distinguish three cases; do not bypass quota controls by switching unrelated accounts.

Tenant shared balance exhausted: authorized admin purchases or tops up Credits per current site rules.

Tenant has available balance but total limit reached: check whether the tenant total limit needs adjustment.

Only the current identity hits its personal limit: adjust the default personal limit or that identity's individual limit.

If the problem is missing project info, job info, or time range, supplement context first; incomplete input is a different issue from insufficient permissions or quota.

Next Step: Ask a Complete Question

Now open MaxAgent, copy the example to ask your question, verify one conclusion with the official docs, then review the history session and the quota page. After completing this exercise, pick a work-relevant question that requires no resource changes and continue exploring.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

MaxComputeAI assistantAlibaba Cloudexternal tablesDataAgentbig data tutorialCredits quotaMaxAgent
Alibaba Cloud Big Data AI Platform
Written by

Alibaba Cloud Big Data AI Platform

The Alibaba Cloud Big Data AI Platform builds on Alibaba’s leading cloud infrastructure, big‑data and AI engineering capabilities, scenario algorithms, and extensive industry experience to offer enterprises and developers a one‑stop, cloud‑native big‑data and AI capability suite. It boosts AI development efficiency, enables large‑scale AI deployment across industries, and drives business value.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.