Features

Updated at:
Copy as MD

ADA provides capabilities that cover the entire workflow from conversational interaction to enterprise management. Each capability can be enabled independently and implemented in phases based on business priorities. You do not need to activate all capabilities at once.

Key concept

ADA capabilities are organized into four areas based on "who is using them" rather than by technical module.

Workbench is the daily workspace for users. It includes interaction entry points, built-in agents, connectors, skills, knowledge and context, scheduled tasks, IM channels, and artifact management. Capability Store is the supply side, responsible for the discovery, subscription, and usage management of official agents, skills, and connectors. Enterprise Management is the control plane, responsible for account and data permissions, usage monitoring, data security, and auditing. Subscription & Billing handles product subscriptions, credit metering, and invoicing.

The following table lists the specific capability domains and corresponding documents for these four areas. The Area column indicates the area to which each row belongs.

Area

Capability domain

Feature

Documentation

Workbench

Interaction entry points

Web workbench conversations, IM channel conversations and delivery

Start a conversation, IM channel management

Workbench

Built-in agents

Ada primary agent, domain-specific agents, enterprise custom agents, agent cloning and customization

Agent management

Workbench

Team collaboration

Agent configuration sharing, cross-member collaboration, ownership control

Team collaboration and resource sharing

Workbench

Knowledge and context

Knowledge base (including semantic models and knowledge graphs), Wiki references, file and image uploads, voice input

Knowledge base management

Workbench

Data connection

Big data engine and MCP service connectors, connectivity detection, Data Catalog browsing

Overview of data source access, Create a data connection

Workbench

Scheduled tasks

Scheduled triggering, execution records, result delivery

Scheduled tasks

Workbench

Artifact management

Artifact browsing and search, sharing and export, share expiration control

Artifact management

Workbench

Skills and Open Ecosystem

Built-in skill invocation, custom skill packaging and registration, and skill version deployment

Skill management

Capability Marketplace

Capability discovery and activation

Discover, activate, and view the status of official agents, skills, and MCP services

AI Capability Center

Enterprise management

Account and data permissions

Members and limits, data permission policies, row-level and column-level rules, and auditing

Members and permissions, security and compliance

Enterprise management

Portal and usage

View activated services, analyze consumption, and check usage

ADA Portal Console

Subscription and billing

Subscription and credits

Edition subscription, pay-as-you-go, credit metering, and billing

Billing overview

Capabilities come from two sources, each with a different access method and update responsibility.

Comparison dimension

Built-in platform capability

Official subscription capability

Access method

Available immediately after the service is activated

Available after subscription in the Capability Marketplace

Coverage

Conversations, built-in agents, knowledge, connectors, scheduled tasks, and artifact management

Official agents, official skills, and MCP services for each engine

Depth

Cross-engine general capabilities

In-depth diagnosis and development assistance for a single engine

Update responsibility

Updated with platform releases

Continuously updated by the engine product team

Cost

Counted toward subscription seats and credit consumption

Requires a separate subscription with independent consumption metering

Use case

Analysis and operations scenarios

When your work primarily involves answering business questions, the recommended combination is the chat workspace plus the data analysis expert, supplemented by a knowledge base to unify metric definitions. The reason for this combination is that after metric definitions are centrally maintained, the same metric produces consistent results across different users and different chat sessions. When a session is linked to a knowledge base, responses include wiki citations that you can expand to see which metric definitions were used. The implementation sequence is to first connect to a data source, then enter metric definitions, and finally verify the results in a chat session. For more information, see knowledge base management.

Data development and O&M scenarios

When your work focuses on job development, performance diagnostics, and cluster inspection, the built-in platform capabilities are not sufficient to cover engine-specific details. You need to subscribe to official agents by engine in the AI capability center. Official agents are released by the respective engine product teams. After ADA connects to their capabilities, it can observe engine compute resources and job statuses and provide diagnostics and optimization suggestions without requiring you to manually export metrics for analysis. For more information, see AI capability center.

Periodic delivery scenarios

For daily reports, inspections, and monitoring tasks that follow stable workflows, we recommend that you first complete a task in a chat session, then save it as a skill with fixed input and output formats, and finally package it as a scheduled task for periodic execution. The value of this approach is that it removes human effort from the execution phase and retains only result verification. For more information, see scheduled tasks.

Integration with your own systems

When alerting platforms, internal data portals, or other systems need to trigger data analysis within their own workflows, you can use Open APIs and SDKs to call agents, or use the MCP protocol to connect your own tools to agents. The choice between APIs and MCP depends on the call direction: use APIs when your system calls ADA, and use MCP when ADA calls your system. For more information, see API overview and authentication.

FAQ

  • Q: What is the relationship between platform capabilities and the products in the AI capability center?

    A: Platform capabilities are features provided by ADA itself and are available as soon as the service is activated. Official agents, skills, and MCP services in the AI capability center are subscribable capability products. After you subscribe, they are integrated into the same chat, knowledge base, and permission system. No duplicate agent with the same name is created in the agent list.

  • Q: What features are available if I have not subscribed to any official capabilities?

    A: The chat workspace, built-in scenario agents, custom agents, knowledge base management, data connections, scheduled tasks, and artifact management are available as soon as the service is activated. Engine-specific capabilities require subscriptions to the corresponding official agents. Examples include Hologres query performance diagnostics and EMR task exception diagnostics.

  • Q: Do scenario agents and official agents have overlapping capabilities?

    A: There is some overlap, but responsibilities are divided by depth. Scenario agents cover cross-engine general tasks such as data retrieval, modeling suggestions, and quality inspections. Official agents cover in-depth operations and diagnostics for a single engine. During orchestration, Ada automatically connects to the capabilities of the corresponding official agents based on task requirements. You do not need to manually select them.

  • Q: Which capability should I start with?

    A: Start by completing an end-to-end conversational analysis to verify that data connections and metric definitions are correct, and then expand based on your work priorities. For more information, see the quickstart guide on completing your first conversational analysis.