This document describes how to create and manage Agents in DataBuddy, covering the high-code development creation method, as well as Agent state transitions, permission management, and list filtering.
Overview
The Agent module currently provides the following creation methods:
High-code Development Agent: A code template based on frameworks such as OpenAI and LangChain/LangGraph. Developers edit and deploy code through Studio.
After all Agents are created, they will appear in the "My Agents" list and be deployed and run as Apps.
Prerequisites
Before you start creating an Agent, make sure the following prerequisites are met:
You have activated the DataBuddy service and completed workspace initialization.
The MLflow experiment required by the Agent is ready. For GenAI application and Agent types, you can create a new one during the creation process.
To access assets in a Catalog, such as data tables, you must have the corresponding Catalog access permissions.
Operation Steps
Step 1: Going to the Agent List
1. Log in to the DataBuddy console.
2. Click a workspace to enter it.
3. In the left sidebar, select Agent Development.
Step 2: Selecting a Setup Method
Select the corresponding card based on your use case:
|
Code development Agent (OpenAI framework) | Go to the App installation page and select the OpenAI template to create. |
Code development Agent (LangGraph framework) | Go to the App installation page and select the LangGraph template to create. |
Step 3: Filling Out the Creation Form
Form fields vary depending on the creation method. Enter the required information as prompted on the page.
Step 4: Saving and Going to the Agent Overview
After successful creation, you will be automatically redirected to the Agent Overview page, where you can:
View the Agent identity information, including its name, type, and APPID.
View the run status (Stopped, Starting, Deploying, Running, Error).
Chat directly through the preview window (ChatUI) on the right.
Go to the submodules such as Code Development / Deployment and Ops / Tracking and Evaluation / Intelligent Tuning.
Creation Parameters
High-Code Agent Development
Entry: In the Agent list, click the High-Code Development Agent card to go to the App installation page.
|
Agent name | Unique agent identifier | Yes | - | Uppercase and lowercase letters, digits, and -, up to 30 characters |
MLflow experiment | GenAI experiment associated with the Agent, carrying data such as traces and evaluations | Yes | - | GenAI&Agent experiments for which the current user has CAN_MANAGE permission; can be created instantly in the selector. |
Agent description | Describe the purpose of the Agent. | No | - | Within 500 characters |
Note:
To create an MLflow experiment immediately, click Create in the experiment dropdown list, enter a name for the experiment in the pop-up window (up to 50 characters), and the experiment will be automatically selected after creation. If no project location is specified, the experiment is stored in the current user's home directory by default.
After creation, you can go to the Agent code template in Studio to directly edit, debug, and deploy the code.
My Agent List
The list displays the agents that the current user has permission to use and manage, including those across workspaces under the same root account:
|
Agent name | Filled in at creation |
Agent type | Custom code Agent |
Agent description | Description filled in at creation |
Owner | Agent owner |
Invocation Endpoint | Endpoint name of the Agent. For an App-type Agent, it is the App name. |
Last updated | Last updated time, sorted from newest to oldest, with ascending/descending order switchable. |
Status Transitions
An Agent has the following states during its lifecycle:
|
Stopped | Starting | Starting |
Starting | - | Deploying (initial creation) / Running |
Deploying | Deployment | Running (succeeded) / Previous state (failed) |
Running | Stop / Deploy | Stopping / Deploying |
Stopping. | - | Stopped. |
Error | Start / Stop | Starting / Stopping |
Note:
After the Agent is created for the first time, it enters the Deploying state to complete code deployment and then enters the Running state upon startup. For subsequent code changes, click Deploy to redeploy.
The page displays the corresponding run status and deployment progress bar in different states, allowing you to intuitively check the current progress.
Permission Management
The Agent platform inherits the common permission management of DataBuddy: configure permissions in the upper-right corner of the Agent details page.
|
Available | Can chat and view traces |
Manageable | Can modify, delete, deploy, and configure permissions |
The authorization scope covers members across workspaces under the entire root account.
Note:
The created Agent does not appear in the asset list of the workspace, but it appears in global search results and Recent Access.
The Agent does not appear in Favorites.
Modification and Deletion
Modify: Go to the Agent overview page and use the left-side navigation to jump to the corresponding submodule (Code Development, Tool Configuration, or Intelligent Tuning) to make modifications.
Rename: Edit the name in the info pop-up on the overview page.
Delete: Delete the Agent in the row operation of the Agent list. A confirmation dialog box appears before deletion.
Warning:
Deleting an Agent is irreversible. We recommend that you disable the Agent on the overview page and confirm that there are no online calls before deleting it.
FAQs
Q: Why Does My Agent Remain in "Deploying" Status for a Long Time?
A: There are three common causes: ① Insufficient permissions on the associated MLflow experiment. ② Failed installation of dependency packages in the code template. ③ The platform base model allowlist is not enabled. We recommend that you check the error message on the deployment progress bar or contact the administrator to confirm the allowlist.
References