Product Positioning
Buddy is a built-in AI assistant in DataBuddy, positioned as a digital data employee. It encapsulates capabilities such as Data Integration, data modeling, data development, data analysis, intelligent Ops, and metric management on the DataBuddy platform into callable Skills, enabling data engineers and data analysts to complete tasks through conversational interaction in natural language that would otherwise require clicking through multiple modules and writing extensive code.
Buddy is not a standalone tool but the AI entry point for the entire DataBuddy platform: it can be invoked as a sidebar in every module, automatically identifies the current page context (for example, it automatically associates with the current SQL in the SQL editor), and enables AI assistance throughout all data workflows.
Entries
Main Entry: Left Navigation
Buddy exists as a top-level menu at the top of the left sidebar in the DataBuddy console. After you click to enter, you will see the Buddy main page, which contains:
Left navigation area: mode switcher (data development, data analysis), space list (data analysis mode only), history session list, scheduled tasks, and knowledge base.
Middle conversation area: sample scenario cards and a conversation input box.
Right-side artifact preview area: displays artifact previews when artifacts are generated during conversations (such as HTML reports and dashboards).
The main page provides three types of expert scenario cards. After you click a card, a sample Prompt is automatically filled into the input box:
Data Engineering Expert: covers data ingestion, ETL development, workflow orchestration, and Ops diagnostics.
Data Analysis Expert: covers data querying, anomaly attribution, report generation, and dashboard building.
Data Governance Expert: covers data modeling, metric registration, and knowledge base exploration.
Sidebar Entry (in Other Modules)
In modules such as Studio, Workflow, and Catalog, Buddy exists as a collapsible right sidebar. After it is opened, it automatically identifies the current page context:
|
Studio SQL editor | Currently open SQL file and selected statement |
Studio Notebook | All cells and run history of the current Notebook |
Workflow editor | Current workflow DAG and selected task node |
Catalog asset page | Currently selected table / view / model |
Conversation Mode
Two mode switches are provided at the bottom of the Buddy dialog box:
Mode | Description | Scenario |
Agent Mode (Default) | Buddy can operate on platform assets, such as generating and saving SQL files, creating data ingestion tasks, and registering metrics. | Real business task execution: data retrieval, table creation, and workflow orchestration. |
Ask Mode | Buddy only provides knowledge-based Q&A and does not operate on platform assets. | Learning, principle discussion, and concept clarification. |
For detailed behavior differences, see Agent Basic Capabilities.
Dialog Box Structure
The dialog box structure on the Buddy main page is consistent with that in the sidebar, containing the following from bottom to top:
1. Sample Scenario Area: cards for Buddy's expert capabilities. After you click a card, a sample Prompt is filled into the input box.
2. Multiline text input box: supports new lines with Enter and sending with Shift+Enter. The maximum length is determined by the model context window.
3. Bottom toolbar (from left to right):
Folder selector: opens the Workspace file tree for you to select the working directory for Buddy in this session. After you select a folder:
The model prioritizes this directory as its main workspace, understanding the project structure and reading relevant files.
Generated artifacts are stored in this directory by default.
Select a folder to not restrict the access scope. If the user requests to read content from other paths, Buddy can still access it.
Upload File: supports uploading CSV, Excel, JSON, SQL, image, and other formats as conversation context. After upload, the file name, size, and format summary are displayed, and multiple files can be uploaded at the same time. For details, see Agent Basic Capabilities.
@ Reference: supports referencing three types of context in a conversation: Catalog assets (tables, metrics), Workspace entities (SQL files, workflows, and so on), and knowledge bases. For details, see Agent Basic Capabilities.
Mode Switch: switches between Agent and Ask modes.
Model Selector: selects the large model used in this conversation. Each option displays the model icon and name.
Skills Selection: In Agent mode, you can temporarily select a subset of Skills to enable.
Send Button: disabled when the input box is empty.
Session Organization
Buddy adopts the same conversation organization design as CodeBuddy, which is divided into two types: Tasks and File Spaces.
Task
Definition: A single, independent conversation session that is not bound to a specific Workspace folder.
Output Scope: Buddy can generate files in any folder within the Workspace where you have permissions.
Applicable Scenarios: ad hoc queries, quick consultations, and concept clarifications.
Conversation Attribution: After the conversation ends, it is archived to the Tasks list in the left navigation.
File Space (Project)
Definition: A collection of conversations bound to a specific folder in the Workspace.
Output Scope: All file artifacts in this conversation are strictly limited to the selected folder.
Applicable Scenarios: continuous data engineering development, team collaboration, and standardized projects.
Conversation Attribution: After the conversation ends, it is archived under Projects in the left navigation and categorized by folder name.
Folder selection rules:
|
Selection granularity | Any folder node in the Studio file tree. |
Permission validation | The current account must have CAN EDIT or higher permission on the selected folder. |
Change restriction | After the project is created, the bound folder cannot be modified. |
Standard Elements of Returned Content
Buddy's responses consist of several standard elements. In a complex analysis conversation, you may see the following in sequence:
|
Plain text / rich text | Analysis of reasoning processes, data insight conclusions, and recommendation text. |
Todo Task Progress | Collapsible multi-step task list. Each step is marked with one of three statuses: Completed / In Progress / Pending, and is checked off as execution progresses. |
Intent Clarification Confirmation Box | When your requirements are vague or can be interpreted in multiple ways, Buddy confirms the intent through multiple-choice questions before proceeding, avoiding ineffective execution. |
Code block | A SQL or Python code block with syntax highlighting on a dark background. The upper-right corner supports copying, viewing the full SQL, and running operations. |
Command execution + terminal output | System commands executed by Buddy and terminal output results. |
Interactive HTML preview | A right-side drawer opens HTML visualization artifacts (reports, dashboards, and more), supporting interactive viewing of chart details. |
Markdown table | Data query results are rendered in table form. |
File artifact card | Generated SQL files, reports, workflows, scripts, and more are displayed as cards. Click a card to preview it on the right or jump to the corresponding DataBuddy module. |
Tool Invocation Status | A collapsible block with messages such as "Reading file...", "Running command...", and "Querying data..." that shows the real-time status of Buddy calling external tools. |
Outputs and Navigation
All artifacts generated by Buddy during the conversation are collected in the Artifacts list, which is unique to each session. Artifacts fall into two categories:
Artifacts That Can Be Previewed in Buddy:
Markdown/HTML files (displayed as rendered by default)
SQL/Notebook/Python files (displayed as source code files)
Artifacts That Can Only Be Viewed by Jumping to the Corresponding Module:
Dashboard: Go to Dashboard.
At the bottom of each conversation, quick access cards for the first two artifacts are also displayed, along with a View All Artifacts button that jumps to the complete artifact list.
For detailed rules, see Agent Basic Capabilities.
Session History
The History sessions in the left navigation are sorted in reverse chronological order, and Buddy automatically generates a session title based on the first message of the session. Click any historical session to continue the conversation, and the complete context (messages, artifacts, selected model, and selected mode) will be restored.
References
Two Major Scenario Modes: Data Analysis vs. Data Development