The Data Agent workbench provides an agent-based intelligent chat entry. Users can directly select a pre-configured agent to quickly complete tasks, such as data queries, knowledge retrieval, and analysis report generation, through natural language chats.
Daily data queries: Business personnel can directly ask agents questions to quickly obtain business data such as sales amounts, inventory, and orders.
Analysis by multiple rounds of chats: Users can ask follow-up questions within the same chat to gradually delve deeper into business issues.
Chat history review: Users can quickly find and resume previous analysis chats via the left history panel.
Taking the chain operation analysis scenario as an example, select the Chain Operation Analysis agent, and enter the question "Sales amount per state in 2025". The agent will automatically parse the question, fetch data from the data source, and return the results. The following figure shows the effect.
Follow-up questions, for example, "Top 3 states in the Southwestern United States", can be asked to further refine the analysis dimensions. The following figure shows the effect.
Log in to Data Agent, switch to the Experts page, and click the corresponding expert agent card to enter the chat page.
Return to Experts: Click the back arrow in the upper left corner of the page to return to the homepage of Experts with one click.
Switch to another agent: Click the current agent's name at the top, and search and select another agent in the drop-down list to switch chats without returning to the homepage.
Quickly enter the configuration page: Click the configuration button on the right to jump directly to the console configuration page of the current agent.
To start a new analysis task, click the left New Chat button to clear the current chat context and begin a fresh chat. Chat history is displayed on the left panel. You can click one to enter a historical chat and continue unfinished analysis.
In the chat input box, enter your question (for example, "What can you help me with?"), and click the sending button (or press Enter) to submit the question.
Moreover, two ways to add data resources are available on the left side of the input box to supplement agent analysis materials:
Local file upload: Click the + button to upload local files, such as Excel tables, Word files, PDF files, and Markdown files.
FineBI resource selection: Click the icon to select FineBI resources from Directory or My Analysis on the FineBI platform.
The added files/data can be referenced in questions using the @ symbol, allowing the agent to complete the analysis based on the corresponding data.
The agent will automatically parse the intent of the question, invoke the corresponding skill, and return the result within the chat area.
For further analysis, you can continue to enter follow-up questions in the same chat. Currently, the following types of follow-up questions are supported:
(1) Direct follow-up questions: Directly enter a follow-up question in the chat input box for a secondary Q&A.
(2) Word-marked follow-up questions: Select part of the content in the returned results, click Ask Further, and then describe the specific follow-up question in the chat input box.
(3) Source/output-referenced follow-up questions: Select and reference the file content from Source or Output in the workspace, and describe the specific follow-up question in the chat input box.
Chat correction is primarily applicable to the following two types of business scenarios:
Agent response deviation scenario: If the agent's output contains errors or fails to meet user expectations, users can directly raise objections, correction requests, or adjust instruction strategies. The agent will automatically review and reflect on its output, and regenerate a compliant result.
User query error scenario: When a user-entered question is deviated and the agent has already entered the response generation process, users can directly resend a concise and correct question in the chat box. The agent will automatically interrupt the original task and respond based on the latest correct question.
Only when the current version supports scheduled tasks, the Scheduled Task entry is displayed on the left side of the agent chat page. You can view or set scheduled tasks related to the current agent through this entry.
The clone is used to display the historical chats with the agent through the bot in an IM tool. You can view these historical chats in the clone and continue the chats.
During the agent analysis, you can perform related operations on the chat content. For details, refer to Chat Operations.