Using AskTable
Data dashboards
Create dashboards through conversation or combine existing results to monitor business metrics.
Guide · Module: product · Users · Deployment: Not specified · Confirmed product versions: Not specified · Source: Existing documentation
Data dashboards let you monitor business metrics over time. A dashboard contains multiple components, each linked to a refreshable data table and displayed as a table or chart.
There are two ways to build one:
| Method | Best for | Entry point |
|---|---|---|
| Build through conversation | Starting from scratch: describe what you want to see and let the agent arrange it | “+” menu to the left of the conversation input → “Build Dashboard” |
| Build manually | Arranging results you have already saved | Main navigation on the left → Data dashboards → “Create dashboard” |
Build through conversation
Enter building mode
Open a conversation, click “+” to the left of the input, and select Build Dashboard.
The input placeholder changes to “Describe the Dashboard you want…”, and a “Build Dashboard” pill appears to the left of the input to indicate the mode is on. Click × on the pill to exit and return to a normal conversation.
Describe what you want to see
Describe the dashboard's contents, such as the metrics, time range, comparison dimensions, and sort order.
With no components yet, the interface shows only the conversation. Once the dashboard has its first component, it splits into two columns: “Dashboard preview” on the left and the conversation on the right. When the left side is empty, it says: “No elements yet. Ask AI in the conversation on the right to create your first chart or table.”
Let the agent explore data and test queries
No action is needed from you at this step. The agent searches your data semantics for relevant tables and fields and reads table structures and business documents. After writing SQL, it runs it first to validate column names and data before creating a dashboard component. It does not put an unverified query on the dashboard.
Refine through conversation
Describe changes naturally: “Show the last 30 days' trend in the top left”, “Remove average order value”, “Change the bar chart to a line chart”, or “Add a store ranking”.
Changes take effect in place: update titles or SQL, change display formats, or adjust width and height while retaining existing positions. The agent only adds, updates, or removes components; it does not rearrange the layout you have already organized.
Finish on the dashboard detail page
Click “Open detail page” in the top right of the preview to open the full dashboard page, where you can configure scheduled refreshes, share links, and make finer layout adjustments.
Metric cards (BigNumber)
Two kinds of single-number metric cards are available. Both use table display mode and a fixed height of 2 grid units:
| Type | Purpose | Required SQL result |
|---|---|---|
| BigNumber single value | One core KPI, such as today's order count | One row and one column, displayed as a large centered number |
| BigNumber trend | Compare the current period with the previous period | One row and two columns: current_value and previous_value |
The trend card's column names, current_value and previous_value, are fixed conventions recognized by the display layer. Keep these exact English names without translating them or adding prefixes or suffixes. The display shows current_value and calculates the percentage change using both columns.
Key constraints when building through conversation
| Item | Value or behavior |
|---|---|
| Underlying component | Each component is backed by a refreshable data table that stores its data source and SQL |
| SQL | Must be repeatable; use time functions for relative dates instead of hard-coding specific dates |
| Grid | 12 columns. Specify width and height; the system places components from left to right and starts a new row when needed |
| Manual layout takes priority | Positions you have set by dragging and resizing are retained; AI does not rearrange existing components |
| Changes to existing components | Only changed when you request it; changing SQL triggers an immediate refresh |
| Component types | Charts or tables; use tables for detailed lists |
Build and adjust manually
You can also assemble a dashboard from previously saved results.
Open data dashboards
Click “Data dashboards” in the main navigation on the left. The page title is “Data dashboards”, and the top right offers dashboard search by name.
Objects created here are called dashboards: the button is “Create dashboard”. When the page is empty, use the button above to create one.
Create a dashboard
Click “Create dashboard” and enter a name. The list has columns for name, component count, and modification time, with modification time shown relatively. Use “Sort” to order by last modified, name, or creation time.
Add components
Click “Add component” and choose a data artifact. Each item shows its row count, and previously added items are marked. If nothing is available, the project has no saved results yet; go to Conversations and use “Save to Tables” first.
An empty dashboard starts with a blank canvas. Use “Add component” to begin.
Adjust layout and display
Drag components to change their position and size. Each component can switch between table and chart display.
Dragging is unavailable when the window is too narrow; the interface tells you to widen it.
Scheduled refreshes and sharing
Set a refresh schedule
Open “Scheduled refresh”, enable it, and select a frequency: “Every 10 minutes”, “Hourly”, “Daily”, “Weekly”, or “Monthly”.
The panel shows the refresh schedule and next run time. You can also click “Refresh now” to trigger it manually. Refresh records show the last refresh time and a failure status when a refresh fails.
When refreshes happen
After the scheduled time, all data tables refresh the next time someone views the dashboard, rather than immediately in the background. Click “Refresh now” if you need current data immediately.
Share the dashboard
Click “Share dashboard” and enable the public link switch. Anyone with the link can view the dashboard, with data loaded live.
Key constraints
| Item | Value | Where to change it |
|---|---|---|
| Minimum refresh interval | As often as “Every 10 minutes” | Dashboard → “Scheduled refresh” |
| Refresh timing | On the next view after the scheduled time, not a punctual background refresh | — |
| Component source (manual) | Only saved data artifacts can be selected | Use “Save to Tables” in a conversation first |
| Sharing permissions | Public links are read-only; data loads live | Dashboard → “Share dashboard” |
| Component count | Each row shows the component count; no hard limit | — |
| Trend card column names | Must be current_value and previous_value | Determined by the SQL generated through conversation |
Troubleshooting
| Symptom | How to identify it | What to do |
|---|---|---|
| Component refresh fails | The card says “Refresh failed” or shows no data | Open the underlying table, refresh manually, and retry; if the source schema changed, first use “Resync” on the data source page |
| No components available to add | The selection list is empty | The project has no saved results; use “Save to Tables” in a conversation first |
| Component shows no data | The card says no data is available | Click “View details” to inspect the underlying table's status |
| Cannot drag a component | Dragging does nothing and the interface says the window is too small | Widen the browser window |
| Share link does not open | The page says the dashboard does not exist or sharing is disabled | Return to “Share dashboard” and check that the public link is still enabled |
| Numbers are outdated | The card's refresh time is noticeably earlier than now | Click “Refresh now”, or check that scheduled refresh is enabled |
| Metric card does not show a trend | Only one number appears, with no percentage change | Make SQL return exactly one row and two columns named current_value and previous_value |
| Conversation never adds anything to the dashboard | Dashboard preview remains empty | Confirm “Build Dashboard” mode is enabled, indicated by the pill to the left of the input |
Next steps
- Update a component's data: return to Conversations, rerun, and save again
- Make dashboard numbers explainable: improve table and field descriptions to establish consistent definitions
- Let colleagues receive results in Feishu: see Scheduled analysis reports
- Apply a repeatable analysis method automatically: attach analysis skills to the agent