Charts
Charts are the primary analytical output in Lakehousecat. When you ask an analytical question in a session using a Custom Model, the result is automatically saved as a chart and made available in the Analytics section. From here you can view, share, clone, and manage all your charts.
Chart Lifecycle
Understanding how charts are created, persisted, and affected by model changes helps you manage your analytics safely.
Creation and Persistence
A chart is persisted immediately when it is generated — as soon as the LLM returns an answer and the visualization is rendered in a session. No manual save is required. The chart appears in My Charts right away.
Relationship to the Custom Model
Each chart is linked to the Custom Model it was created from. This relationship is recorded at chart creation time.
What happens when a Custom Model is retrained?
The Operator's approach to Custom Model semantic training is deliberately conservative — existing semantic structures are not automatically deleted or replaced. This means charts built on an existing model continue to work after a re-training run, because the underlying structure is preserved.
However, if the Custom Model's semantic layer is deleted and rebuilt from scratch (via the Delete Semantic Layer operation), the structural definitions change. Charts that reference the old structures may no longer function correctly after this.
Only delete and rebuild a Custom Model's semantic layer during development, before the model is shared with users. For a model that is already in productive use, prefer creating a new Custom Model for redesigns and keeping the existing model unchanged.
Locking Productive Models
To protect a Custom Model that users depend on from accidental structural changes, use the Locked toggle in the Custom Model settings. A locked model cannot be edited or deleted. This is the recommended practice for any model that has been shared with users and has live charts depending on it.
Visibility and Access
A chart generated in a session is only visible to you by default. It appears in your chart history (My Charts) and can be accessed from there. Other users cannot see it until you explicitly share it.
To make a chart accessible to others, use the Share action. See Sharing for details.
Charts have internal status values (Draft, Published, Archived) that originate from the underlying Superset system. These statuses are not actively processed or surfaced in the Lakehousecat interface and have no effect on visibility or access. Access is controlled exclusively through the sharing model described above.
Chart Tabs
The Analytics section has three chart-related tabs:
My Charts
Charts you own. Available actions:
| Action | Description |
|---|---|
| Share | Share the chart with users or groups |
| Clone | Create a personal copy of the chart |
| Delete | Permanently remove the chart |
Shared by Me
Charts you have shared with others. The only available action is:
- Unshare — revoke access for all recipients
Shared with Me
Charts shared with you by others. The only available action is:
- Clone — create a personal copy of the chart
Chart Detail View
Click on any chart to open the detail view. It contains:
- Chart visualization — an interactive preview of the chart, embedded via the Superset SDK. If the chart is not yet configured for embedding, a warning is shown instead.
- Chart information — Chart ID, title, chart type, linked Custom Model, Dashboard ID, and embeddable status
- Chart State & Settings — two toggles:
- Enabled — if disabled, the chart is not available for use
- Locked — if locked, the chart cannot be edited or deleted
- Chart Data — on-demand access to the underlying data (see below)
Visualization Actions
When hovering over the chart visualization, action buttons appear:
| Action | Description |
|---|---|
| Refresh | Reload the chart data and visualization |
| Edit | Open the chart in Superset for advanced editing |
| Fullscreen | View the chart in full screen. Press Escape to exit. |
Editing a Chart in Superset
Clicking Edit opens the chart directly in the Superset chart explorer. Authentication is handled automatically — no separate login is required. In Superset you can:
- Change the chart type, dimensions, metrics, and filters
- Adjust colors, labels, and visual settings
Changes are saved in Superset and immediately reflected in Lakehousecat.
Chart Data
The Chart Data section at the bottom of the detail view allows you to inspect the data behind the chart on demand:
| Button | Description |
|---|---|
| Load Sample Data | Loads up to 100 raw rows from the underlying data source. Displays the data as a table and shows the SQL query used to retrieve it. |
| Load Chart Data | Loads up to 100 rows of the aggregated data as processed by Superset for the chart visualization. |
Both views are rendered as interactive tables in the detail view.
Chart Types
The LLM automatically selects the most appropriate chart type based on your data and question. You can also specify a chart type in your prompt (e.g., "show as a line chart") and the system will use it if the data supports it.
Core Types
| Type | Description |
|---|---|
| Bar | Standard bar chart for categorical comparisons |
| Line | Standard line chart for trends |
| Area | Line chart with filled area below |
| Pie | Proportional distribution |
| Scatter | Correlation between two numeric dimensions |
| Table | Tabular data display |
| Pivot Table | Cross-tabulation with row and column grouping |
| Big Number | Single KPI value |
| Big Number with Trendline | KPI value with a sparkline trend indicator |
Time-Series Types
These chart types require a temporal dimension in the data (a date, datetime, or time column).
| Type | Description |
|---|---|
| Time-series Bar | Bar chart over a time axis |
| Time-series Line | Line chart over a time axis |
| Time-series Area | Area chart over a time axis |
| Time-series Scatter | Scatter over a time axis |
| Time-series Step Line | Step-function line over a time axis |
| Time-series Smooth Line | Bezier-smoothed line over a time axis |
| Mixed Chart | Combines bar and line series on a time axis |
Other Types
| Type | Description |
|---|---|
| Heatmap | Two-dimensional intensity grid |
| Histogram | Frequency distribution |
| Funnel | Sequential stage drop-off |
| Gauge | Single value on a radial scale |
| Treemap | Hierarchical proportional rectangles |
| Sunburst | Hierarchical radial chart |
| Sankey | Flow diagram between nodes |
| Waterfall | Cumulative value changes |
| World Map | Geographic data on a world map |
| Box Plot | Statistical distribution with quartiles |
| Radar | Multi-axis comparison on a radial grid |
| Bubble | Scatter with a third dimension encoded as bubble size |
| Word Cloud | Text frequency visualization |
Not all chart types are appropriate for all queries. The LLM selects based on data context. If the selected type does not match your intent, rephrase your prompt with a specific chart type.
Point-Density Map Types
These chart types require geographic point data (latitude/longitude) and are rendered as an interactive map.
| Type | Description |
|---|---|
| Scatter | Individual points plotted at their geographic location |
| Grid | Points aggregated into a regular grid of cells |
| Screen Grid | Points aggregated into a grid sized to the screen resolution |
| Hexagon | Points aggregated into a hexagonal binning grid |
| Heatmap | Smooth density surface based on point concentration |
| Contour | Density surface shown as contour lines |
Chart Variants
The chart detail view offers a Change Type switcher next to the visualization. It lets you switch the current chart to a related type instantly — in place, without asking the LLM a new question and without re-querying the data source.
The offered variants depend on the current chart's type:
- Standard charts (single dimension + single metric) offer: Area, Bar, Line, Pie, Table.
- Point-Density Map charts (Scatter, Grid, Screen Grid, Hexagon, Heatmap, Contour — see above) offer the other five types from the same family instead, since these share the same geographic data shape.
To switch:
- Open a chart in the detail view.
- Click Change Type and select a variant from the list.
- The chart updates immediately in place.
A ↩ reset control returns the chart to the type it had when you opened the detail view.
Only the chart owner can change the type, and the chart must not be Locked. If the current data does not support the selected variant, an error is shown with a link to open the chart in Superset for manual, full-control editing.
Audit Log
The detail view includes an Audit Log tab that records all changes made to the chart — who changed what and when.
Creating Charts
Charts are not created manually. They are generated automatically when you ask an analytical question in a session using a Custom Model. For a step-by-step walkthrough, see Create a Chart.