Loading the documentation…
When a Mind answers with numbers or relationships, it can show them as a chart, a set of KPI cards or a diagram, inside the answer itself, alongside a reference that opens the query or the document behind it. This page explains what DataMind's AI, Kartlos, can draw, the rules each block follows, and how to read each visual. To read answers in general, start with Ask a Mind; for message formatting, tables and code blocks, see Formatting messages and rich answers.
A Mind can draw these kinds of charts:
Series colours come from the chart palette set for your workspace. The chart cannot pick its own colours; any colours inside its settings are ignored. If a pie's data contains a negative value, the same numbers are drawn as a bar chart instead.
| Chart | What it shows | Its rules |
|---|---|---|
| Pie | Parts of one total. | At most six slices; with more groups, the smallest fold into a single "Other" slice. |
| Pie comparison | Two totals side by side. | Exactly two pies. |
| Bar | Comparisons and rankings. | Bars stack when the mix is the point; read a stacked bar as parts of one whole. |
| Line | Trends over time. | Can carry actual figures and a forecast in one line: the forecast continues from the last actual point. |
| Area | Trends where the volume matters. | Drawn like a line. |
| Composed | Two measures in one picture. | Bars and lines or areas together, each series on its own scale when needed. |
| Radar | A profile across several measures. | Best for four to eight measures across two to five items; all values share one scale. |
| Radial bar | Cyclical data, such as months or weekdays. | Values are percentages or normalised figures. |
| Sankey | Flows. | Shows how a quantity splits between stages. |
A profit and loss is always drawn as a Sankey: money runs from revenue through costs to profit, and a deeper structure is split into several diagrams by category.
A visual answer follows a fixed order: KPI cards first, then charts, then tables, and plain text last. Even a small result can arrive as a picture, and plain text stays for single values and short explanations. How a Mind plans and shapes an answer is on How a Mind answers.
Each chart sits in its own block with a small toolbar. Point at a button to see its tooltip:
| Button | What it does |
|---|---|
| Chart view | Shows the drawing. |
| Table view | Shows the same data as a table: search it, show or hide columns, 50 rows per page. |
| View code | Shows the chart's settings, read-only, with a copy button. Desktop only. |
| Maximize | Opens the chart at full screen. |
Point at the drawing to see exact values in a tooltip. Point at or click a legend entry to focus one series; the other series dim. On a pie, the middle of the ring shows the total, and pointing at a slice shows that slice's value and its share of the total.
While the Mind is still writing, the block shows Generating; the finished chart replaces it when the answer completes.
A KPI block shows cards in a grid: two columns on small screens and up to four on wider ones. Each card carries a title, a value and a change indicator:
If a title is shortened, point at it to see the full title.
When the point is structure rather than numbers, the Mind draws a diagram instead of a chart: Gantt charts, flowcharts, graphs and entity relationship diagrams, and where they fit better, sequence, class and state diagrams, timelines, quadrant charts, mindmaps, journeys, kanban boards, XY charts and block diagrams. While a diagram loads, its block shows Loading diagram.
A diagram's labels keep to plain words, without brackets, braces or quotation marks, so it stays drawable; a diagram that cannot be read shows an error card in place of the drawing.
An answer also points to what the conversation produced:
For how a reference looks and opens in the chat, see Formatting messages and rich answers.
Ask for a presentation and the Mind builds it from the conversation. It finishes the analysis first: the deck reads the conversation up to the moment the build starts, so everything it should include must already be in the chat. When the build starts, the chat shows "Your presentation generation has started." The finished deck is saved in Artifacts as a Slide Deck, ready to page through, present or download.
Generate a presentation or slide deck covers the dialog and its settings. For the AI that changes a deck inside the Slides editor, see The Slides AI.
A problem stays inside its block; the rest of the answer remains readable.
A pie cannot show a negative value. When the data contains one, the same numbers are drawn as a bar chart instead. The labels and values stay the same.
Yes. Open Table view on the chart block to read the same data as a table, with search and column controls and 50 rows per page. The Table tab in the maximized view shows the same data.
Click Maximize on the chart, then use Zoom out and Zoom in to move between 0.5x and 4x, or the mouse wheel. Drag the chart to move around while it is zoomed, and click Reset view to return to the normal size.
The code views are desktop only: the View code button and the Code tab in the maximized view. On a phone, the chart and table views remain available.
The chart could not be drawn, and only that block is affected; the rest of the answer stays readable. Click Copy details to copy the failure details, and ask the Mind again if you still need the chart.
The answer ended before all of the chart's data arrived, so the chart shows only the rows it received. The note tells you how many rows are shown. Ask the Mind again if you need the complete chart.
The middle shows the total of all slices. Point at a slice to see its value and its share of the total; the other slices dim while you point.
Yes. Series colours come from the chart palette set for your workspace, and any colours chosen inside the chart's settings are ignored.
No. A pie draws at most six slices; when the data holds more groups, the smallest ones fold into a single "Other" slice, so the picture stays readable.
A profit and loss is a flow, so it is always drawn as a Sankey: you see how money runs from revenue through costs to profit, and where it splits on the way. A deeper structure is split into several diagrams by category.
Whenever the conversation produced a query or a document, the answer includes a reference to it, so you can open the item behind the answer. The reference points at the item itself, not just at its name, so it opens the right one even when two items share a name.
The analysis is done and the deck build has begun. The deck is built from the conversation up to that moment, so anything added later does not change it; when it is ready, it is saved in Artifacts as a Slide Deck.