Heatmaps, Treemaps, and Diagrams That Match the Question

July 19, 2026 · InfraScout Team

The AI has been able to draw charts and diagrams inline in a conversation for a few weeks. Watching what it drew turned up a pattern: the available shapes did not always match the shape of the answer, so it settled for whichever was closest.

Findings by severity across a fleet of hosts is a matrix, and a matrix squeezed into grouped bars is hard to read. Cost across twenty resource groups is a proportion, and twenty pie slices with long labels is not a chart, it is a legend. A remediation plan is a set of items positioned by severity against effort, and there was no form for that at all.

This release adds the missing shapes.

Heatmap, treemap, and pie

Heatmap renders a matrix of intensities — each series becomes a row, the labels become columns, up to twelve of them so the result stays legible. It is the right form for anything with two dimensions and a magnitude: findings by severity across hosts, sign-in failures by hour across days, patch level by device group.

Treemap renders proportions where there are too many categories, or labels too long, for a pie to survive. It holds up to eight cells, which is a constraint rather than a limitation: the intent is that the tail gets aggregated into an "Other" cell rather than every category being shown at a size nobody can read.

Pie joins donut for share-of-whole. Same data, differing only in whether the center is open.

Color that carries meaning

The two new matrix forms take a color scale, and the choice is about semantics rather than taste.

Severity runs green through amber to red, for anything where higher is worse — risk counts, open findings, failures. It reads the way an operator already expects a risk display to read, which means the shape of a problem is legible before anyone consults the legend.

Sequential is a single-hue ramp, for neutral magnitudes like cost, disk usage, or device counts. Size without the implication of danger, because an expensive resource group is not a dangerous one.

Both ramps derive from the portal's own theme rather than a charting library's defaults, so a chart looks like part of the product. Labels are re-inked for contrast against whatever cell color they land on — the alternative is white text on a pale cell, which is a value nobody can read.

Three diagram types that fit assessment work

Diagrams gained three types chosen because they match shapes an assessment produces naturally.

Timeline for a sequence of events — an incident reconstruction, a change history, a patch rollout.

Mindmap for a decomposition, where the point is how a topic branches rather than how it flows.

Quadrant chart for prioritization, and this is the one worth trying deliberately. Ask for remediation work positioned by severity against effort and you get the four quadrants the conversation was implicitly about anyway: fix now, plan it, easy wins, and leave it. A prioritized list makes you infer the tradeoff. A quadrant chart shows it.

All of them now render on the portal's palette. Mermaid's default theme is fine on a white page and jarring in a dark product, so diagram colors are derived from the same theme tokens the charts use, including the section palette that mindmaps and timelines draw from. Charts and diagrams also agree on the order in which they assign colors, so the same series is the same color whichever form it appears in.

They re-ink when you change theme

A quieter fix that anyone who switches theme will notice. Several charts kept the previous theme's colors after a switch and only corrected on a reload — they read their colors from resolved styles, and a change to those styles was invisible to the framework's reactivity.

The dashboard's severity donut, burndown chart, activity heatmap, and category chart all re-derive their colors now, as do charts and diagrams in a conversation. Switching theme repaints rather than leaving stale ink behind. We verified this by reading the actual rendered colors before and after a switch rather than by looking at it, because looking at it is exactly how this class of bug survives.

In the same pass the theme list was trimmed to light and dark. The extra stock themes multiplied the verification surface without anyone having asked for them. Anyone who had selected a removed theme lands on dark without an error.

Try it

Ask for something matrix-shaped — "show me open findings by severity across our domain controllers as a heatmap" — or something proportional across many categories. For the diagram side, ask the AI to plot your current findings by severity against remediation effort and see what the quadrants tell you.

Full reference in Chart Display and Chat. Questions? info@infrascout.cloud.