Operational dashboards over your data items, portals and groups — widgets that filter each other, stored on the server, shared by person, group, link or publicly.
Open Hyz DashboardsA Haiyz Dashboard is an operational screen you assemble from widgets — numbers, gauges, charts, tables, lists, a map, selectors, text and images — each bound to your data and wired to react to the others. It is stored on the server as a versioned document, so it lives with your data and your groups rather than in a file. Yours are at /dashboards.
Four ways to start. On the dashboards page, New dashboard opens the popup — how to begin, then a template, then one name — and the arrow beside it names the ways — From a template, Blank, Import a bundle, and With Hyz AI while the assistant is switched on — so you can go straight to one:
.haiyz-dashboard.json exported elsewhere.In the popup the templates and Blank are cards side by side. Pick one, give the dashboard a name — one is enough; the name in the other language is under More options — and press Create.
The builder then opens: the canvas in the middle, the settings on one rail, the inspector on the other. Add widget opens a popup with every widget type by category — and, while Hyz AI is on, Describe it to Hyz AI: say what the widget shows and it is built over the dashboard's data. Either way the widget opens in the editor with a live preview; Add to dashboard keeps it, Cancel takes it away. In the editor, Change with Hyz AI takes a change in words ("make it a pie chart"), shows it in the preview, and you keep it or discard it. Drag a widget by its header, resize from the corner, or move the selected one with the arrow keys. The inspector has three tabs — Data, Appearance, Actions. Save with ⌘S / Ctrl+S: a save carries the version you opened, so if a colleague saved first you are asked to reload rather than overwrite them.
A widget binds to one of these, or carries a static value:
Every answer says how it was obtained: Layer, read from the data item's Hyz Layer as it stands (edits and approved field records included — the usual case for an item that is online); Live, asked of the provider just now; or Snapshot, computed from the stored file copy, with the time it was taken — the fallback for an item with no Hyz Layer, such as a table. The mode setting is auto (live when the provider can answer, else the snapshot), live only or snapshot only, and the inspector warns in advance whether your query will really resolve live. Not every question can: grouping, buckets and the rarer statistics are computed here, on the snapshot.
The Sources section of a dashboard lists everything it reads, once each, with the widgets that read it — and, for each source you may open yourself, how fresh it is: whether the widgets read the Hyz Layer, the stored file, the portal live or a campaign's records, and as of when; a data item's stage; a linked item's connection and last sync; a campaign's record count. A source you may not open is listed by its kind only.
Only data you can view can be bound, and a public or link-shared dashboard exposes only what is public or yours. Sharing a dashboard never widens access to the data behind it.
| Purpose | Widgets |
|---|---|
| Numbers | Indicator (one value with unit, target and thresholds), Gauge (a value against a range, with bands) |
| Charts | Serial chart (bar, line, area, scatter), Pie chart |
| Records | Table, List, Details (the fields of the selected record), Group card |
| Map | Map, Legend |
| Selectors | Category, Number, Date — these filter the other widgets |
| Content and layout | Text (markdown-lite with tokens), Image, SVG, Embed, Spacer |
Each widget sits in the grid, in the sidebar, or in the header.
A binding is a question, not a table dump. Five kinds:
A filter narrows any of them: field, operator (equals, greater than, contains, in, between, is empty…), value. A filter value may carry a dashboard parameter token, so one control drives many widgets. Widgets can also refresh on a timer.
A chart or an indicator can carry several series — each a statistic over a field, with its own key, bilingual label, colour slot and (for charts) its own bar/line/area form. Grouping by a date field lets you bucket time by hour, day, week or month — in Riyadh local time, so a reading at one in the morning belongs to that day. On a data item that is online, a plain time series is computed where the data lives, however many records it has.
Indicators and gauges accept a calculation over those series keys — pct(load, capacity), (a - b) / b * 100, round(x, 1) — plus a comparison against another series that renders the change as ▲/▼. Text widgets embed other widgets' values with tokens such as {{w.kpi-load}}.
Selectors, charts, tables, lists and the map emit events: a chosen value, a selection, or the map's visible extent. In the Actions tab you say what an event does, and to which widgets:
| Effect | What happens |
|---|---|
| Filter | The target widgets re-ask their question with the emitted clause added |
| Zoom / Pan | The map moves to the emitted geometry or extent |
| Flash | The matching feature blinks on the map |
| Set | A dashboard parameter takes the emitted value |
A header dropdown narrowing every widget to one region, a map whose extent drives the KPIs, a bar you click to filter the table — all are actions. Reset filters clears every one at once.
On a map a viewer selects with a click, by dragging a box, or by drawing a shape (the three select tools at the map's edge). What is selected is features of a data item, so the widgets that read that data item are filtered to them and every other widget is left as it was: pick a tower and the towers' table and figures follow, while a table of something else does not empty. The count of what is selected sits at the map's corner; press it to clear. Filter by this view beside it is the filter by place: on, every widget shows only what is inside the map's view and follows as the map moves; off, the filter is lifted. Moving the map never drops a selection.
A date is the period. A date selector filters the widgets it targets by its field — and a Readings series chart, which has one axis of time, simply draws the period it is sent. On a Readings series the viewer can also drag across the chart to look closer at a span: the chart draws it, the charts its actions name draw the same span, and Whole period puts them back.
The map draws its bound features and any overlays — extra data items under the primary layer, such as transmission lines beneath substations. Each overlay has its own colour slot, line weight, label and an option to obey the dashboard's filters, and can animate as flow: the dashes travel along the line, faster where a numeric field you nominate is larger.
A layer can take its label, size and colour from fields. Colour by gives each value its own colour — or a ramp, for a number with many values — and an asset's alert state reads as it does everywhere: normal green, alerting red, stale amber. So a map over a class's live layer shows which asset needs attention, and a point changes colour when its sensor's state does. For a different symbol, size and colours per layer — a tower, a server, a small dot — style the layers in a Hyz Map (its Style panel) and bind the widget to that map: the widget draws each layer exactly as the map sets it. Clicking a feature selects it; panning emits the extent, and Filter by this view is on from the start so every widget shows what the map shows. A map on a dashboard has no projection toggle: that is the map's own setting. A Legend widget renders any map widget's symbology.
A bar chart grouped by an asset's alert state or service status draws each bar in that state's colour — normal green, alerting red, stale amber — and a chart grouped by a number can read from red to green (Colour scale in its appearance): the lowest value red, the highest green, the rest amber.
For an asset class, Readings by day draws one bar a day over the last week, fortnight, month or quarter: the sensors that were up at the day's end (green — a yes/no measurement read yes, or a number read above the limit you set), down (red) and silent (grey — they sent nothing that day), with the readings accepted that day in the tooltip. Select assets on any map of the dashboard — a click, a box, a lasso — and the chart narrows to their sensors. Click a day to send it to the charts that follow this widget.
A table or list can colour itself by rule. Each rule is: a field, an operator (≥, >, ≤, <, between, equals, not equals, contains, in), a value, one of four colours — green, blue, red or orange — and a scope: just the cell, or the whole row.
Rules are ordered and the first one that matches wins, so put the strongest condition first: a rule for "over 90" placed below a rule for "over 70" will never fire. In a list, the rule can also colour the value chip on the end of the row.
A dashboard can have several pages, shown as tabs. The header carries a title and subtitle, an optional logo and clock, a theme toggle, and any widget placed in the header slot — typically a selector. The sidebar holds selectors and legends, on either side, at a width you set.
Page height decides how a page meets the screen. Fits the screen makes every page exactly one screen: its rows share the height, the header and tabs stay in place, and a table or a list scrolls inside its own card — the way an operations room reads a wall screen. Grows with its content gives every row a fixed height, and a page with many rows scrolls. A new dashboard fits the screen; a page with more rows than the screen can hold scrolls inside the board instead of squeezing its cards. Put what is watched together on one page and the rest on a second one. A figure card that is short shows its icon, number and unit in one row.
Theme is auto (following the site), light or dark; density and background are per dashboard too — Compact brings the cards closer and holds more on one screen. On phones the grid stacks in reading order, or in the mobile order you arrange in Settings — which is also where parameters are declared, named text, number or date values that filters and text widgets reference.
Create takes your chosen sources, reads their schema — field names, types, ranges, distinct counts — and designs a whole dashboard from a sentence: pages, widgets, bindings, actions, even the table colour rules. Revise does the same to a dashboard you have: it starts from the sources it already binds plus the config on screen, and its answer arrives as one undoable step you review and then save.
Both offer an extra-context box (what the data means, what "status = 3" is) and a sample-rows switch letting a few rows travel with the schema, identifiers masked; it is off by default. See Hyz AI.
Hyz Insight (Ask, on a dashboard that reads data items or a field campaign) answers a question about the data behind it: the model plans the queries, the platform computes the numbers from the data item's Hyz Layer or the campaign's records — their review status, collector and channel included — and every answer shows the query that produced it so you can check it.
Explain this widget: hover a widget that reads a data item or a campaign and choose the sparkle. Hyz Insight then answers about that widget alone — what it shows and what stands out — with the numbers computed by the platform, never by the model, and reading the same rows the widget reads.
Save as template in the builder keeps the layout, the widgets and the fields each widget reads — and nothing about your sources: every source becomes a placeholder. Starting from a template (New dashboard → Saved templates, or Use on a catalog entry) shows one slot per source with the fields it needs; pick your data item, and missing fields can be aliased to one of its columns or that slot's widgets are left out. A public template can be picked by anyone; a private one only by you.
Visibility is private, link (anyone holding a link you can rotate to revoke) or public. On top of that you can invite people as viewer or editor, and grant a group view or edit. Viewers open and query; editors also save the layout; only the manager changes sharing, grants and deletion. See Sharing.
Export downloads the dashboard as a bundle: layout, widgets, bindings by column name, map symbology — never data. On import, every source the file names is re-mapped to something local, with a live "columns found" check so a wrong pick shows before the import, not after.
While a dashboard is open, Haiyz watches the data items and maps it depends on. When one publishes a new version every widget refetches and the dashboard says the data updated — no reload needed.