| name | present-data |
| description | Choose the right visualization for a dataset and render it — table, chart, network graph, or globe — then place it well. Use when an agent has structured data to show, for "chart this", "show it as a table", "graph these relationships", or comparisons. Triggers: chart, graph, table, visualize, plot, compare, breakdown, show the data, trend. |
| license | MIT |
| compatibility | Requires JarvisVR agent-backend stage runtime. Widget props validated against holo-tools/registry.json. |
| metadata | {"agent":"stage-agent","category":"stage","version":"1.0","author":"jarvisvr"} |
| allowed-tools | show_data_table show_chart show_graph show_panel arrange_holograms |
Present Data
Pick the visualization that best fits the data's shape and the user's question,
build valid props, and place it comfortably.
Choosing the widget
| Data shape / intent | Widget | Tool |
|---|
| Rows × columns, look up values | data_table | show_data_table |
| Trend / comparison of series | chart_3d (line/bar) | show_chart |
| Distribution / parts of whole | chart_3d (pie/scatter) | show_chart |
| Entities + relationships | graph_3d | show_graph |
| Geographic points / flows | volumetric_globe / map_3d | show_globe / show_map |
| Short prose / KPIs | panel | show_panel |
Steps
- Inspect the data (types, cardinality) and the user's question.
- Select from the table above; prefer the simplest that answers it.
- Build props to schema (e.g.
chart_3d requires chart_type + series;
data_table requires columns + rows).
- Render, then place via
arrange_holograms (charts/tables read best
world/surface at ~1.2 m).
- Narrate the takeaway in
agent.speech — the headline, not every number.
Output
chart_3d (show_chart, props per registry.json):
{ "widget_type": "chart_3d",
"props": { "chart_type": "bar", "title": "Quarterly Revenue",
"labels": ["Q1","Q2","Q3","Q4"],
"series": [ { "name": "FY2026", "values": [14,20,15,26], "color": "#4FC3F7" } ],
"y_axis_label": "Revenue ($M)", "show_legend":
data_table (show_data_table):
{ "widget_type": "data_table",
"props": { "title": "Devices",
"columns": [ { "key": "name", "label": "Name", "type": "string" },
{ "key": "watts", "label": "Watts", "type": "number" } ],
"rows": [ { "name": "Ceiling Light", "watts": 12 }, { "name"
Edge cases
- Tiny data (1–2 numbers) → a
panel/text_label, not a chart.
- Too many series/rows → aggregate or paginate; keep it legible.
- Categorical vs continuous → bar for categories, line for time series.
- Selection events (
select_point, select_row) → drill down or explain.
- Raw source, not data → that's
research-agent; you visualize what it
returns.
