Dashboards

A dashboard is a collection of widgets arranged in a 12-column grid. Each widget queries a specific event and displays a visualization.

Concept map

Dashboards sit between the raw event model and the browser runtime: they define the surface that later gets published, embedded, and scoped.

Data Events Widgets only become useful when the underlying event names, dimensions, and measures are coherent. Browser Embed Runtime Once a dashboard is published, the runtime mounts and controls that surface in the host page. Isolation Tenants The same published dashboard can be reused across many customers while the session keeps data scoped. Surface Dashboards How widgets, filters, draft state, and the published embed surface fit together. Control Permissions Permissions decide what stays visible or interactive once the published dashboard is embedded.

Surface lifecycle

A dashboard goes through three practical stages before a customer sees it inside your product.

Step 1 Draft Build widgets, change filters, and iterate in the builder. This version is only visible in preview and is not served to customer embeds.

Step 2 Publish Freeze the current draft into the live snapshot. This is the moment the dashboard becomes the customer-facing surface your signed sessions can mount.

Step 3 Embed Mint a tenant-scoped signed session and reuse the same published dashboard across customers. The browser runtime then adds filters, period changes, drill, and resize behavior.

Validation path: open the live demo to see the published surface as a customer would. Then read Embed Runtime for the browser contract and API Reference for /v1/embed-sessions.

Widget types

Kind	Charts	Use case
`kpi`	stat, progress	Single number or quota-style progress card
`timeseries`	bar, area, line, stacked_bar	Trend or composition over time
`breakdown`	list, horizontal_bar, donut	Top-N, mix, or share by dimension
`table`	table	Detailed rows

Metrics

count — number of events
count_unique — distinct user_id count
count_unique_approx — approximate distinct user_id count via ClickHouse HLL, with HLL rollup merges on safe hourly/daily paths
sum — sum of a numeric property
avg — average of a numeric property
p95 / p99 — percentile metrics for numeric properties like latency or duration

Filters

Dashboard config can include interactive filters:

Period selector: 1h, 24h, 7d, 30d, 90d
Dimension filters: single-select or multi-select on string properties
Numeric range: min/max on numeric properties
Date range: custom from/to dates

For common SaaS dimensions like model, endpoint, status, plan, and source, safe count trend and grouped widgets can stay on filtered hourly/daily rollups instead of always falling back to raw events.

Draft vs Published

Dashboards have two states:

Draft: Editable in the builder. Changes are visible only in preview and do not affect the customer-facing embed.
Published: The live snapshot shown in embed iframes. Click "Publish" in the builder to replace the currently served snapshot with the latest draft.

If you unpublish a dashboard, the live embed goes offline. Emban keeps the last published snapshot so you can restore it into the draft later, but it is not served while the dashboard is in draft.

Auto-create

Emban can scan your events and automatically generate a dashboard. Use the "Auto-create" button in the builder or call:

POST /v1/discover/auto-create
Authorization: Bearer YOUR_ADMIN_API_KEY

This creates a draft dashboard with KPIs, timeseries, share, and ranking widgets based on your event schema.