Reports & Dashboards

Reports give you visual, interactive dashboards backed by your data in BigQuery. Each report combines KPI cards, charts, tables, and AI-generated insights into a single view — no spreadsheets required.

How It Works

1. A dashboard definition connects to a BigQuery view in your data provider
2. Filters (date range, category selects) control which data is displayed
3. Layout sections render the data as cards, charts, or tables
4. AI insights analyze the filtered data and surface trends automatically

Reports are scoped to your organization — you only see data that belongs to you.

Dashboard Components

KPI Cards

Summary metrics displayed as large numbers with labels. Each card applies an aggregation (sum, average, count, max, min) to a column in your data.

Cards support color coding and number formatting (plain numbers, percentages, currency).

Stacked Bar Charts

Interactive bar charts that break down a metric by category. Click any bar segment to filter the data table below — the chart drives drill-down navigation.

Charts support horizontal and vertical orientations, custom series colors, and pivot columns for grouping.

Data Tables

Sortable, filterable tables with badge-based highlighting. Define threshold rules to color-code cells — for example, show values below target in red and above target in green.

Tables support:

• Column visibility — show or hide columns per breakpoint (mobile, tablet, desktop)
• Badge thresholds — color cells based on value ranges
• Drill-down links — click a row to navigate to a related detail report with filters pre-applied
• Special row styling — highlight summary or total rows

Insight Boxes

Static text cards with colored borders. Use these for fixed context, definitions, or callouts that don't change with the data.

AI-Powered Insights

The most powerful feature of reports is playbook-driven AI analysis. Instead of reading raw numbers, you get written insights that explain what the data means.

How AI Insights Work

1. A report section is configured with a playbook name
2. When you open the report (or click Generate), the current filtered data is sent to the playbook
3. The playbook analyzes the data — it can use knowledge files, multi-step reasoning, and custom prompts
4. The AI returns numbered insights explaining trends, anomalies, and recommendations
5. Results are cached by data content — the same data produces the same insights without re-running

Caching

Insights are cached based on a hash of the underlying data. When your filters change, the insight shows a stale indicator and offers to regenerate. If the data hasn't changed, the cached result loads instantly.

Configuration

AI insight sections reference a playbook by name. The playbook receives:

Input	Description
$rows	The current filtered dataset as JSON
$title	The dashboard title for context
$filters	Active filter values

This means you can swap or improve the analysis playbook without changing the dashboard layout.

Filters

Reports support interactive filters that control all sections on the page:

Filter Type	Description
Date range	Start and end date picker
Select	Dropdown with values from the data

Filter state is reflected in the URL — share a filtered view by copying the link. Browser back/forward navigation preserves filter state.

Access Control

• System dashboards are visible to all organizations (created by administrators)
• Organization dashboards are scoped to a single organization
• Data queries are always filtered to the authenticated organization's data provider
• BigQuery view names are validated against an allowlist — arbitrary queries are not possible

Data Sources

Reports query BigQuery views through your organization's configured data provider. The provider must be active and the view name must be in the allowed list. This ensures reports can only access pre-approved datasets.