data

Overview

The data parent type covers data-centric content including published datasets, statistical reports, dashboards, financial data, and scientific data. This type captures the metadata needed for agents to evaluate data quality, scope, and applicability.

Subtypes

Subtype	Qualified Type	Description
`dataset`	`data.dataset`	Published dataset or data download
`statistics`	`data.statistics`	Statistical summaries or reports
`report`	`data.report`	Data-driven report or analysis
`dashboard`	`data.dashboard`	Dashboard or data visualization page
`financial`	`data.financial`	Financial data, earnings, market data
`scientific`	`data.scientific`	Scientific data and measurements

type_data fields

Field	Type	Required	Description
`dataset_name`	`string`	Yes	Name of the dataset or data source
`instrument`	`string`	No	Measurement instrument or data collection method
`time_range`	`object`	No	Time period the data covers
`frequency`	`string`	No	Data update frequency (e.g., “daily”, “quarterly”)
`units`	`string`	No	Primary unit of measurement
`methodology`	`string`	No	Brief description of data collection methodology
`data_points`	`integer`	No	Number of data points or records

Example type_data

{
"type_data": {
  "dataset_name": "US Consumer Price Index (CPI-U)",
  "instrument": "Bureau of Labor Statistics survey",
  "time_range": {
    "start": "2020-01-01",
    "end": "2025-02-28"
  },
  "frequency": "monthly",
  "units": "index (1982-84=100)",
  "methodology": "Probability-based survey of approximately 94,000 prices from 23,000 retail and service establishments across 75 urban areas.",
  "data_points": 62
}
}

Time range schema

Field	Type	Description
`start`	`string` (ISO 8601 date)	Start of the time range
`end`	`string` (ISO 8601 date)	End of the time range

Notes

dataset_name should be the canonical name of the dataset, not a page title or description.
instrument describes how the data was collected, not the platform hosting it.
frequency uses human-readable terms: "real-time", "daily", "weekly", "monthly", "quarterly", "annual", "one-time".
data_points is an approximate count when exact counts are not available.
For financial data pages, entities should include the companies or instruments covered, and claims should capture key numerical assertions.
For scientific data, methodology should summarize the experimental design or measurement protocol.