Explore public-data API connections

Unity includes API connections for St. Louis Fed (FRED), World Bank, and the US Census Bureau. Use them to discover supported economic, development, and demographic data from the report workspace.

These connections fetch data from external providers. They are separate from the personal Unity API key used to call Unity from your own application.

Add a public-data connection

1. Open the data connection setup.
2. Choose API under Select Database Type.
3. Select the provider.
4. For FRED, enter the required FRED API Key.
5. Save and attach the connection to your report.
6. Make it active and use Data Chat to discover and analyze a topic.

The World Bank and US Census connection forms do not require an API key. Provider availability, supported datasets, and request constraints still affect the results.

Choose a provider

Topic lists are starting points, not a guarantee that every subject maps to one uniquely defined series. Ask the assistant to discover the exact series, indicator, dataset, and geography.

Discover before analyzing

Using the active FRED connection, find the series that matches the
unemployment measure I need. Explain its definition, units, frequency,
and date coverage before fetching observations and building a trend.

For country comparisons:

Using the World Bank connection, find a suitable population indicator.
Compare the requested countries over their common available years.
Explain missing values and the indicator definition.

For Census data:

Using the US Census connection, identify the correct dataset, vintage,
geography, and population variable. Inspect sample rows before treating
any coded time field as a calendar year.

State the countries, regions, time span, and measure you intend. A source's metadata can be different from its observations endpoint.

Understand provider queries

The API connection path uses a structured provider request descriptor rather than ordinary database SQL. It identifies the provider, endpoint, and request parameters.

Supported processing can then use DuckDB SQL over the returned api_data table to reshape or filter observed columns. For example, multiple series can be aligned by date and pivoted into the numeric columns expected by a visual.

The assistant should inspect returned fields before writing the transformation. It should not invent a database schema or assume a provider's coded field means a calendar year.

FRED credentials are supplied server-side. They do not belong in chat examples or a saved query's visible descriptor.

Inspect a result

Review the selected series/indicator, units, frequency, geography, time coverage, missing values, request descriptor, transformation, and output mapping.

Source values can be updated or revised. A previous chat answer retains its earlier result, so use a fresh request when you need the latest available observations.

Multiple provider requests do not automatically create a cross-source SQL join with your database. Explicit preparation is needed when one calculation combines differently shaped sources.

Use the supported authoring path

API connections have their own provider-query path and topic browsing. The HTML dataset contract described elsewhere in this catalogue expects supported model-backed read-only SQL.

Do not assume that a provider descriptor can be pasted into an HTML SQL dataset or SQL Playground. Inspect the available execution path for the intended visual and source.

Troubleshoot provider data

For your own warehouse, use Database connections. For local tabular data, use CSV imports.