FXMacroData for dbt

Build warehouse models over FXMacroData data loaded by your ingestion pipeline. This native dbt package provides a lossless staging model, numeric and timed-record views, a discoverable public operation catalog, and data tests for row counts, response envelopes and unknown publication times.

Explore FXMacroData

Add the package

Use dbt Core 1.11.15 or later with sqlparse>=0.6.0,<0.7.0. The reference runtime is dbt Core 1.11.15, dbt-duckdb 1.11.0 and DuckDB 1.4.3. Install your warehouse adapter using its normal setup instructions before adding this package.

Add the package to your project's packages.yml, pinned to a release tag:

packages:
  - git: https://github.com/fxmacrodata/dbt-fxmacrodata.git
    revision: v0.1.0

Alternatively, extract this package beside your dbt project and reference it with - local: ../dbt-fxmacrodata.

Configure the relation containing your ingested records in your project's dbt_project.yml:

vars:
  fxmacrodata_schema: macro_data
  fxmacrodata_table: fxmacrodata_records
  # fxmacrodata_database defaults to target.database

Then install and build:

dbt deps
dbt build --select package:fxmacrodata

The package uses your existing dbt warehouse profile. It does not need an FXMacroData API credential or call the service from SQL. Load records first with a customer-side FXMacroData connector, using the shared public-contract row schema described below. Keep API keys in the ingestion platform's secret manager and warehouse credentials in your normal dbt profile or environment.

Models

Resource	Contents
stg_fxmacrodata	Every ingested record, including null-record envelope rows.
fxmacrodata_numeric_records	Records with a numeric value.
fxmacrodata_timed_records	Records with an original announcement_datetime.
fxmacrodata_operation_catalog	Public REST operations and packaged MCP tools, with JSON input schemas.

The source relation needs these columns: operation, source_url, record_id, currency, indicator, date, announcement_datetime, value, value_text, record_json and response_json. value is nullable numeric; the remaining columns are nullable strings. Identity, operation, provenance and JSON envelope columns must be present. Original JSON fields are never parsed away or replaced.

record_id is content identity, so repeated snapshots can share an identity. The package preserves the rows you loaded and does not silently deduplicate them. date remains the original observation date. Unknown publication times stay null; calendar schedules and period dates are not converted into observed release times. The operation catalog describes capabilities, not which operations you have loaded or which your API plan permits. MCP tool names use an mcp_ prefix.

Use the models in your project

Add a model that attaches release context to your own positions:

select
    positions.currency,
    positions.position_usd,
    events.indicator,
    events.announcement_datetime,
    events.record_json
from {{ ref('positions') }} positions
join {{ ref('fxmacrodata_timed_records') }} events
  on positions.currency = events.currency
where events.operation = 'release_calendar'

Choose one operation without rewriting the standard projection:

{{ fxmacrodata.fxmacrodata_records(operation='indicator_history') }}

Explore available operation arguments in the seeded catalog:

select name, description, input_schema_json
from {{ ref('fxmacrodata_operation_catalog') }}
order by name

The package uses dbt's native cross-database type macros and standard SQL. DuckDB is the reference validation adapter. Validate the package with your warehouse adapter before scheduling it; output materialization and source relation names remain configurable through normal dbt settings. The package does not download history, infer announcements, or create data-provider infrastructure.