Skip to main content
The FXMacroData source reads economic releases and supporting market data using an HTTP manifest. It targets API version v1 and is alpha. Public announcement, forecast coverage, and FX source responses have been checked live; authenticated reads and paid resources have not yet been tested against an account.

​
Configuration

Use an API key with access to the selected currencies and data families. Public USD availability does not establish access to paid resources. See the FXMacroData API reference for authentication, subscription access, and supported currency and indicator values. The data_catalogue resource exposes indicator availability for your currency. Use separate pipelines and destination tables for different currencies, indicators, or pairs. Some resources use dates as keys within the configured series; those keys do not identify observations across multiple series.

​
Resources

Announcements, the data catalogue, latest announcements, and the release calendar are selected by default. Other resources are optional and may require additional subscription access. Only selected resources are written. Historical resources read the configured date window. Predictions contain one row per announcement with its predictions nested as JSON. Native official forecasts are read separately by official_forecasts, which follows every publisher in official_forecast_sources for the configured currency. That source list is fetched even if it is not selected for output. The indicator endpoint’s capped official_forecasts section remains disabled in favor of the paginated native resource. official_forecast_calendar reads verified publication schedules for those publishers. Native forecasts are not restricted by the pipeline date window. Press releases read the available feed without date bounds. Intraday FX reads the configured date window, from midnight UTC on start_date through the end of end_date. Curves return the latest available nodes up to end_date, subject to the API’s lookback rules. announcement_revisions requests all revisions and retains the nested revision data. The data catalogue, global forecast coverage, latest announcements, latest commodities, market sessions, and curve views each produce a single snapshot row. Nested maps and arrays remain JSON. Other resources emit one row per item in their response array. The raw column retains unmapped fields from each row; it does not retain the surrounding response envelope for array resources. research_panel returns one snapshot containing values, series metadata, completeness, quality, and dataset version. Supply the complete request as the research_panel_request string, for example:
The API accepts 1–6 series and 1–10,000 decision times per request. These dates come from the JSON body, independently of the pipeline date fields. The body also accepts a known dataset_version; it is sent unchanged. See the point-in-time guide.

​
Modes

All REST resources support full reads. Historical dates describe observations, not when they were last revised, so incremental reads are not supported. Rereading the chosen date window refreshes revisions within that window. Use full replace for resources without a primary key and to remove rows that are no longer returned. Keyed resources also support full upsert. The release_changes resource reads the currently retained event window on each run; it does not maintain a persistent change-feed cursor or provide a complete historical event archive.

​
Behavior

  • Auth: requests use X-API-Key against https://api.fxmacrodata.com.
  • Pagination: historical lists follow pagination.next_offset while pagination.has_more is true, using 100 rows per request. Press releases use 50. Predictions follow next_cursor through before_date; release changes follow next_cursor through since. Native forecasts follow official_forecast_next_cursor through before_id for each publisher. Snapshot and unpaginated endpoints make one request per read.
  • Rate limiting: one request per second per connection. The API’s account limits also apply across other clients. Filament honors Retry-After when throttled but does not track monthly usage.
  • Revisions and consistency: announcements request the latest revision. The connector does not capture a first-page dataset_version and reuse it across subsequent pages. A changing dataset can therefore shift offset pages during a read; it is not a pinned point-in-time snapshot.
  • Date-window fallbacks: the API can return a latest available observation outside the requested window. Such rows are retained, including any row-level fallback flags in raw.
  • FX semantics: reference rates and derived crosses are not executable quotes. Consult their source and derivation fields before using them.
All 26 non-streaming data paths in the production API index are represented. SSE is excluded because the existing engine does not reconnect or persist Last-Event-ID for replay. Health and ping are service probes rather than data resources. Endpoint coverage does not mean every parameter combination is read: the connector does not automatically traverse every currency, indicator, pair, or series variant. Outside native forecast publisher discovery, it uses the API’s default source and series choices. Automatic dataset pinning, durable stream replay, and authenticated verification remain outside this implementation. Native forecasts, forecast calendars, and research panels have been implemented from the official contracts but still need credentialed live validation. A public native-forecast request returned an API-key-required error.