Reference

Endpoint

GET /api/v1/resources/{resource_id}/data

Path parameter	Type	Description
`resource_id`	integer	The resource to read. A non-numeric value is `422`.

Headers

Header	Required	Description
`X-API-Key`	Yes	Your API key.
`workspace_id`	Yes	Numeric id of the workspace that owns the resource.

Query parameters

Parameter	Type	Default	Description
`page`	integer	`1`	One-based page number.
`page_size`	integer	`1000`	Rows per page. Minimum `1`, maximum `10000` — higher values are capped silently.
`fields`	string	(all columns)	Comma-separated column names to return.
`ordering`	string	(unordered)	One column name. Prefix with `-` for descending.
anything else	string	—	A filter. See below.

Filters

Any parameter that isn't one of the four reserved names is read as a filter on the column of that name.

Form	Operator	Example
`column=value`	Equal to	`status=RECONCILED`
`column__exact=value`	Equal to	`status__exact=RECONCILED`
`column__gt=value`	Greater than	`A.amount__gt=1000`
`column__gte=value`	Greater than or equal	`A.date__gte=2026-09-01`
`column__lt=value`	Less than	`A.amount__lt=5000`
`column__lte=value`	Less than or equal	`A.date__lte=2026-09-30`
`column__in=a,b,c`	Any of	`A.country__in=Colombia,Mexico`

Multiple filters are combined with AND. Column names are the labels shown in Simetrik and are case-sensitive. Any other operator suffix is 400.

page is one-based, and paging is offset-based: page=3 skips the first 2 × page_size rows. Read pages until returned_count is lower than meta.page_size — that's the end of the result set, and it's a more robust test than comparing against total_pages.

Always send ordering when you read more than one page. Without an explicit sort, the order rows come back in isn't guaranteed to be the same between requests, so a row can be skipped or returned twice as you walk the pages. Sort by a column with unique values — an id is ideal. If rows are being written while you read, also filter to a closed window (a date range that's already over) so new rows stay out of the result set. That doesn't freeze it — a row that arrives late with a date inside the window can still shift the pages — so when completeness matters, de-duplicate on the unique column you sort by and treat total_count as a sanity check, not proof: a repeated row and a missed one can cancel out. If the counts disagree, re-read the window.

Caching

Identical requests are served from a cache for five minutes. Identical means the same workspace, resource, page, page size, fields, ordering and filters — change any one of them and it's a different entry.

So data can be up to five minutes stale, and polling faster than that returns the same response, including the same total_count. Cache entries are scoped per workspace and ownership is checked before the cache is read, so a cached response never crosses workspaces.

Response

200 OK:

{
  "data": [
    {"A.id": "TX-0001", "A.amount": 1500, "status": "RECONCILED"}
  ],
  "meta": {
    "page": 1,
    "page_size": 1000,
    "returned_count": 1,
    "total_count": 1,
    "total_pages": 1
  }
}

Field	Type	Description
`data`	array	The rows. Each object is keyed by column name. Empty when nothing matched.
`meta.page`	integer	Page returned.
`meta.page_size`	integer	Page size actually applied, after capping.
`meta.returned_count`	integer	Rows in this response.
`meta.total_count`	integer	Rows matching the filters across the resource.
`meta.total_pages`	integer	`total_count` ÷ `page_size`, rounded up.

Responses larger than about 1 KB are gzip-compressed.

Errors

Status	`detail`	Cause
`400`	`workspace_id must be a valid integer`	The header isn't numeric.
`400`	`Filter field '<name>' is not a valid column for this resource.`	Unknown column in a filter — often a misspelled reserved parameter.
`400`	`Field '<name>' is not a valid column for this resource.`	Unknown column in `fields`.
`400`	`Ordering field '<name>' is not a valid column for this resource.`	Unknown column in `ordering`.
`400`	`Operator '<op>' is not supported.`	Unknown `__suffix`.
`400`	`The requested filters or ordering could not be applied.`	The query was rejected downstream — usually a value whose type doesn't match the column.
`401`	`Could not validate credentials`	Missing or invalid `X-API-Key`.
`403`	`Resource does not belong to the specified workspace`	Ownership mismatch.
`404`	`Workspace not found`	No workspace with that id.
`404`	`Resource not found`	No resource with that id.
`422`	Validation error object	Missing `workspace_id` header, `page_size` below `1`, or a non-numeric `resource_id`. Headers and declared parameters are validated first, so a `workspace_id` sent in the query string fails here as a missing header rather than being read as a filter.
`503`	`Data platform is temporarily unavailable. Please try again later.`	Transient. Retry with backoff, and with a smaller `page_size` if it repeats.

Errors carry a detail field. 422 follows the standard validation-error shape, with detail as an array identifying the offending parameter.

Health

GET /health

Returns {"status": "ok"}. No authentication, and no check of the data platform behind it.

OpenAPI

The service publishes its own OpenAPI document at /api/v1/openapi.json, which is handy for generating a client.