Purpose and constraints
“API data publication” determines which flow table an external consumer receives. Add it from “API: receive and publish data” and select a REST API connection with direction “Publish flow data”. Connect the required result to its input: a transformation output, selected router branch or data written by a destination.
The block ends its branch, has no outputs and does not modify table values. A flow may contain only one publication block. An endpoint belongs to one published flow; create a separate flow for another endpoint. Occupied endpoints are unavailable as new free connections.
Connection and serving mode
The connection defines the address, API version, JSON/JSON Lines/CSV/XML format, public access and page limit. Block properties show the complete address and technical reference link. The block itself sets response behavior and parameters.
“Stored results of the latest run” serves an existing table from the latest completed Production run. GET does not execute the flow; a schedule or another trigger maintains freshness. “Run the flow for every request” executes Production and waits 1–120 seconds. A reuse setting of 0–3600 seconds permits reading a completed result for an identical request. Zero disables reuse of completed results.
Filters and run parameters
Add up to 20 API parameters as needed. For each, set name, description, purpose, target field or run parameter, type and required status. Names start with a Latin letter and contain up to 40 Latin letters, digits or underscores; offset, limit and run are reserved.
A filter selects results by text, number or date. Allowed values constrain selection; multiple selections repeat the parameter name in the URL. An inclusive range uses _from and _to suffixes. All filters apply to the full dataset before pagination. A run parameter is available only in execution-on-demand mode and supplies the corresponding flow parameter.
Publication, access and verification
Save and publish the flow. For stored results, complete an initial run; without one, no dataset exists to read. Grant the API client “Read”; for execution on demand, check the owner's flow execution permission. Public access is enabled separately in the connection: data can then be read without a key, so expose only an approved public dataset.
Verify a control GET, filtered row counts and the next page. If execution continues, the 202 response includes next; request it after Retry-After while preserving run.id. Unknown parameters are rejected. For 404, check that a completed result exists; for 409, check publication state or data freshness. Detailed examples appear in API publication configuration.