Skip to main content

Windows

🚧 Coming Soon — This API is under active design. Endpoints return 501 Not Implemented until released.

Processing windows control how the platform accumulates and flushes records before publishing to Kafka. Instead of processing every record individually, windows let you batch records by time, count, or external trigger — enabling micro-batch and aggregation patterns at scale.

Window Types​

TypeDescription
TUMBLINGFixed-size non-overlapping windows (e.g. every 5 minutes)
SLIDINGOverlapping windows that advance by a step interval
SESSIONWindows closed by inactivity gap
TIME_BASEDCron-scheduled flush (e.g. daily at 2 AM)
COUNT_BASEDFlush when N records accumulate
EVENT_TRIGGEREDExternal signal triggers the flush

Planned Workflow​

  1. Create a window: POST /api/v1/windows
  2. Reference the window ID in a transformation job
  3. Records accumulate until the window condition is met
  4. Window flushes — records are published to Kafka as a batch
  5. Monitor window state and late-arrival handling

Operations​

📄️ List window instances

🚧 *Coming Soon* — Returns all active and recently closed window instances across all profiles.

📄️ Get a window instance

🚧 *Coming Soon* — Retrieve full details of a single window instance including event counts and status.

📄️ Force-close an open window

🚧 *Coming Soon* — Immediately close an open window and trigger its configured actions, regardless of whether the normal close condition has been met.

📄️ Reprocess a closed window

🚧 *Coming Soon* — Re-run all actions for a previously closed window. Useful for replaying after a downstream failure.

📄️ List events collected in a window

🚧 *Coming Soon* — Returns a paginated list of raw events that were collected into this window instance.

Schema References​