Documentation by ClonePartner

Start typing to search.

Using Envoy

Tasks and task runs

View Markdown

Manage the Task library, start standalone Runs with bound Connectors, review versions, and inspect Run logs and metrics.

Tasks and task runs

A Task is a YAML pipeline definition; a Run is one execution of a Task, started standalone from the Task page or as part of a Job Run. These four screens — the Tasks list, Task detail, Task editor, and the Runs list — are where operators manage the Task library and review execution history.

Screen overview: Tasks list

Open Tasks in the sidebar. The page header reads Tasks ("Manage your task definitions and versions.") with a refresh button, an Import button, and a New Task button. A search box ("Search tasks...") and a Status filter (All statuses / Active / Archived) sit above the table.

Columns: a selection checkbox, Name (sortable), Description, Version (sortable), Status, and Created (sortable). Each row has inline Run and Edit icon buttons (Run is disabled for archived Tasks — "Archived tasks cannot be run") plus a ⋮ menu with Run, Edit, Duplicate, Export, Archive or Restore, and Delete permanently. Selecting rows enables a bulk Export. Deleting asks for confirmation: "This will permanently delete '<name>' and all its versions and run history. This cannot be undone."

Task detail

Clicking a Task opens its detail page: the Task name with a version badge (v3), its description, connector slot chips, and a metrics summary. The main panel, YAML Configuration, shows the Task YAML in a read-only editor with a version selector. The right column lists Recent Runs (the last eight, clickable) and Version History — selecting a version loads its YAML into the viewer.

Header actions: Run Now, Edit, and a ⋮ menu with Download YAML, Export with history, Archive/Restore, and Delete permanently.

To start a standalone run

  1. On the Task detail page, select Run Now (or the Run icon on the Tasks list, which opens the same flow).
  2. The Run Configuration dialog opens ("Configure connectors and arguments to run this task."). Under Connectors, bind a stored Connector to each slot — each select is filtered to the slot's type, and a slot is pre-filled when exactly one matching Connector exists.
  3. Under Arguments (shown when the Task declares arguments), fill each field. Required arguments are marked; defaults are pre-filled.
  4. Select Run. The button stays disabled until every slot and required argument is filled. The browser navigates to the new Run's detail page.

If the Task has no connector slots and no arguments, Run Now starts the Run immediately without a dialog.

Task editor

New Task and Edit open the editor (admin or superadmin only). The top bar shows New Task or Edit: <name>, with AI Assist, the save button (Create Task for new Tasks, Save Version when editing), and a ⋮ menu with Archive/Restore and Delete permanently.

The metadata row has Name (required; locked after creation), Description, and — when editing — Change Summary, which records what changed in this version. Below it is a full-height YAML editor.

Saving validates the YAML on the server. If validation fails, a "Validation failed" banner lists each issue with its severity, step, field, and message, and a Fix with AI button sends the issues to the AI assistant. Each successful save of an existing Task creates a new version; earlier versions stay browsable on the Task detail page, and existing Runs keep the version they executed.

This page is for maintaining the library, not learning YAML — the task YAML language itself is covered in Task anatomy and authoring workflow.

Screen overview: Runs list

Open Runs in the sidebar. The page header reads Runs ("Execution history across all tasks.") with a refresh button. A search box ("Search runs...") and a Status filter (Running, Success, Completed, Failed, Cancelled, Pending) sit above the table.

Columns: a selection checkbox, ID, Task, Status, Trigger, Version, Duration (sortable), and Started (sortable). Each row's ⋮ menu has Delete; selecting rows enables bulk Delete ("Active runs will be skipped."). Clicking a row opens the Run detail page.

Run detail

The Run page header reads Run #<id> with a status badge and a subtitle linking to the Task and its version. A metadata bar shows the trigger type, duration (ticking live while running), items processed, started → completed timestamps, and the executor that ran it. A failed Run shows its error message in an alert.

Below the metadata:

  • Arguments — the values passed to this Run, as read-only JSON with a Copy button.
  • Metrics — per-step counters, shown when the Run reported metrics.
  • The log viewer — the Run's captured output, streaming live for active Runs. See Live run monitoring.

Header actions depend on state: Stop Run while the Run is running or pending; Re-run once it is terminal, which starts a new Run with the same arguments and connector mapping; and a ⋮ menu (terminal Runs only) with Clear Logs and Delete Run.