Skip to content

Reference & Data Model

Execution (workflow_execution)

An execution is a single run of a workflow or agent: status, trigger, steps, child runs and transcript.

The execution is the sanitized projection of an engine run. It contains only metadata, status and times; inputs, outputs and credentials stay server-side. Every run – including every subagent and every workflow started by a rule – gets its own row and is assigned to its originating run via parent_run_id. (Data type: runtime data)

Page Ausführungen (Executions, /executions), reached via a workflow run in the Polaris sidebar: on the left the list with the tabs „Workflows“ and „Chats“ (newest first, live updates), on the right the selected run with header, „Children (n)“ and trace. „Export“ delivers „JSON“ or „AI digest“. Subagents and workflows started from Polaris additionally appear in its „Agenten“ (Agents) sidebar. See Guides › Executions.

Field Technical Type Required Meaning Notes
Run ID run_id Text yes engine identifier; shown truncated, can be copied System
Kind kind Selection yes „Rules“ (cel, rules workflow) / „Agent“ (agent) System; chat runs show „Chat“ (surface = chat)
Status status Selection yes see States System
Failure label failure_label Text no execution failed or ceiling reached System; never the error message itself
Trigger trigger Selection yes „Manual“, „CDC“, „Chat“ (conversation) System
Started by initiator_user_id Reference no initiating person; empty = „system“ System
Started started_at Date yes point in time (displayed e.g. „yesterday 16:04“, „14m“) System
Finished / Duration finished_at Date no duration = end − start, counting up for running runs System
Workflow definition_id Reference (text) no definition with link; empty for chat runs System
Version revision Number no deployment version, e.g. „v3“ System
Steps total steps_total Number no number of rules for rules runs; empty for agents System; progress is derived from the trace
Error rule error_node_id Text no rule at which the run broke; marked in the trace („bei Regel …“, at rule …) System
Root run / Parent run / Depth root_run_id, parent_run_id, depth Reference / Number – tree of child runs System
Origin spawn_source Selection yes external (trigger), planner (Polaris/agent), workflow (agent or workflow started by a rule) System
Activity activity Selection no active / waiting („Wartet“ (Waiting) in the Polaris sidebar) System; meaningful only while running

The trace consists of steps: node (node_id), type (node_type, e.g. cel_eval, cel_call_db, cel_spawn, llm_step), status (status), summary (summary, a sanitized one-liner), tool name (tool_name, never the arguments), start and end (started_at, finished_at), order (seq). For rules runs Aligia shows the steps per rule („Regeln“ (Rules): ran, failed, condition false, not reached). Child runs appear under „Children“. For agent steps Aligia additionally shows the output of the language model (transcript). Raw inputs and outputs of a step are visible only to the creator.

Related to Kind Required On archive On delete
Workflow definition belongs to (text, no hard reference) no kept kept
Step has many (via run_id) no – –
Execution (parent run) belongs to (via parent_run_id) no – –

Executions are neither deleted nor archived; they are a log.

Value Label (UI) Meaning
pending Pending created, not yet started
running Running in progress
completed Success finished successfully
failed Failed error; failure label execution failed
ceiling Failed step limit reached; failure label ceiling reached
cancelled Cancelled aborted

Steps: pending Pending, running Running, done Success, failed Failed, cancelled Cancelled, skipped Skipped. A hint „child failed“ appears when a child run failed but the main run shows „Success“.

  • Without a selection of your own, the page opens the most recent failed run.
  • You stop a running chat run in Polaris via „Stoppen“ (Stop); the executions page offers no cancel button.

None. Runs do not feed into any KPI.

Trigger · Conversation · Private data