Update a rule
A new RuleVersion is minted automatically whenever rule_text, window_member_scope,
or period boundaries change. Editing name or alert_message updates the rule in place
and does not mint a version. When a new version is minted, evaluation against
historical targets is queued automatically; backfill_enqueued reflects this.
Path Parameters
Response
Rule updated
A rule on a project's project-scope rule set. Identity is stable; rule_text lives on the latest RuleVersion, while alert_message lives on the rule itself.
Versioned action reach. Requires a windowed aggregate unless none. none acts on the firing data point alone; all reaches same-slug data points in the fired windows; repeats keeps that window boundary and requires a predecessor with the same value, preserving the reading that opens the run. Existing repeats versions retain this behavior. run requires a flat-window condition for the subject and extends the action in both directions through the same original value, including the opener and readings outside the detection window. A changed value, conflicting values at one timestamp, the rule's effective-period boundary, or a gap that neither touching spans nor the event type's cadence connects ends the run. A reading another rule corrected is judged by its measured value: the walk passes through it and it is not attributed to this rule. A single recorded zero or missing reading outside the rule's scope can be crossed only when the same value resumes immediately; that dropout is not itself attributed. Selecting run creates a new version. Scope-only changes can recover original values through consecutive earlier versions in the same period with unchanged condition, action, and method chain. Manual edits, reverts, or another rule's corrections reset that evidence. The reference population retains repeats-based window exclusions independently of the expanded repair reach, so selecting run alone does not widen reference exclusions. No method may draw its replacement from the run it is replacing. Attributed result context records the firing data point and detection-window end; run attribution also records the run bounds. Dashboard and API v1 authoring are supported; MCP authoring parity is a follow-up.
none, all, repeats, run automatic rules are evaluated as data arrives and over history when the rule is saved. on_demand rules run only when a run is started over a period (POST rule_runs), so a substitute rule may read DATA_CAPTURE_RATE or TIME_NOT_CAPTURED of a data point other than its subject. An on_demand model rule is not evaluated when its model run finishes or is recomputed.
automatic, on_demand