Pipeline Failure Auto-Fix Feature Generator

Available since: v1.0.61

The Pipeline Failure Auto-Fix Feature Generator is an event-driven workflow that monitors recurring pipeline failures and automatically creates high-priority remediation features when a failure crosses a critical recurrence threshold. This closes the loop between failure detection and autonomous repair — no human triage required.

How It Works

The workflow watches the pipelineFailures table and evaluates each record against three conditions:

resolved          = false
occurrenceCount   >= 3
autoFixFeatureId  IS NULL

When all three conditions are met, the system:

Creates a new feature in the features table with:
- source = 'pipeline_failure'
- priority = 'high'
Writes the new feature's ID back to the pipelineFailures record as autoFixFeatureId, preventing duplicate features from being generated on subsequent evaluations.

Trigger Modes

Event-Driven (Real-Time)

The workflow fires immediately when a pipelineFailures record is inserted or updated and occurrenceCount crosses the threshold of 3. This ensures the platform responds to persistent failures as soon as they are detected.

Daily Sweep (Scheduled)

A daily background sweep re-evaluates all pipelineFailures records where resolved = false and autoFixFeatureId IS NULL. This acts as a safety net for:

Records that were updated during platform downtime
Historical failures that pre-date the workflow
Any edge cases the real-time trigger may have missed

Feature Attributes

Auto-generated features are distinguishable from manually created or AI-discovered features by their source field:

Field	Value	Description
`source`	`pipeline_failure`	Identifies the feature as auto-generated from a failure record
`priority`	`high`	Ensures the fix is prioritized above normal backlog items
`autoFixFeatureId`	(UUID)	The ID written back to the originating failure record

Deduplication

The autoFixFeatureId foreign key on pipelineFailures acts as a deduplication guard. Once a failure record has a non-null autoFixFeatureId, it is excluded from all future evaluations — both real-time and scheduled. This means:

One feature per failure record — no duplicate remediation tasks
Safe to re-run — the daily sweep is idempotent

Workflow Summary

pipelineFailures insert/update
        │
        ▼
 occurrenceCount >= 3?
        │ YES
        ▼
  resolved = false?
        │ YES
        ▼
autoFixFeatureId IS NULL?
        │ YES
        ▼
  Create feature (source='pipeline_failure', priority='high')
        │
        ▼
  Write autoFixFeatureId back to pipelineFailures record
        │
        ▼
       DONE

Related Entities

pipelineFailures — Source table. The workflow reads resolved, occurrenceCount, and autoFixFeatureId; it writes autoFixFeatureId upon feature creation.
features — Destination table. Auto-fix features are inserted here and enter the standard development pipeline.