Usage Guide
This guide is organized as a practical user journey so you can move from foundational checks to advanced rule logic in a predictable order.
Start with single-instance and graph-level validation to confirm structural correctness, then expand into uniqueness and time-series quality, and finally use conditional and chained conditional logic for derived operational outcomes and multi-step reasoning.
You can follow the journey end-to-end when onboarding, or jump directly to the section that matches your current use case.
Rule management model used across this guide:
- Data Product + RuleSet is the primary way to manage and persist rules.
- YAML is the CDF Toolkit representation of Data Product and RuleSet bindings, plus deployment/runtime configuration.
- Direct TTL files are a legacy transition path while RuleSet rollout is completed.
This Usage guide follows one validation journey:
- Single instance data quality validation - Expected outcome: per-instance pass/fail findings for required fields and basic constraints.
- Graph consistency - Expected outcome: detected relationship and model-consistency issues across connected instances.
- Uniqueness - Expected outcome: duplicate-value detection with actionable violation records for cleanup and rerun.
- Time Series - Expected outcome: signal-quality findings for gaps, stale data, range violations, and temporal anomalies.
- Conditional logic - Expected outcome: derived rule outputs (
RuleEngineResult) for operational classification and automation. - Chained conditional logic - Expected outcome: multi-stage inferred outcomes with lineage (
causedBy) across dependent rules.
Start with context
Validation journey
- Single instance data quality validation
- Graph consistency
- Uniqueness
- Time Series
- Conditional logic
- Chained conditional logic