aibi-dashboard-guardrails

What problem does it solve?

It prevents repeated, high-impact mistakes when generating or updating Databricks AI/BI dashboards from semantic YAML, schemas, and user requirements.

Core Features & Use Cases

• Prompt-vs-YAML conflict enforcement: Treat the user prompt as the source of truth and explicitly flag any contradictions before changes are made.
• KPI and SQL validation: Require KPI formulas to be validated with explicit SQL prior to publishing updates.
• Display vs raw dataset correctness: Ensure presentation-ready (formatted/rounded) tables are only used when the user expects display-shaped output, not raw calculation datasets.
• Ambiguity handling: Clarify ambiguous measure names (gross/net, booked/realized, filtered/raw, period/lifetime) so the dashboard logic matches intended business meaning.
• Cross-check and pre-publish discipline: When users provide expected values, cross-check rendered results before publishing; otherwise run a structured pre-publish checklist.

Quick Start

Use aibi-dashboard-guardrails to verify that your proposed dashboard changes (KPI formulas, formatting, and datasets) align with the prompt, validate with SQL, and pass any user-provided expected-value cross-checks.