AI SqlGuard — Live Preview

Will this AI-generated SQL survive production?

Deterministic checks for sensitive columns, destructive statements, policy violations, and column-level lineage — evaluated against a sample catalog before the SQL reaches your database. The page calls the same-origin API /api/sqlguard/report.

What this preview is — and what it is not

What it is

A live, browser-only demo of the sqlguard semantic-governance engine.
Real parser (GSP TGSqlParser) + catalog binder + 9 L6 policy rules + column-level lineage.
Sample catalogs (ecommerce, broker, hr-analytics, finance) plus inline mode for your own JSON schema.
Stable snake_case JSON contract (the response shape your integration would consume).

What it is not

Not a production governance gate. SQL is never executed; no database connection happens.
Not connected to a real catalog. Sample tags are illustrative; bring your schema via the inline-catalog demo.
Not an LLM-quality evaluator, SQL rewriter, IAM bridge, or approval workflow.
Phase 1 Evaluation Kit deliverables (CLI, batch corpus runner, Readiness Score, full report) ship with the design-partner package — not the live preview.

Evaluating for a real workflow? Send us 50–100 anonymized SQL samples with your expected decisions; we'll send back a Readiness Report against your corpus.

Try a rule

Picking a demo replaces the SQL and option fields below; click Run to see the rule fire.

SQL

Dialect Sample catalog Policy preset

Tenant ID (optional — enables MISSING_TENANT_FILTER) Question (optional — enables AGGREGATION_GRAIN_MISMATCH)

Ready.

Need this inside your own network?

Request a full AI SqlGuard assessment

For regulated environments, the readiness engine runs locally as a parser, catalog, policy, and lineage layer before generated SQL is executed.

Request on-premise readiness assessment