Stop AI Agents from Guessing SQL with Semantic Operator on Kubernetes
summary
Semantic Operator runs Apache Ossie semantic models as Kubernetes CRDs and turns certified metrics and dimensions into deterministic, governed SQL, demonstrated working end-to-end against StarRocks with Iceberg/Polaris metadata and DataHub enrichment. The demo shows agents producing 28 wrong answers across 90 prompts when doing text‑to‑SQL but returning consistent results when routed through the operator, which validates schemas, detects physical drift, enforces row/column/metric governance via OPA/Ranger and token exchange, and exposes MCP and REST access for agents and apps on StarRocks or Trino.
AI-written summary · about
chapters (13)
- 0:00Semantic Operator project intro
- 1:22AI agents generating SQL issues
- 3:55Apache RC for data semantics
- 5:45Semantic Operator on Kubernetes
- 8:13Demo: Kubernetes and Starox
- 10:20Custom resource with CLI
- 14:36Enrich models with DataHub
- 17:30Metrics and governance setup
- 22:55Model authoring and validation
- 23:38REST endpoint interactions
- 26:52Deterministic data querying
- 30:32MCP server for agents
- 33:27Q&A
related talks
29mLakeside AI: Bringing Models to Data with Open Lakehouses & Agents)
56mDeveloping AI Applications with SQL AI: Ground to Cloud to Fabric
44mAgentic Observability — The Path to AI Co-SRE
28mPowering Generative AI with Semantic Kernel
01h 13mBuilding a Semantic AI Layer in Snowflake: From Metrics to AI Insights
50mThe hell of SQL DBs on Kubernetes versus the promise of operators
35mSmarter Databases in Kubernetes: Applying AI for Query Optimization and Self-Tuning
27mFrom Chatbots to Agents: Orchestrating Agentic AI on Existing Enterprise Kubernetes Infrastructure.