Snowflake Cortex

Connect Snowflake Cortex to the full platform so agent answers are traceable, semantic definitions stay accurate, and new business questions can be modeled faster.

Agentic Analytics

Why Snowflake Cortex matters

Cortex lets anyone ask natural-language questions against Snowflake data. But answer quality is bounded by the accuracy of the semantic definitions it relies on and the completeness of what they model. Stale definitions produce wrong answers with full confidence. Missing definitions mean the agent simply cannot answer. And Cortex operates within Snowflake, blind to metric definitions in external semantic layers, business logic in BI tools, and data quality issues in upstream sources.

Typedef connects Cortex to the full platform graph, so every answer is traceable end to end, semantic drift is caught before it reaches agents, and data teams can model new metrics faster because the upstream context is already mapped.

What Typedef unlocks

Explainable agent answers

Trace any Cortex answer back through the semantic definition, warehouse tables, transformations, and source systems that produced it.

Semantic accuracy for agents

Surface when upstream changes have made the definitions Cortex relies on stale, so teams can update them before agents return wrong answers.

Faster semantic expansion

Model new metrics into the semantic layer faster because Typedef maps the upstream sources and transformations, removing the manual detective work.

Agent-to-dashboard reconciliation

Show where Cortex answers diverge from dashboards or other agents and trace both paths to the point of divergence.

Related integrations

Trace any Databricks AI/BI answer back through definitions, the lakehouse, and source systems so users can verify the reasoning.

Trace any WisdomAI answer through the Context Layer, the warehouse, and the upstream pipeline that built the data, so users can verify results end to end.