Snowflake semantic view metadata reference

Omni reads a semantic view’s definition and writes it into the topic it generates. This page lists what each part of the definition becomes in Omni.

What Snowflake metadata becomes in Omni

From Snowflake	In Omni
`COMMENT` on the semantic view, a logical table, or a field	`description` on the topic, view, or field
`SEMANTIC_VIEWS.AI_QUESTION_CATEGORIZATION`	Topic `ai_context`, prefixed `Use these instructions for question categorization:`
`SEMANTIC_VIEWS.AI_SQL_GENERATION`	Topic `ai_context`, prefixed `Use these instructions for SQL generation:`
`WITH SYNONYMS` on a dimension or metric	`synonyms` on the field
`DIMENSIONS` and `FACTS`	Dimensions on the logical table’s view
`METRICS`	Measures on the logical table’s view
`LABELS = (FILTER)`	A filter-only field named `<field>__filter` that users can apply in workbooks. It uses `bind_to` to apply the filter through a hidden dimension holding the fact’s expression
`PRIMARY KEY`	`primary_key: true` on the key dimension, or `custom_compound_primary_key_sql` for a compound key
Key columns not declared as dimensions	Hidden dimensions, so keys and joins resolve
`RELATIONSHIPS`	Relationships from the foreign-key table to the referenced table
Logical table defined by an inline SQL query	A SQL view with the declared dimensions and metrics