cortex-agents-tutorial

Build Snowflake Cortex AI agents by creating semantic views for structured data and Cortex Search services for unstructured documents, then using a CREATE AGENT specification to orchestrate queries across both data types.

Deploying Cortex Agents requires a Snowflake account with Cortex AI enabled, ACCOUNTADMIN role for environment setup, and a configured warehouse for execution, along with properly defined tool_resources and execution_environment blocks.

Cortex agent orchestration routes text-to-SQL queries by using an agent specification that directs structured retail sales questions to Cortex Analyst and unstructured product documentation queries to Cortex Search services.

Test and parse Cortex Agent responses by executing DATA_AGENT_RUN and using Python examples to surface the agent's thinking process, generated SQL statements, and returned tool results for verification.

Common errors when creating semantic views and Cortex Agents involve missing ACCOUNTADMIN permissions, incorrect execution_environment configurations, and improperly defined tool_resources blocks that prevent successful agent orchestration.

Yes, Snowflake Cortex Agents support retail sales and product documentation RAG scenarios by orchestrating structured retail data through semantic views and unstructured product documents through Cortex Search services.