This article details how to monitor and troubleshoot Snowflake's CoWork deployments, emphasizing the fragmented nature of usage data across multiple account views. It explains that tracing a CoWork question requires joining data from `SNOWFLAKE_INTELLIGENCE_USAGE_HISTORY` and `CORTEX_AGENT_USAGE_HISTORY` using request IDs. The piece also covers setting up a dedicated monitoring warehouse, creating health dashboards, implementing automated alerts for runaway queries, and leveraging Cortex AI Guardrails for prompt injection detection. AI
IMPACT Provides operational guidance for managing AI deployments, focusing on cost and performance monitoring.
RANK_REASON Article provides operational guidance and best practices for using an existing product (Snowflake CoWork), rather than announcing a new product or frontier model.
- AI_CLASSIFY
- AI_EXTRACT
- CORTEX_AGENT_USAGE_HISTORY
- CORTEX_AI_FUNCTIONS_USAGE_HISTORY
- Cortex AI Guardrails
- CORTEX_AI_GUARDRAILS_USAGE_HISTORY
- PARENT_REQUEST_ID
- request.id
- Snowflake
- SNOWFLAKE_INTELLIGENCE_USAGE_HISTORY
- Snowsight
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
