views 'Use Cases / Backend' {
    dynamic view anomaly_detection {
        title 'Anomaly Detection (Single-Tenant)'
        description '
        Detecting anomalies for a single store.
        AI agent queries Redshift tenant database serving views directly.
        '
        
        kdsai.aibackend -> kdsai.aibackend 'Scheduler triggers anomaly detection process'
        kdsai.aibackend -> kdsai.operationalDb 'Get Monitor Spec'
        kdsai.aibackend <- kdsai.operationalDb 'Returns monitor spec'

        parallel {
            kdsai.aibackend -> kdsai.bedrock 'Agent orchestrates anomaly detection process using LLMs'
            kdsai.aibackend <- kdsai.bedrock 'Tool call request'
            kdsai.aibackend -> kdsai.warehouseDb 'Execute SQL in the selected tenant database serving views'
            kdsai.aibackend <- kdsai.warehouseDb 'Returns tenant-scoped data for anomaly detection'
            kdsai.aibackend -> kdsai.bedrock 'Send data to LLM for anomaly detection'
            kdsai.aibackend <- kdsai.bedrock 'Returns anomaly detection results'
        }

        kdsai.aibackend -> kdsai.operationalDb 'Store anomaly detection results'
        kdsai.aibackend -> user 'Alert notifications if anomalies detected'

        user -> kdsai.frontend 'Checks dashboard for anomaly detection results'
        kdsai.frontend -> kdsai.aibackend 'Polls for anomaly detection results to display on dashboard'
        kdsai.frontend <- kdsai.aibackend 'Returns anomaly detection results for dashboard'
        user <- kdsai.frontend 'Views anomaly detection results on dashboard'

        // This is required to have user at the start of the flow diagram
        include user, kdsai.frontend
    }

    dynamic view anomaly_detection_crosstenant {
        title 'Anomaly Detection (Cross-Tenant)'
        description '
        Detecting anomalies across multiple stores (chain/district manager scenario).
        AI agent queries Athena over curated S3 Parquet instead of a single Redshift database.
        Platform injects WHERE tenant_id IN (...) based on user_tenant_access.
        '

        kdsai.aibackend -> kdsai.aibackend 'Scheduler triggers cross-tenant anomaly detection'
        kdsai.aibackend -> kdsai.operationalDb 'Get Monitor Spec and resolve user tenant access'
        kdsai.aibackend <- kdsai.operationalDb 'Returns monitor spec and authorized tenant list'

        parallel {
            kdsai.aibackend -> kdsai.bedrock 'Agent orchestrates cross-tenant anomaly detection using LLMs'
            kdsai.aibackend <- kdsai.bedrock 'Tool call request'
            kdsai.aibackend -> kdsai.athena 'Execute SQL with platform-injected tenant_id filter'
            kdsai.athena -> kdsai.glueCatalog 'Resolves table schemas and partitions'
            kdsai.athena -> kdsai.etl.curatedStorage 'Reads curated Parquet from S3'
            kdsai.aibackend <- kdsai.athena 'Returns cross-tenant data for anomaly detection'
            kdsai.aibackend -> kdsai.bedrock 'Send cross-tenant data to LLM for anomaly detection'
            kdsai.aibackend <- kdsai.bedrock 'Returns anomaly detection results'
        }

        kdsai.aibackend -> kdsai.operationalDb 'Store anomaly detection results'
        kdsai.aibackend -> user 'Alert notifications if anomalies detected across stores'

        user -> kdsai.frontend 'Checks dashboard for cross-store anomaly results'
        kdsai.frontend -> kdsai.aibackend 'Polls for cross-store anomaly results'
        kdsai.frontend <- kdsai.aibackend 'Returns cross-store anomaly results for dashboard'
        user <- kdsai.frontend 'Views cross-store anomaly results on dashboard'

        include user, kdsai.frontend
    }
}