Play 20
Anomaly Detection
High🔧 Skeleton
Real-time anomaly detection with streaming analysis and AI enrichment.
Events flow through Event Hub, Stream Analytics detects statistical anomalies in real time, GPT-4o enriches alerts with natural language explanations and suggested actions. Cosmos DB stores event history for trend analysis. Azure Functions trigger downstream workflows (PagerDuty, Teams, email).
Architecture Pattern
Streaming anomaly detection, event-driven, AI enrichment, alerting
Azure Services
Event HubStream AnalyticsAzure OpenAI (gpt-4o)Azure FunctionsCosmos DB
DevKit (.github Agentic OS)
- agent.md — anomaly detective persona
- instructions.md — detection logic guide
- plugins/ — event processor, anomaly scorer, alerter
TuneKit (AI Config)
- config/detection.json — detection models, sensitivity, thresholds
- config/alerts.json — alert rules, severity mapping
- config/enrichment.json — AI analysis prompts
Tuning Parameters
Detection thresholdsAlert promptsSensitivity levelsDetection windows
Estimated Cost
Dev/Test
$100–250/mo
Production
$1.2K–4K/mo