Skip to main content

Deployment Overview

Anomstack offers flexible deployment options to fit different organizational needs and infrastructure requirements. This page helps you choose the right deployment pattern for your use case.

Deployment Modes​

🎯 Full Stack Deployment​

Deploy the complete Anomstack platform including dashboard, backend services, and database infrastructure.

✅ Best for:

  • New implementations
  • Teams wanting the full Anomstack experience
  • Organizations needing the dashboard interface
  • Proof of concepts and demos

📦 Includes:

  • Interactive dashboard for metrics visualization
  • Dagster UI for pipeline management
  • Complete alerting system (Email, Slack, LLM)
  • Built-in storage for metrics and models

🤖 Headless Deployment​

Deploy only the Dagster orchestration engine to integrate with your existing infrastructure.

✅ Best for:

  • Organizations with existing analytics infrastructure
  • Enterprise environments with strict data governance
  • Teams preferring their current dashboards/alerting
  • Microservices architectures

📦 Includes:

  • Anomaly detection pipeline only
  • Writes results to your existing database
  • Configurable alert outputs (database, webhooks, etc.)
  • Model storage in your preferred system

Deployment Platforms​

Local Development​

Perfect for development, testing, and small-scale deployments.

MethodComplexityBest For
Docker Compose🟢 LowQuick start, local development
Python Virtual Env🟡 MediumDevelopment, debugging

Cloud Platforms​

Scalable options for production workloads.

PlatformComplexityScalabilityBest For
Fly.io🟢 Low🟡 MediumGlobal edge deployment
Google Cloud🟡 Medium🟢 HighGCP-native integration
Dagster Cloud🟢 Low🟢 HighServerless, managed

Containerized Deployment​

OptionUse Case
Docker ComposeSingle-node deployment
KubernetesMulti-node, enterprise scale
Docker SwarmSimple orchestration

Configuration Management​

FeatureDescriptionCompatibility
Deployment ProfilesEnvironment-specific configurations (demo, production, dev)Fly.io, Docker, Custom
Environment VariablesRuntime configuration overridesAll platforms
Hot ReloadDynamic configuration updatesDocker, Local

Architecture Patterns​

# Everything in one deployment
make docker          # or
make fly-deploy      # or  
make dagster-cloud

Pros: Simple setup, everything included Cons: Single point of failure, harder to scale components independently

Pattern 2: Service Separation​

Pros: Independent scaling, better reliability Cons: More complex setup and management

Pattern 3: Fully Distributed​

Pros: Maximum scalability and reliability Cons: Most complex to set up and debug

Choosing Your Deployment​

Quick Decision Tree​

Configuration Examples​

Full Stack Configuration​

# .env for full stack
ANOMSTACK_DUCKDB_PATH=/data/anomstack.db
ANOMSTACK_DASHBOARD_HOST=0.0.0.0
ANOMSTACK_DASHBOARD_PORT=8080
ANOMSTACK_ALERT_EMAIL_FROM=alerts@company.com
ANOMSTACK_ALERT_EMAIL_TO=team@company.com
ANOMSTACK_SLACK_CHANNEL=#anomalies

Headless Configuration​

# .env for headless deployment
ANOMSTACK_DB=bigquery
ANOMSTACK_TABLE_KEY=analytics.anomaly_detection.metrics
ANOMSTACK_MODEL_PATH=gs://company-ml-models/anomstack/
ANOMSTACK_DISABLE_DASHBOARD=true
ANOMSTACK_ALERT_WEBHOOK_URL=https://api.company.com/alerts

Next Steps​

Getting Started​

  1. 🚀 Quick Demo: Visit anomstack-demo.fly.dev to see Anomstack in action
  2. 📖 Follow Guides: Choose your deployment method from the guides below
  3. ⚙️ Configure Metrics: Set up your first metric batch
  4. 🔔 Test Alerts: Configure and test your alerting channels

Deployment Guides​

Advanced Topics​

Support​

Choose your deployment path and get started with reliable, open-source anomaly detection! 🎉