The Skyflo.ai Kubernetes Controller is a robust operator designed to facilitate the deployment, configuration, and management of the Skyflo.ai platform within Kubernetes environments. By leveraging a single Custom Resource Definition (CRD), the controller streamlines the orchestration of Skyflo.ai's components, ensuring efficient and secure operations.

The controller is developed in Go and built using the Kubebuilder framework with controller-runtime library, adhering to Kubernetes operator best practices. It introduces a singular CRD to manage the Skyflo.ai deployment, encompassing the UI, Engine, and MCP components.

• SkyfloAI (skyfloais.skyflo.ai):
  • Spec Fields (Required: ui, engine, mcp):
    • ui: Configuration parameters for the Skyflo.ai UI component.
      • image (required)
      • replicas
      • resources
      • env variables
    • engine: Settings for the Engine component.
      • image (required)
      • replicas
      • resources
      • databaseConfig (PostgreSQL configuration)
      • redisConfig (Redis configuration)
      • env variables
    • mcp: Parameters governing the MCP component.
      • image (required)
      • replicas
      • resources
      • kubeconfigSecret
      • env variables
    • imagePullSecrets: Secrets for pulling images from private registries.
    • nodeSelector: Node selection constraints for scheduling pods.
    • tolerations: Tolerations for scheduling pods on tainted nodes.
    • affinity: Affinity rules for pod scheduling.
  • Status Fields:
    • uiStatus: Current status of the UI component.
    • engineStatus: Status of the Engine component.
    • mcpStatus: Status of the MCP component.
    • conditions: Overall conditions and health indicators of the Skyflo.ai deployment.

• Controller Manager:
  • Watches for changes to the SkyfloAI custom resource.
  • Reconciles the desired state by managing Deployments, Services, and other Kubernetes resources.
  • Provides metrics endpoint for monitoring (:8080)
  • Implements health probes for liveness and readiness (:8081)
  • Supports leader election for high availability
• RBAC Management:
  • Ensures the MCP component has the necessary permissions to interact with cluster resources.
  • Configures Role-Based Access Control (RBAC) policies based on the specified access level.
  • Implements cluster-admin role binding for MCP service account

• Namespace Isolation:
  • All Skyflo.ai components are deployed within the skyflo-ai namespace
  • Network policies controlling inter-component communication
• Standard Kubernetes Resources:
  • UI Deployment with NodePort service (30080)
  • API Service Deployment with associated service
  • MCP Deployment with NodePort service (30081)
  • StatefulSets for Redis and PostgreSQL with persistent storage
• Security Model:
  • Non-root container execution with specific UIDs
  • Restricted capabilities and privilege escalation controls
  • Secure inter-service communication through network policies

• Unified Deployment:
  • Simplifies the deployment process by managing all Skyflo.ai components through a single custom resource.
  • Integrated database and cache management with PostgreSQL and Redis
• Dynamic Configuration:
  • Supports on-the-fly updates to component configurations, scaling, and resource allocations.
  • Environment variable customization for all components
  • Configurable resource requests and limits
• Secure Cluster Interaction:
  • Provides configurable access levels for the MCP component through RBAC
  • Network policies controlling pod-to-pod communication
  • Non-root security model with minimal container capabilities
• Resource Efficiency:
  • Utilizes standard Kubernetes resources where appropriate
  • Built-in monitoring through metrics endpoint
  • Health checking capabilities for all components
  • Persistent storage for stateful components

• Docker
• KinD (Kubernetes In Docker)
• kubectl
• helm (optional, for Helm-based installation)

Refer to the Local Development guide for more information.

• api/v1/skyfloai_types.go:
  • Defines the SkyfloAI custom resource schema.
• controllers/skyfloai_controller.go:
  • Implements the reconciliation logic for managing Skyflo.ai components.
• config/:
  • Contains Kubernetes manifests for CRDs, RBAC, and sample resources.

• Website
• Discord Community
• Twitter/X Updates
• GitHub Discussions