skyflo

Installing Skyflo on Kubernetes

Deploy Skyflo to any Kubernetes cluster with a single command. The interactive installer provisions all resources, configures your LLM provider, and generates secrets automatically.

curl -sL https://skyflo.ai/install.sh | bash

You will be prompted for:

1. Target namespace (default: skyflo-ai)
2. LLM model in provider/model format (e.g. openai/gpt-4o, anthropic/claude-sonnet-4-6)
3. Provider API key
4. Self-hosted LLM endpoint (if applicable)

Secrets, database credentials, and internal service URLs are generated automatically.

Verify

kubectl get pods -n skyflo-ai

All pods should reach Running status. The Engine waits for PostgreSQL and Redis to become healthy before starting.

Ingress

Expose Skyflo through an Ingress resource. The manifest below is cloud-agnostic. Add provider-specific annotations for your environment.

1. Create the Ingress manifest:

# skyflo-ai-ingress.yaml
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: skyflo-ai-ingress
  namespace: skyflo-ai
  annotations:
    # Add cloud-specific annotations below
spec:
  ingressClassName: nginx
  tls:
  - hosts:
    - your-domain.com
    secretName: skyflo-ai-tls
  rules:
  - host: your-domain.com
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service:
            name: skyflo-ai-ui
            port:
              number: 80

AWS ALB

```yaml metadata: annotations: kubernetes.io/ingress.class: alb alb.ingress.kubernetes.io/scheme: internal alb.ingress.kubernetes.io/target-type: ip alb.ingress.kubernetes.io/subnets: subnet-xxxxx, subnet-yyyyy alb.ingress.kubernetes.io/security-groups: sg-xxxxx alb.ingress.kubernetes.io/certificate-arn: arn:aws:acm:region:account:certificate/xxxxx alb.ingress.kubernetes.io/listen-ports: '[{"HTTP":80,"HTTPS":443}]' alb.ingress.kubernetes.io/ssl-redirect: "443" spec: ingressClassName: alb ```

GCP Cloud Load Balancer

```yaml metadata: annotations: kubernetes.io/ingress.class: gce networking.gke.io/managed-certificates: skyflo-ai-cert kubernetes.io/ingress.global-static-ip-name: skyflo-ai-ip spec: ingressClassName: gce ```

Azure Application Gateway

```yaml metadata: annotations: kubernetes.io/ingress.class: azure/application-gateway appgw.ingress.kubernetes.io/ssl-redirect: "true" appgw.ingress.kubernetes.io/backend-protocol: "http" spec: ingressClassName: azure-application-gateway ```

1. Apply:

kubectl apply -f skyflo-ai-ingress.yaml

1. Retrieve the external address:

kubectl get ingress -n skyflo-ai

1. Point your domain to the load balancer address via DNS (CNAME for hostname, A record for IP).

---

Reference

Prerequisites

The installer validates these automatically.

• Kubernetes cluster (v1.19+)
• kubectl
• gettext (for envsubst)
• curl
• openssl

Version Pinning

VERSION=v0.5.0 bash <(curl -sL https://skyflo.ai/install.sh)

Deployed Resources

Resource	Description
Engine Deployment	FastAPI backend with LangGraph agent (port 8080)
MCP Deployment	FastMCP tool server with kubectl, Helm, Argo, Jenkins (port 8888)
UI Deployment	Next.js frontend with Nginx proxy sidecar (port 80)
Controller Deployment	Kubernetes operator for SkyfloAI CRD (Go)
PostgreSQL StatefulSet	Primary database with 5Gi persistent volume
Redis StatefulSet	Pub/sub, rate limiting, stop signals with 1Gi volume
ConfigMaps	Non-sensitive configuration for Engine, MCP, and UI
Secrets	JWT secret, database URLs, LLM API keys
NetworkPolicy	Restricts MCP ingress to Engine pods only
RBAC	ServiceAccounts, ClusterRole, ClusterRoleBinding
CRD	skyfloais.skyflo.ai custom resource definition

Generated Values

The following are created automatically unless pre-set as environment variables:

Variable	Default
JWT_SECRET	Random 32-byte base64
POSTGRES_PASSWORD	Random URL-safe string
POSTGRES_DATABASE_URL	In-cluster PostgreSQL
REDIS_URL	redis://skyflo-ai-redis:6379/0
MCP_SERVER_URL	http://skyflo-ai-mcp:8888/mcp

LLM Configuration

The installer configures your LLM provider interactively. Use the reference below to pre-set values as environment variables or to modify the configuration after installation.

Supported Models

Set LLM_MODEL using provider/model format:

LLM_MODEL=gemini/gemini-2.5-pro       # Gemini (recommended)
LLM_MODEL=moonshot/kimi-k2.5          # Moonshot
LLM_MODEL=deepseek/deepseek-reasoner  # Deepseek
LLM_MODEL=groq/openai/gpt-oss-120b    # Groq
LLM_MODEL=anthropic/claude-sonnet-4-6 # Anthropic
LLM_MODEL=openai/gpt-5.3-codex        # OpenAI
LLM_MODEL=ollama/llama3.1:8b          # Ollama (self-hosted)

Skyflo connects to LLM providers through LiteLLM. See the full list of supported models.

API Keys

ANTHROPIC_API_KEY=sk-...    # Anthropic
MOONSHOT_API_KEY=sk-...     # Moonshot
DEEPSEEK_API_KEY=sk-...     # Deepseek
GROQ_API_KEY=gsk-...        # Groq
GEMINI_API_KEY=AI-...       # Gemini
OPENAI_API_KEY=sk-...       # OpenAI

Additional providers: AWS Bedrock, HuggingFace, Databricks, Fireworks AI, Together AI, NVIDIA NIM, Perplexity, xAI, and others.

Self-hosted Models

For Ollama or other self-hosted endpoints:

LLM_HOST=http://your-model-host:port

Reasoning Models

Skyflo auto-detects reasoning capabilities using LiteLLM’s model registry. Models with native reasoning support (OpenAI o-series, Anthropic Claude with extended thinking, DeepSeek-R1) are enabled automatically at high effort. The reasoning process streams to the UI in collapsible thinking blocks.

Setting LLM_REASONING_EFFORT to high yields the best results with Skyflo.

To override defaults, set these environment variables on the Engine:

LLM_REASONING_EFFORT=high              # low, medium, high
LLM_THINKING_BUDGET_TOKENS=10000       # Anthropic-specific
LLM_MAX_TOKENS=16384                   # Max tokens when thinking is enabled

Building from Source

See deployment/README.md.

Uninstalling

curl -sL https://skyflo.ai/uninstall.sh | bash

Pin a specific version:

VERSION=<version> bash <(curl -sL https://skyflo.ai/uninstall.sh)

The uninstaller prompts for the target namespace and whether to delete persistent volume claims. Confirming deletion permanently removes all PostgreSQL and Redis data.

Troubleshooting

1. Check pod status:

kubectl get pods -n skyflo-ai
kubectl describe pod <pod-name> -n skyflo-ai

1. View logs:

kubectl logs <pod-name> -n skyflo-ai

1. Common issues:
   • Pods stuck in Pending: insufficient cluster resources
   • ImagePullBackOff: verify container registry access and image tags
   • Engine not starting: confirm PostgreSQL and Redis pods are healthy
   • Connection failures: verify services with kubectl get svc -n skyflo-ai

Community

• Discord
• Website

This site is open source. Improve this page.