Sign inSign up

sudeshmu/kubeagentic

By sudeshmu

•Updated about 1 year ago

Deploy AI Agents on Kubernetes https://kubeagentic.com/

Machine learning & AI
Data science
0

211

sudeshmu/kubeagentic repository overview

⁠KubeAgentic - Multi-Architecture Kubernetes AI Agent Operator

Docker Pulls Image Size GitHub Stars Architectures

Deploy and manage AI agents on Kubernetes with simple YAML configurations. KubeAgentic is a powerful Kubernetes operator that simplifies the deployment, management, and scaling of AI agents in your cluster.

ā šŸ—ļø Multi-Architecture Support

āœ… Native support for both AMD64 and ARM64 architectures

  • Intel/AMD x86_64: Traditional servers, VMs, most cloud instances
  • ARM64: Apple Silicon (M1/M2), AWS Graviton, GCP T2A, Azure Ampere
  • Automatic Selection: Kubernetes automatically picks the right architecture
  • Single Manifest: One image reference works on all platforms

ā šŸ·ļø Available Images

docker pull sudeshmu/kubeagentic:operator-latest
  • Size: ~219MB (Multi-stage optimized)
  • Base: Red Hat UBI Micro
  • Architectures: linux/amd64, linux/arm64
  • Purpose: Kubernetes operator for managing agents
  • Build: Multi-architecture using Docker Buildx
⁠Agent Runtime
docker pull sudeshmu/kubeagentic:agent-fixed
  • Size: ~1.25GB (Includes all AI frameworks)
  • Base: Red Hat UBI Minimal
  • Architectures: linux/amd64, linux/arm64
  • Purpose: Python 3.11 + FastAPI + LangGraph/LangChain
  • Features: Direct + LangGraph framework support
⁠⚔ Architecture-Specific Pulls
# Automatically selects your platform
docker pull sudeshmu/kubeagentic:agent-fixed

# Force specific architecture
docker pull --platform linux/amd64 sudeshmu/kubeagentic:agent-fixed
docker pull --platform linux/arm64 sudeshmu/kubeagentic:agent-fixed

ā šŸš€ Quick Start

⁠1. Deploy the Operator
# Pull the optimized operator image
docker pull sudeshmu/kubeagentic:operator-latest

# Deploy to Kubernetes
kubectl apply -f https://raw.githubusercontent.com/KubeAgentic-Community/kubeagentic/main/deploy/all.yaml
⁠2. Create Your First Agent
apiVersion: ai.example.com/v1
kind: Agent
metadata:
  name: my-assistant
spec:
  provider: openai
  model: gpt-4
  framework: direct  # Choose: direct or langgraph
  systemPrompt: "You are a helpful assistant."
  apiSecretRef:
    name: openai-secret
    key: api-key
⁠3. Test Multi-Architecture Deployment
# View image manifests
docker buildx imagetools inspect sudeshmu/kubeagentic:operator-latest
docker buildx imagetools inspect sudeshmu/kubeagentic:agent-fixed

# Test on different architectures
docker run --rm --platform linux/amd64 sudeshmu/kubeagentic:agent-fixed python --version
docker run --rm --platform linux/arm64 sudeshmu/kubeagentic:agent-fixed python --version

⁠✨ Multi-Architecture Build Features

ā šŸ—ļø Docker Buildx Integration
  • Cross-platform builds: Single command builds for AMD64 + ARM64
  • Manifest lists: One image tag serves all architectures automatically
  • Parallel builds: Simultaneous compilation for faster CI/CD
  • Registry optimization: Efficient layer sharing between architectures
ā šŸ”’ Security Hardening
  • šŸ”’ Non-root execution with user ID 1001 (agent) and 65532 (operator)
  • šŸ›”ļø Red Hat Universal Base Images (UBI) for enterprise security
  • šŸ” Minimal attack surface with only required packages
  • 🚫 No package managers in runtime images
  • šŸ·ļø Attestation manifests for supply chain security
⁠⚔ Performance Optimizations
  • šŸ–„ļø Native execution on both x86_64 and ARM64
  • ⚔ Virtual environments for isolated Python dependencies
  • šŸ—œļø Layer optimization with combined commands
  • šŸ’¾ Efficient caching with .dockerignore patterns
  • šŸš€ Platform-specific optimizations during build

ā šŸ—ļø Supported AI Providers

ProviderModelsAuthentication
OpenAIGPT-4, GPT-3.5-turboAPI Key
AnthropicClaude-3 (Opus, Sonnet, Haiku)API Key
GoogleGemini Pro, Gemini Pro VisionAPI Key
vLLMSelf-hosted modelsOptional API Key

ā šŸ”§ Configuration Examples

ā šŸš€ High-Performance Setup (Direct Framework)
apiVersion: ai.example.com/v1
kind: Agent
metadata:
  name: high-performance-agent
spec:
  provider: openai
  model: gpt-4
  framework: direct  # Low latency, simple workflows
  replicas: 5
  systemPrompt: "You are a high-performance assistant."
  resources:
    requests: {cpu: 200m, memory: 256Mi}
    limits: {cpu: 500m, memory: 512Mi}
  apiSecretRef:
    name: openai-secret
    key: api-key
⁠🧠 Complex Workflow Setup (LangGraph Framework)
apiVersion: ai.example.com/v1
kind: Agent
metadata:
  name: workflow-agent
spec:
  provider: anthropic
  model: claude-3-sonnet-20240229
  framework: langgraph  # Complex multi-step workflows
  systemPrompt: "You are a workflow automation assistant."
  langgraphConfig:
    graphType: conditional
    nodes:
      - name: analyze
        type: llm
      - name: tools
        type: tool
    edges:
      - from: analyze
        to: tools
        condition: needs_tools
    entrypoint: analyze
  tools:
    - calculator
    - web_search
  apiSecretRef:
    name: anthropic-secret
    key: api-key
⁠🌐 Multi-Architecture Deployment
# Works automatically on both AMD64 and ARM64 nodes
apiVersion: apps/v1
kind: Deployment
metadata:
  name: kubeagentic-operator
spec:
  replicas: 2
  template:
    spec:
      containers:
      - name: manager
        image: sudeshmu/kubeagentic:operator-latest  # ← Auto-selects architecture
        resources:
          requests: {cpu: 100m, memory: 128Mi}
          limits: {cpu: 200m, memory: 256Mi}

ā šŸ“Š Multi-Architecture Image Details

ComponentAMD64 SizeARM64 SizeArchitecturesBase Image
Operator~219MB~219MBāœ… BothUBI Micro
Agent~1.25GB~1.25GBāœ… BothUBI Minimal
ā šŸ·ļø Image Manifests
# Each image tag contains multiple architecture-specific manifests
sudeshmu/kubeagentic:operator-latest
ā”œā”€ā”€ linux/amd64 → sha256:2335acc4...
ā”œā”€ā”€ linux/arm64 → sha256:08d4833d...
└── attestations (security metadata)

sudeshmu/kubeagentic:agent-fixed  
ā”œā”€ā”€ linux/amd64 → sha256:c33d00cb...
ā”œā”€ā”€ linux/arm64 → sha256:2cdf8f8e...
└── attestations (security metadata)

ā šŸ› ļø Development & Customization

ā šŸ—ļø Build Multi-Architecture Images
# Clone the repository
git clone https://github.com/KubeAgentic-Community/KubeAgentic.git
cd KubeAgentic

# Setup buildx for multi-architecture builds
make buildx-setup

# Build and push multi-architecture images
make docker-buildx-all

# Or build individually
make docker-buildx-operator  # Build operator for AMD64 + ARM64
make docker-buildx-agent     # Build agent for AMD64 + ARM64

# Build locally without pushing (for development)
make docker-buildx-local-all

# Legacy single-architecture builds (if needed)
docker build -f Dockerfile.operator -t my-kubeagentic:operator .
docker build -f Dockerfile.agent -t my-kubeagentic:agent .
ā šŸ” Inspect Multi-Architecture Images
# View detailed manifest information
make inspect-images

# Or manually inspect
docker buildx imagetools inspect sudeshmu/kubeagentic:operator-latest
docker buildx imagetools inspect sudeshmu/kubeagentic:agent-fixed
⁠Environment Variables
# Operator
- LOG_LEVEL=info
- METRICS_ADDR=:8080
- HEALTH_PROBE_ADDR=:8081

# Agent Runtime
- PORT=8080
- LOG_LEVEL=info
- PYTHONUNBUFFERED=1

ā šŸ“‹ System Requirements

ā šŸ–„ļø Multi-Architecture Support
  • AMD64 nodes: Intel/AMD x86_64 processors
  • ARM64 nodes: Apple Silicon, AWS Graviton, GCP T2A, Azure Ampere
  • Mixed clusters: Automatic architecture selection
⁠Minimum Requirements
  • Kubernetes: v1.19+
  • CPU: 100m per agent (both architectures)
  • Memory: 128Mi per agent (both architectures)
  • Storage: 1.5Gi for images (both architectures cached)
  • CPU: 200-500m per agent
  • Memory: 256Mi-1Gi per agent
  • Replicas: 2+ for high availability (can span architectures)
  • Monitoring: Prometheus + Grafana
  • Node selection: Mix of AMD64 and ARM64 for cost optimization
ā ā˜ļø Cloud Provider Compatibility
ProviderAMD64 SupportARM64 SupportARM Instance TypesCost Savings
AWSāœ…āœ…Graviton2/3 instancesUp to 40%
GCPāœ…āœ…T2A instancesUp to 35%
Azureāœ…āœ…Ampere Altra instancesUp to 50%
Localāœ…āœ…M1/M2 Macs, ARM SBCsNative performance

ā šŸ·ļø Tags & Versioning

TagDescriptionArchitecturesUpdate Frequency
operator-latestLatest stable operatorAMD64, ARM64On releases
agent-fixedProduction-ready agent with LangGraphAMD64, ARM64On stable releases
agent-latestDevelopment agent buildsAMD64, ARM64On commits
ā šŸ“¦ Multi-Architecture Tags
# These tags automatically serve the correct architecture:
sudeshmu/kubeagentic:operator-latest  # āœ… AMD64 + ARM64
sudeshmu/kubeagentic:agent-fixed      # āœ… AMD64 + ARM64 (RECOMMENDED)
sudeshmu/kubeagentic:agent-latest     # āœ… AMD64 + ARM64 (Development)

# Architecture-specific tags (if needed):
sudeshmu/kubeagentic:operator-latest@sha256:2335acc4...  # AMD64 only
sudeshmu/kubeagentic:operator-latest@sha256:08d4833d...  # ARM64 only

ā šŸ¤ Contributing

We welcome contributions! Please see our Contributing Guide⁠ for details.

ā šŸ“„ License

Apache License 2.0 - see LICENSE⁠


šŸš€ Built with ā¤ļø for the Multi-Architecture Kubernetes AI Community

These multi-architecture images are built using Docker Buildx with Red Hat UBI base images, ensuring security, performance, and native execution on both AMD64 and ARM64 architectures. Deploy once, run anywhere!

ā šŸŽÆ Why Multi-Architecture?
  • šŸ’° Cost savings: Use ARM-based cloud instances (up to 50% cheaper)
  • ⚔ Performance: Native execution on Apple Silicon and ARM servers
  • šŸŒ Flexibility: Deploy on any Kubernetes cluster architecture
  • šŸ”® Future-proof: Ready for the ARM64 adoption wave

Experience the power of truly portable AI agents! šŸ¤–āœØ

Tag Summary

No tags have been pushed to this repository yet.