Deploy AI Agents on Kubernetes https://kubeagentic.com/
211
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 Native support for both AMD64 and ARM64 architectures
docker pull sudeshmu/kubeagentic:operator-latest
linux/amd64, linux/arm64docker pull sudeshmu/kubeagentic:agent-fixed
linux/amd64, linux/arm64# 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
# 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
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
# 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
| Provider | Models | Authentication |
|---|---|---|
| OpenAI | GPT-4, GPT-3.5-turbo | API Key |
| Anthropic | Claude-3 (Opus, Sonnet, Haiku) | API Key |
| Gemini Pro, Gemini Pro Vision | API Key | |
| vLLM | Self-hosted models | Optional API Key |
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
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
# 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 # ā Auto-selects architecture
resources:
requests: {cpu: 100m, memory: 128Mi}
limits: {cpu: 200m, memory: 256Mi}
| Component | AMD64 Size | ARM64 Size | Architectures | Base Image |
|---|---|---|---|---|
| Operator | ~219MB | ~219MB | ā Both | UBI Micro |
| Agent | ~1.25GB | ~1.25GB | ā Both | UBI Minimal |
# Each image tag contains multiple architecture-specific manifests
sudeshmu/kubeagentic:operator-latest
āāā linux/amd64 ā sha256:2335acc4...
āāā linux/arm64 ā sha256:08d4833d...
āāā attestations (security metadata)
sudeshmu/kubeagentic:agent-fixed
āāā linux/amd64 ā sha256:c33d00cb...
āāā linux/arm64 ā sha256:2cdf8f8e...
āāā attestations (security metadata)
# 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 .
# 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
# Operator
- LOG_LEVEL=info
- METRICS_ADDR=:8080
- HEALTH_PROBE_ADDR=:8081
# Agent Runtime
- PORT=8080
- LOG_LEVEL=info
- PYTHONUNBUFFERED=1
| Provider | AMD64 Support | ARM64 Support | ARM Instance Types | Cost Savings |
|---|---|---|---|---|
| AWS | ā | ā | Graviton2/3 instances | Up to 40% |
| GCP | ā | ā | T2A instances | Up to 35% |
| Azure | ā | ā | Ampere Altra instances | Up to 50% |
| Local | ā | ā | M1/M2 Macs, ARM SBCs | Native performance |
| Tag | Description | Architectures | Update Frequency |
|---|---|---|---|
operator-latest | Latest stable operator | AMD64, ARM64 | On releases |
agent-fixed | Production-ready agent with LangGraph | AMD64, ARM64 | On stable releases |
agent-latest | Development agent builds | AMD64, ARM64 | On commits |
# These tags automatically serve the correct architecture:
sudeshmu/kubeagentic:operator-latest # ā
AMD64 + ARM64
sudeshmu/kubeagentic:agent-fixed # ā
AMD64 + ARM64 (RECOMMENDED)
sudeshmu/kubeagentic:agent-latest # ā
AMD64 + ARM64 (Development)
# Architecture-specific tags (if needed):
sudeshmu/kubeagentic:operator-latest@sha256:2335acc4... # AMD64 only
sudeshmu/kubeagentic:operator-latest@sha256:08d4833d... # ARM64 only
We welcome contributions! Please see our Contributing Guideā for details.
Apache License 2.0 - see LICENSEā
š 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!
Experience the power of truly portable AI agents! š¤āØ
No tags have been pushed to this repository yet.