zededa · adithya-zededa · Aug 28, 2025
diff --git a/EdgeFieldDay2025/EdgeAI_on_ZKS/edge-ai-car-classification-main/.gitignore b/EdgeFieldDay2025/EdgeAI_on_ZKS/edge-ai-car-classification-main/.gitignore
@@ -0,0 +1 @@
+volume-for-model/**
diff --git a/EdgeFieldDay2025/EdgeAI_on_ZKS/edge-ai-car-classification-main/README.md b/EdgeFieldDay2025/EdgeAI_on_ZKS/edge-ai-car-classification-main/README.md
@@ -0,0 +1,306 @@
+# Edge AI ML Inference Platform
+
+A complete, production-ready machine learning inference platform with dynamic model management on Kubernetes.
+
+## 🚀 Quick Start
+
+### Deploy with Docker Hub (Recommended)
+```bash
+# Build any local changes
+./build.sh
+
+# Deploy using pre-built Docker Hub images
+cd helm-chart/edge-ai-car-classification
+helm dependency update
+helm install edge-ai-car-classification . --namespace edge-ai-car-classification --create-namespace \
+  --set server.ovms.image=adithyazededa/edgeai-ovms-server:latest \
+  --set server.sidecar.image=adithyazededa/model-sync-sidecar:latest \
+  --set client.image=adithyazededa/edgeai-client-app:latest
+```
+
+### Deploy with Local Build
+```bash
+# Build all images locally
+./build.sh
+
+# Deploy with locally built images
+cd helm-chart/edge-ai-car-classification
+helm dependency update
+helm install ml-platform . --namespace ml-platform --create-namespace
+```
+
+### Cleanup
+```bash
+./cleanup.sh
+```
+
+## 🏗️ Architecture
+
+```
+┌─────────────────────────────────────────────────────────────┐
+│                    Kubernetes Cluster                       │
+│                                                             │
+│  ┌─────────────────┐    ┌─────────────────┐                │
+│  │   MinIO Pod     │    │   Server Pod    │                │
+│  │                 │    │                 │                │
+│  │  ┌───────────┐  │    │ ┌─────────────┐ │                │
+│  │  │   MinIO   │  │    │ │    OVMS     │ │                │
+│  │  │  Storage  │  │◄───┤ │  Container  │ │                │
+│  │  │           │  │    │ └─────────────┘ │                │
+│  │  └───────────┘  │    │                 │                │
+│  │                 │    │ ┌─────────────┐ │                │
+│  └─────────────────┘    │ │    Sync     │ │                │
+│                         │ │   Sidecar   │ │                │
+│                         │ └─────────────┘ │                │
+│  ┌─────────────────┐    └─────────────────┘                │
+│  │   Client Pod    │                                       │
+│  │                 │    ┌─────────────────┐                │
+│  │ ┌─────────────┐ │    │ Shared Storage  │                │
+│  │ │   Client    │ │    │     (PVC)       │                │
+│  │ │Application  │ │◄───┤   /models/      │                │
+│  │ └─────────────┘ │    │                 │                │
+│  └─────────────────┘    └─────────────────┘                │
+│                                                             │
+└─────────────────────────────────────────────────────────────┘
+```
+
+## ✨ Features
+
+- **Dynamic Model Loading** - Upload models to MinIO, they're automatically loaded
+- **Multi-Model Support** - Client discovers and uses all available models
+- **Zero Downtime** - Add new models without restarting services
+- **Production Ready** - Complete Kubernetes deployment with monitoring
+- **Auto-scaling** - Horizontal pod autoscaling based on demand
+- **Security** - RBAC integration and configurable secrets management
+
+## 📁 Project Structure
+
+```
+edge-ai-car-classification/
+├── client-container/          # Multi-model client application
+├── server-container/          # OpenVINO Model Server setup
+├── volume-for-model/          # Model files
+├── sync-sidecar/              # Dynamic model loading component
+├── helm-chart/                # Kubernetes deployment charts
+├── jupyter-notebook/          # ML model tracking notebook
+├── build.sh                   # Build automation script
+├── cleanup.sh                 # Cleanup automation script
+└── README.md                  # This file
+```
+
+## 🔧 Prerequisites
+
+- Kubernetes cluster (minikube, kind, or any K8s cluster)
+- kubectl configured to access your cluster
+- Helm 3.x installed
+- Docker for building images
+
+## 📦 Deployment Options
+
+### Option 1: Docker Hub Deployment (Easiest)
+Uses pre-built images from Docker Hub - no building required:
+
+```bash
+cd helm-chart/edge-ai-car-classification
+helm dependency update
+helm install ml-platform . --namespace ml-platform --create-namespace \
+  --set server.ovms.image=adithyazededa/edgeai-ovms-server:latest \
+  --set server.sidecar.image=adithyazededa/model-sync-sidecar:latest \
+  --set client.image=adithyazededa/edgeai-client-app:latest
+```
+
+### Option 2: Local Development Deployment
+Builds images locally and deploys:
+
+```bash
+# Build all container images
+./build.sh
+
+# Deploy the platform with local images
+cd helm-chart/edge-ai-car-classification
+helm dependency update
+helm install ml-platform . --namespace ml-platform --create-namespace
+```
+
+### Option 3: Custom Registry Deployment
+Build and push to your own registry:
+
+```bash
+# Build and push to custom registry
+REGISTRY=your-username PUSH=true ./build.sh
+
+# Deploy with custom images
+cd helm-chart/edge-ai-car-classification
+helm dependency update
+helm install ml-platform . --namespace ml-platform --create-namespace \
+  --set server.ovms.image=your-username/edgeai-ovms-server:latest \
+  --set server.sidecar.image=your-username/model-sync-sidecar:latest \
+  --set client.image=your-username/edgeai-client-app:latest
+```
+
+## 🔄 How Dynamic Model Loading Works
+
+1. **Upload Model**: Add new model to MinIO bucket with correct structure:
+   ```
+   ml-models/
+   └── model_name/          # e.g., efb3_car_onnx
+       └── version/         # e.g., 1, 2, latest  
+           └── model_files  # e.g., model.onnx
+   ```
+
+2. **Sync Detection**: Sidecar detects new model via polling (every 30s)
+3. **Download**: Sidecar downloads model to shared volume at `/models/`
+4. **Config Update**: Sidecar updates `/models/config.json` with new model
+5. **OVMS Reload**: OVMS automatically detects config change and loads model
+6. **Ready**: New model is available for inference at `/v1/models/model_name:predict`
+
+## 📊 Monitoring and Management
+
+### Check Deployment Status
+```bash
+# Check all pods status
+kubectl get pods -n ml-platform
+
+# Check services and endpoints
+kubectl get svc,endpoints -n ml-platform
+```
+
+### View Logs
+```bash
+# View OVMS Server logs
+kubectl logs -l app.kubernetes.io/component=server -c ovms -n ml-platform -f
+
+# View Model Sync Sidecar logs  
+kubectl logs -l app.kubernetes.io/component=server -c model-sync -n ml-platform -f
+
+# View Client Application logs
+kubectl logs -l app.kubernetes.io/component=client -n ml-platform -f
+```
+
+### Upload Models
+```bash
+# Access MinIO console
+kubectl port-forward svc/ml-platform-minio-console 9001:9001 -n ml-platform
+
+# Upload models via web UI at http://localhost:9001
+# Login with: minioadmin / minioadmin123
+```
+
+### Scale Components
+```bash
+# Scale server pods
+helm upgrade ml-platform ./helm-chart/edge-ai-car-classification --set server.replicaCount=3 --reuse-values -n ml-platform
+
+# Scale client pods  
+helm upgrade ml-platform ./helm-chart/edge-ai-car-classification --set client.replicaCount=2 --reuse-values -n ml-platform
+```
+
+## 🔒 Security Considerations
+
+1. **Change default MinIO credentials** in production
+2. **Use image pull secrets** for private registries
+3. **Enable RBAC** with minimal permissions
+4. **Use security contexts** to run as non-root
+5. **Enable network policies** to restrict traffic
+
+## 🆘 Troubleshooting
+
+### Common Issues
+
+#### Models not loading
+- Check sync sidecar logs: `kubectl logs -l app.kubernetes.io/component=server -c model-sync -n ml-platform`
+- Verify MinIO connectivity and bucket exists
+- Ensure correct model directory structure
+
+#### Client connection errors
+- Verify service names and ports
+- Check OVMS server is running: `kubectl get pods -l app.kubernetes.io/component=server -n ml-platform`
+
+#### Image pull failures
+- Check registry credentials and image names
+- For minikube: ensure images are loaded with `minikube image load`
+
+#### Storage issues
+- Verify PVC creation: `kubectl get pvc -n ml-platform`
+- Check storage class availability: `kubectl get storageclass`
+
+### Debug Commands
+```bash
+# Check all platform resources
+kubectl get all -l app.kubernetes.io/instance=ml-platform -n ml-platform
+
+# Get detailed resource information
+kubectl describe all -l app.kubernetes.io/instance=ml-platform -n ml-platform
+
+# Check pod events
+kubectl get events --sort-by=.metadata.creationTimestamp -n ml-platform
+
+# Access container shell
+kubectl exec -it deployment/ml-platform-edge-ai-car-classification-server -c ovms -n ml-platform -- /bin/bash
+```
+
+### Missing MinIO Bucket
+```bash
+# Create bucket manually using MinIO client
+kubectl port-forward svc/ml-platform-minio 9000:9000 -n ml-platform
+mc alias set local http://localhost:9000 minioadmin minioadmin123
+mc mb local/ml-models
+```
+
+## 🎯 Production Deployment
+
+For production environments:
+
+1. Use the production values template:
+   ```bash
+   cp helm-chart/edge-ai-car-classification/examples/values-prod.yaml production-values.yaml
+   # Edit with production settings
+   helm install ml-platform ./helm-chart/edge-ai-car-classification -f production-values.yaml
+   ```
+
+2. Key production settings:
+   - Multiple replicas for high availability
+   - Resource limits and requests
+   - Ingress configuration
+   - Auto-scaling enabled
+   - Monitoring integration
+   - Security hardening
+
+## 📚 Additional Documentation
+
+- **[Client Container](client-container/README.md)** - Multi-model client application details
+- **[Server Container](server-container/README.md)** - OpenVINO Model Server setup
+- **[Sync Sidecar](sync-sidecar/README.md)** - Dynamic model loading component
+- **[Helm Chart](helm-chart/edge-ai-car-classification/README.md)** - Kubernetes deployment details
+
+## 🎉 What You've Achieved
+
+You now have a **complete, production-ready ML inference platform** that delivers:
+
+✅ **Dynamic Model Management**: Models are automatically detected, downloaded, and loaded when uploaded to MinIO storage  
+✅ **Zero-Downtime Operations**: Add new models without restarting inference services  
+✅ **Container Orchestration**: Full Kubernetes-native deployment with Helm chart management  
+✅ **Multi-Model Support**: Platform automatically detects and serves multiple ML models simultaneously  
+✅ **Scalable Infrastructure**: Horizontal pod autoscaling for inference servers based on demand  
+✅ **High Availability**: Multi-replica deployments with load balancing and health checks  
+✅ **Fully Automated Deployment**: One-command deployment with `./deploy.sh` including all dependencies  
+✅ **Production Readiness**: Security best practices, monitoring, and comprehensive documentation
+
+## 📞 Support
+
+For issues or questions:
+1. Check the troubleshooting section above
+2. Review logs using the provided kubectl commands
+3. Consult the detailed documentation in component README files
+
+---
+
+**Ready to deploy?** Start with `./build.sh` followed by the Helm deployment commands above for the quickest setup!
+
+## 📦 Docker Images (Docker Hub)
+
+- `adithyazededa/edgeai-ovms-server:latest`
+- `adithyazededa/edgeai-model-sync-sidecar:latest` 
+- `adithyazededa/edgeai-client-app:latest`
+
+Deploy anywhere with: `./deploy.sh --dockerhub`