Show HN: Lovable Clone with K8s based orchestration Demo of POC (github.com)

Lovable Clone (elbavol) is a proof-of-concept repo that demos an end-to-end Kubernetes-based orchestration for LLM/agent workflows. The stack combines Go services (ingress, orchestrator, prime), TypeScript/Bun apps (control-pod, serve-pod), and infra (Kafka, Zookeeper, MinIO) with scripts for fast local iteration via docker-compose and full local K8s using minikube or kind. Out-of-the-box commands (docker-compose up -d, minikube start + minikube image load, kind create cluster) bootstrap infra; helper scripts create Kafka topics and deploy YAMLs. Images are published as aivalacoder/elbavol-*, and the repo includes ops/ manifests for GKE/EKS, port-forwarding examples and SSE-based tool execution updates in the pipeline flow. This matters because it packages a realistic orchestration pattern for AI products: a message-driven (Kafka) control plane, an orchestrator that runs plan->execute->validate loops, and separate “prime” API and agent pods to scale LLM tasks independently. Key technical implications include the need for durable Kafka topics, object storage (MinIO), image management for K8s, and operational best practices (secrets, PVs for Kafka, resource limits, monitoring/backup, imagePullPolicy). The repo is useful as a template for teams building production-grade LLM agents, giving reproducible local dev, CI/CD hooks, and clear migration paths to managed cloud services.