Back

RunAI Platform

Powered by a production-grade Kubernetes homelab
LIVE

A reusable, production-grade AI runtime built to power many products from a single backend. The platform handles conversational AI, multi-agent systems, real-time streaming, persistent memory, and multi-model routing. Everything runs on a fully self-managed Kubernetes cluster. Provisioned, secured, and operated without any managed cloud. Automated GitOps pipelines, encrypted secrets, zero-trust networking, and automated disaster recovery make the infrastructure as reliable and hardened as the applications it runs. The infrastructure isn't just scaffolding. It's the second half of the project.

AI Runtime

  • FastAPI + LangGraph orchestration
  • Real-time SSE streaming
  • LiteLLM multi-provider routing
  • Persistent memory via PostgreSQL
  • Auto title gen & history summarization
  • Passwordless auth (SuperTokens)
  • Token tracking & cost visibility

Homelab Infra

  • Self-managed Kubernetes cluster: no cloud dependency
  • ArgoCD GitOps: fully automated delivery
  • Sealed Secrets: encrypted secrets in git
  • Tailscale + Zero-Trust: no open ports
  • Velero + S3: automated disaster recovery
  • Longhorn: resilient distributed storage
  • Observability: Prometheus, Grafana, Jaeger, Langfuse

Python FastAPI LangGraph LiteLLM Streaming SSE Docker Kubernetes DevOps ArgoCD PostgreSQL Vector Database Observability Prometheus Grafana Langfuse Security Next.js TypeScript
View live project