Kubernetes-native LLM routing service using Thompson Sampling bandit for intelligent model selection.
- ✅ Reinforcement Learning - Multi-armed bandit learns optimal model routing
- ✅ Kubernetes Native - Deployments, Services, ConfigMaps, HPA, Ingress
- ✅ Feedback Loop - Every request logged for analysis
- ✅ Health Checks - Liveness/readiness probes
- ✅ Horizontal Scaling - Auto-scales based on CPU/memory
- ✅ Multi-Model Support - GPT-4, Llama (via Ollama), extensible
make local