# Ollama — local LLM inference with GPU
resource "kubernetes_namespace" "ollama" {
  metadata {
    name = "ollama"
    labels = {
      managed-by = "terraform"
      tier       = "application"
    }
  }
}

resource "kubernetes_persistent_volume_claim" "ollama_models" {
  metadata {
    name      = "ollama-models"
    namespace = kubernetes_namespace.ollama.metadata[0].name
  }

  spec {
    access_modes = ["ReadWriteOnce"]
    resources {
      requests = {
        storage = "50Gi"
      }
    }
  }
}

# Deployment + Service managed by Argo CD — see servers/legion/apps/ollama.yaml