resource "kubernetes_namespace_v1" "this" {
  metadata {
    name = var.namespace
    labels = {
      "app.kubernetes.io/part-of" = "llm-platform"
    }
  }
}

resource "kubernetes_horizontal_pod_autoscaler_v2" "llm" {
  count = var.hpa.enabled ? 1 : 0

  metadata {
    name      = "${var.release_name}-llm-app"
    namespace = kubernetes_namespace_v1.this.metadata[0].name
  }
  spec {
    scale_target_ref {
      api_version = "apps/v1"
      kind        = "Deployment"
      name        = "${var.release_name}-llm-app"
    }
    min_replicas = var.hpa.min_replicas
    max_replicas = var.hpa.max_replicas

    metric {
      type = "Pods"
      pods {
        metric {
          name = var.hpa.metric_name
        }
        target {
          type          = "AverageValue"
          average_value = var.hpa.target_average_value
        }
      }
    }
  }

  depends_on = [helm_release.llm]
}

resource "helm_release" "llm" {
  name             = var.release_name
  chart            = var.chart_path
  namespace        = kubernetes_namespace_v1.this.metadata[0].name
  create_namespace = false
  atomic           = false
  wait             = true
  timeout          = 1800

  values = [
    yamlencode({
      replicaCount = var.replicas

      image = {
        repository = var.image_repository
        tag        = var.image_tag
        digest     = var.image_digest
        pullPolicy = "IfNotPresent"
      }

      model = {
        name        = var.model_name
        alias       = var.model_alias
        maxModelLen = var.max_model_len
        dtype       = var.dtype
      }

      server = {
        port       = 8000
        ompThreads = var.omp_threads
        extraArgs  = var.extra_args
      }

      resources = var.resources

      ingress = {
        enabled   = true
        className = var.ingress_class
        host      = var.ingress_host
      }

      monitoring = {
        serviceMonitor = {
          enabled  = true
          interval = "15s"
          labels = {
            release = var.service_monitor_release_label
          }
        }
      }

      modelCache = {
        sizeLimit = var.model_cache_size
      }
    }),
  ]
}