From d3e770254de0bb301815ca87257c8b1a357d06c4 Mon Sep 17 00:00:00 2001
From: Your Name <you@example.com>
Date: Sun, 26 Apr 2026 21:02:47 +0800
Subject: hehe

---
 terraform/modules/llm/variables.tf | 112 +++++++++++++++++++++++++++++++++++++
 1 file changed, 112 insertions(+)
 create mode 100644 terraform/modules/llm/variables.tf

(limited to 'terraform/modules/llm/variables.tf')

diff --git a/terraform/modules/llm/variables.tf b/terraform/modules/llm/variables.tf
new file mode 100644
index 0000000..3a7d8f7
--- /dev/null
+++ b/terraform/modules/llm/variables.tf
@@ -0,0 +1,112 @@
+variable "release_name" {
+  type        = string
+  description = "Helm release name."
+}
+
+variable "namespace" {
+  type        = string
+  description = "Kubernetes namespace to deploy into."
+}
+
+variable "chart_path" {
+  type        = string
+  description = "Path to the local llm-app chart."
+}
+
+variable "replicas" {
+  type    = number
+  default = 1
+}
+
+variable "model_name" {
+  type        = string
+  description = "HuggingFace repo id, passed as vLLM model_tag (positional)."
+}
+
+variable "model_alias" {
+  type        = string
+  description = "Value clients pass in the OpenAI 'model' field (maps to --served-model-name)."
+}
+
+variable "max_model_len" {
+  type    = number
+  default = 2048
+}
+
+variable "dtype" {
+  type    = string
+  default = "bfloat16"
+}
+
+variable "omp_threads" {
+  type        = number
+  default     = 0
+  description = "OMP_NUM_THREADS for vLLM CPU backend. 0 = autodetect."
+}
+
+variable "extra_args" {
+  type        = list(string)
+  default     = []
+  description = "Extra CLI args passed to `vllm serve`, appended after the stock set."
+}
+
+variable "resources" {
+  type = object({
+    requests = object({ cpu = string, memory = string })
+    limits   = object({ cpu = string, memory = string })
+  })
+}
+
+variable "ingress_host" {
+  type = string
+}
+
+variable "ingress_class" {
+  type    = string
+  default = "nginx"
+}
+
+variable "image_repository" {
+  type    = string
+  default = "public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo"
+}
+
+variable "image_tag" {
+  type        = string
+  default     = "latest"
+  description = "Used only when image_digest is empty."
+}
+
+variable "image_digest" {
+  type        = string
+  default     = ""
+  description = "Optional sha256:abc... content-addressable digest. Takes precedence over image_tag."
+}
+
+variable "service_monitor_release_label" {
+  type        = string
+  default     = "kube-prometheus-stack"
+  description = "Must match the release label the Prometheus Operator selects on."
+}
+
+variable "model_cache_size" {
+  type    = string
+  default = "10Gi"
+}
+
+variable "hpa" {
+  type = object({
+    enabled              = bool
+    min_replicas         = number
+    max_replicas         = number
+    metric_name          = string
+    target_average_value = string
+  })
+  default = {
+    enabled              = false
+    min_replicas         = 1
+    max_replicas         = 3
+    metric_name          = "vllm:num_requests_running"
+    target_average_value = "500m"
+  }
+}
-- 
cgit