Make apiVersion configurable for inferencePool in the helm charts (#1542)

* Make apiVersion configurable in helm chart for inferencePool

* Update helm chart to accomodate for targetPortNumber in v1alpha2

* updating helm charts to work on the cluster with v1 and v1a2

---------

Co-authored-by: Kellen Swain <[email protected]>

Author: rahulgurnani

config/charts/inferencepool/templates/inferencepool.yaml

@@ -1,5 +1,26 @@
+{{ if eq .Values.inferencePool.apiVersion "inference.networking.x-k8s.io/v1alpha2"}}
+apiVersion: {{ .Values.inferencePool.apiVersion }}
+kind: InferencePool
+metadata:
+  name: {{ .Release.Name }}
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "gateway-api-inference-extension.labels" . | nindent 4 }}
+spec:
+  targetPortNumber: {{ .Values.inferencePool.targetPortNumber | default 8000 }}
+  selector:
+    {{- if .Values.inferencePool.modelServers.matchLabels }}
+    {{- range $key, $value := .Values.inferencePool.modelServers.matchLabels }}
+    {{ $key }}: {{ quote $value }}
+    {{- end }}
+    {{- end }}
+  extensionRef:
+    name: {{ include "gateway-api-inference-extension.name" . }}
+    portNumber: {{ .Values.inferenceExtension.extProcPort | default 9002 }}
+    failureMode: {{ .Values.inferenceExtension.failureMode | default "FailClose" }}
+{{ else }}
 {{ include "gateway-api-inference-extension.validations.inferencepool.common" $ }}
-apiVersion: inference.networking.k8s.io/v1
+apiVersion: "inference.networking.k8s.io/v1"
 kind: InferencePool
 metadata:
   name: {{ .Release.Name }}
@@ -22,5 +43,6 @@ spec:
     name: {{ include "gateway-api-inference-extension.name" . }}
     port:
       number: {{ .Values.inferenceExtension.extProcPort | default 9002 }}
+{{- end }}
 
 

config/charts/inferencepool/values.yaml

@@ -44,10 +44,15 @@ inferencePool:
   targetPorts:
     - number: 8000
   modelServerType: vllm # vllm, triton-tensorrt-llm
+  apiVersion: inference.networking.k8s.io/v1 
   # modelServers: # REQUIRED
   #   matchLabels:
   #     app: vllm-llama3-8b-instruct
 
+  # Should only used if apiVersion is inference.networking.x-k8s.io/v1alpha2, 
+  # This will soon be deprecated when upstream GW providers support v1, just doing something simple for now.
+  targetPortNumber: 8000
+
 provider:
   name: none