Merge branch 'main' into fix/grafana-dashboard-improvements

Author: sd109

chart/templates/api/config-map-grafana-dashboard.yml

@@ -788,22 +788,4 @@
       "title": "GPU Power Draw",
       "type": "timeseries"
     }
-  ],
-  "refresh": "",
-  "schemaVersion": 39,
-  "tags": [],
-  "templating": {
-    "list": []
-  },
-  "time": {
-    "from": "now-30m",
-    "to": "now"
-  },
-  "timepicker": {},
-  "timezone": "",
-  "title": "LLM Dashboard",
-  "uid": "c4114a95-8172-4a8b-99c8-963ba7d31385",
-  "version": 1,
-  "weekStart": ""
-}
 {{- end -}}

chart/templates/api/deployment.yml

@@ -35,15 +35,14 @@ spec:
         {{- if .Values.huggingface.secretName }}
         envFrom:
         - secretRef:
-            name: huggingface-token
-        {{- else if .Values.huggingface.token }}
+            name: {{ .Values.huggingface.secretName }}
+        {{- end }}
         env:
         - name: DO_NOT_TRACK
-          value: 1
+          value: "1"
+        {{- if .Values.huggingface.token }}
         - name: HUGGING_FACE_HUB_TOKEN
           value: {{ quote .Values.huggingface.token }}
-        {{- else if contains "Llama" .Values.huggingface.model -}}
-        {{- fail "Either secretName or token value must be set for Llama and other gated models" }}
         {{- end }}
         readinessProbe:
           httpGet:
@@ -60,4 +59,4 @@ spec:
         - name: shm
           emptyDir:
             medium: Memory
-            sizeLimit: 1Gi
+            sizeLimit: 1Gi

chart/values.yaml

@@ -35,7 +35,7 @@ api:
   # Container image config
   image:
     repository: vllm/vllm-openai
-    version: v0.4.1
+    version: v0.4.2
   monitoring:
     enabled: true
   # Service config
@@ -98,7 +98,7 @@ ui:
   # Container image config
   image:
     repository: ghcr.io/stackhpc/azimuth-llm-ui-base
-    version: "0.2.0-rc1"
+    version: "0.3.0"
   # Service config
   service:
     name: web-app