Improve Grafana dashboard and other minor fixes

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

@@ -31,6 +31,103 @@ data:
       "links": [],
       "liveNow": false,
       "panels": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "Request Rate",
+                "axisPlacement": "auto",
+                "axisWidth": 3,
+                "barAlignment": 0,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "auto",
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": null
+                  },
+                  {
+                    "color": "red",
+                    "value": 80
+                  }
+                ]
+              },
+              "unitScale": true
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 12,
+            "w": 24,
+            "x": 0,
+            "y": 0
+          },
+          "id": 7,
+          "options": {
+            "legend": {
+              "calcs": [],
+              "displayMode": "list",
+              "placement": "bottom",
+              "showLegend": true
+            },
+            "tooltip": {
+              "mode": "single",
+              "sort": "none"
+            }
+          },
+          "targets": [
+            {
+              "datasource": {
+                "type": "prometheus",
+                "uid": "prometheus"
+              },
+              "editorMode": "code",
+              "expr": "sum by (container, namespace, method, path) (rate(requests_total_counter[1m]))",
+              "instant": false,
+              "legendFormat": "__auto",
+              "range": true,
+              "refId": "A"
+            }
+          ],
+          "title": "API Requests",
+          "type": "timeseries"
+        },
         {
           "datasource": {
             "type": "prometheus",
@@ -95,7 +192,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 0,
-            "y": 0
+            "y": 12
           },
           "id": 2,
           "options": {
@@ -191,7 +288,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 12,
-            "y": 0
+            "y": 12
           },
           "id": 4,
           "options": {
@@ -287,7 +384,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 0,
-            "y": 8
+            "y": 20
           },
           "id": 3,
           "options": {
@@ -383,7 +480,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 12,
-            "y": 8
+            "y": 20
           },
           "id": 5,
           "options": {
@@ -479,7 +576,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 0,
-            "y": 16
+            "y": 28
           },
           "id": 1,
           "options": {
@@ -579,7 +676,7 @@ data:
             "h": 8,
             "w": 12,
             "x": 12,
-            "y": 16
+            "y": 28
           },
           "id": 6,
           "options": {
@@ -619,14 +716,14 @@ data:
         "list": []
       },
       "time": {
-        "from": "now-24h",
+        "from": "now-30m",
         "to": "now"
       },
       "timepicker": {},
       "timezone": "",
       "title": "LLM Dashboard",
       "uid": "c4114a95-8172-4a8b-99c8-963ba7d31385",
-      "version": 4,
+      "version": 1,
       "weekStart": ""
     }
 {{- end -}}

chart/templates/api/deployment.yml

@@ -38,6 +38,8 @@ spec:
             name: huggingface-token
         {{- else if .Values.huggingface.token }}
         env:
+        - name: DO_NOT_TRACK
+          value: 1
         - name: HUGGING_FACE_HUB_TOKEN
           value: {{ quote .Values.huggingface.token }}
         {{- else if contains "Llama" .Values.huggingface.model -}}

chart/values.schema.json

@@ -67,7 +67,7 @@
                         "llm_top_k": {
                             "type": "integer",
                             "title": "LLM Top K",
-                            "description": "The [top k](https://docs.vllm.ai/en/latest/dev/sampling_params.html) value to use when generating LLM responses (must be an integer).",
+                            "description": "The [top k](https://docs.vllm.ai/en/stable/dev/sampling_params.html) value to use when generating LLM responses (must be an integer).",
                             "default": -1,
                             "minimum": -1
                         },

chart/web-app/app.py

@@ -122,7 +122,7 @@ def inference(latest_message, history):
         raise gr.Error("Unexpected error encountered - see logs for details.")
 
 
-# UI colour theming
+# UI theming
 theme = gr.themes.Default(**settings.theme_params)
 if settings.theme_background_colour:
     theme.body_background_fill = settings.theme_background_colour