After story 1

monitoring/grafana/dashboards/active_learning.json

@@ -0,0 +1,97 @@
+{
+  "annotations": { "list": [] },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 1,
+  "id": null,
+  "links": [],
+  "panels": [
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 8, "x": 0, "y": 0 },
+      "id": 1,
+      "title": "Review Queue Size Over Time",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "queue_depth_review",
+          "legendFormat": "Queue Size"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 8, "x": 8, "y": 0 },
+      "id": 2,
+      "title": "Annotation Rate (Labels/Hour)",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "rate(videos_processed_total{routing_decision=\"REVIEW\"}[1h])",
+          "legendFormat": "Reviews/hour"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 8, "x": 16, "y": 0 },
+      "id": 3,
+      "title": "Model Version Timeline",
+      "type": "state-timeline",
+      "targets": [
+        {
+          "expr": "models{status=\"ACTIVE\"}",
+          "legendFormat": "{{version}} (ACTIVE)"
+        },
+        {
+          "expr": "models{status=\"CANDIDATE\"}",
+          "legendFormat": "{{version}} (CANDIDATE)"
+        },
+        {
+          "expr": "models{status=\"ARCHIVED\"}",
+          "legendFormat": "{{version}} (ARCHIVED)"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percentunit", "min": 0, "max": 1 } },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
+      "id": 4,
+      "title": "F1 Score by Model Version",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "models_f1_score",
+          "legendFormat": "{{version}}"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percentunit", "max": 1 } },
+      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 8 },
+      "id": 5,
+      "title": "ECE by Model Version",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "models_ece_score",
+          "legendFormat": "{{version}}"
+        }
+      ]
+    }
+  ],
+  "schemaVersion": 38,
+  "style": "dark",
+  "tags": ["VideoDetect"],
+  "templating": { "list": [] },
+  "time": { "from": "now-24h", "to": "now" },
+  "timepicker": {},
+  "title": "VideoDetect - Active Learning",
+  "uid": "videodetect-active-learning",
+  "version": 1
+}

monitoring/grafana/dashboards/processing_details.json

@@ -0,0 +1,115 @@
+{
+  "annotations": { "list": [] },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 1,
+  "id": null,
+  "links": [],
+  "panels": [
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "s" } },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
+      "id": 1,
+      "title": "Processing Duration per Video",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "histogram_quantile(0.5, rate(video_processing_duration_seconds_bucket[5m]))",
+          "legendFormat": "P50"
+        },
+        {
+          "expr": "histogram_quantile(0.95, rate(video_processing_duration_seconds_bucket[5m]))",
+          "legendFormat": "P95"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 6, "x": 12, "y": 0 },
+      "id": 2,
+      "title": "Frame Count per Video",
+      "type": "histogram",
+      "targets": [
+        {
+          "expr": "rate(frame_count_per_video[5m])",
+          "legendFormat": "Frames"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 6, "x": 18, "y": 0 },
+      "id": 3,
+      "title": "Face Count per Video",
+      "type": "histogram",
+      "targets": [
+        {
+          "expr": "rate(face_count_per_video[5m])",
+          "legendFormat": "Faces"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percentunit" } },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
+      "id": 4,
+      "title": "Confidence Score by Model Version",
+      "type": "barchart",
+      "targets": [
+        {
+          "expr": "avg by (model_version) (confidence_score)",
+          "legendFormat": "{{model_version}}"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percentunit" } },
+      "gridPos": { "h": 8, "w": 6, "x": 12, "y": 8 },
+      "id": 5,
+      "title": "Routing Decision Distribution",
+      "type": "piechart",
+      "targets": [
+        {
+          "expr": "increase(videos_processed_total{routing_decision=\"MATCH\"}[1h])",
+          "legendFormat": "MATCH"
+        },
+        {
+          "expr": "increase(videos_processed_total{routing_decision=\"REVIEW\"}[1h])",
+          "legendFormat": "REVIEW"
+        },
+        {
+          "expr": "increase(videos_processed_total{routing_decision=\"SKIP\"}[1h])",
+          "legendFormat": "SKIP"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "reqps" } },
+      "gridPos": { "h": 8, "w": 6, "x": 18, "y": 8 },
+      "id": 6,
+      "title": "Inference Throughput",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "rate(faces_detected_total[5m])",
+          "legendFormat": "Faces/sec"
+        }
+      ]
+    }
+  ],
+  "schemaVersion": 38,
+  "style": "dark",
+  "tags": ["VideoDetect"],
+  "templating": { "list": [] },
+  "time": { "from": "now-6h", "to": "now" },
+  "timepicker": {},
+  "title": "VideoDetect - Processing Details",
+  "uid": "videodetect-processing-details",
+  "version": 1
+}

monitoring/grafana/dashboards/system_overview.json

@@ -0,0 +1,108 @@
+{
+  "annotations": {
+    "list": []
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 1,
+  "id": null,
+  "links": [],
+  "panels": [
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "color": { "mode": "palette-classic" }, "unit": "videos/h" } },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
+      "id": 1,
+      "title": "Throughput (Videos/Hour)",
+      "type": "timeseries",
+      "targets": [
+        {
+          "expr": "rate(videos_processed_total[5m]) * 3600",
+          "legendFormat": "{{routing_decision}}"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 6, "x": 12, "y": 0 },
+      "id": 2,
+      "title": "Queue Depth",
+      "type": "stat",
+      "targets": [
+        { "expr": "queue_depth_pending", "legendFormat": "Pending" },
+        { "expr": "queue_depth_processing", "legendFormat": "Processing" },
+        { "expr": "queue_depth_review", "legendFormat": "Review" }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percent", "max": 100 } },
+      "gridPos": { "h": 8, "w": 6, "x": 18, "y": 0 },
+      "id": 3,
+      "title": "GPU Utilization",
+      "type": "gauge",
+      "targets": [
+        { "expr": "gpu_utilization_percent{gpu_id=\"0\"}", "legendFormat": "GPU 0" },
+        { "expr": "gpu_utilization_percent{gpu_id=\"1\"}", "legendFormat": "GPU 1" }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "bytes", "max": 24000000000 } },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
+      "id": 4,
+      "title": "GPU Memory Usage",
+      "type": "timeseries",
+      "targets": [
+        { "expr": "gpu_memory_used_bytes{gpu_id=\"0\"} / 1073741824", "legendFormat": "GPU 0 (GB)" },
+        { "expr": "gpu_memory_used_bytes{gpu_id=\"1\"} / 1073741824", "legendFormat": "GPU 1 (GB)" }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "short" } },
+      "gridPos": { "h": 8, "w": 6, "x": 12, "y": 8 },
+      "id": 5,
+      "title": "Error Rate (per 100 videos)",
+      "type": "stat",
+      "targets": [
+        {
+          "expr": "rate(inference_errors_total[5m]) / (rate(videos_processed_total[5m]) + 0.001) * 100",
+          "legendFormat": "Errors/100"
+        }
+      ]
+    },
+    {
+      "datasource": { "type": "prometheus", "uid": "prometheus" },
+      "fieldConfig": { "defaults": { "unit": "percentunit" } },
+      "gridPos": { "h": 8, "w": 6, "x": 18, "y": 8 },
+      "id": 6,
+      "title": "Confidence Distribution",
+      "type": "piechart",
+      "targets": [
+        {
+          "expr": "sum(increase(confidence_score_distribution_bucket{le=\"0.45\"}[1h]))",
+          "legendFormat": "SKIP (<0.45)"
+        },
+        {
+          "expr": "sum(increase(confidence_score_distribution_bucket{le=\"0.75\"}[1h])) - sum(increase(confidence_score_distribution_bucket{le=\"0.45\"}[1h]))",
+          "legendFormat": "REVIEW (0.45-0.75)"
+        },
+        {
+          "expr": "sum(increase(confidence_score_distribution_bucket{le=\"1\"}[1h])) - sum(increase(confidence_score_distribution_bucket{le=\"0.75\"}[1h]))",
+          "legendFormat": "MATCH (>0.75)"
+        }
+      ]
+    }
+  ],
+  "schemaVersion": 38,
+  "style": "dark",
+  "tags": ["VideoDetect"],
+  "templating": { "list": [] },
+  "time": { "from": "now-1h", "to": "now" },
+  "timepicker": {},
+  "title": "VideoDetect - System Overview",
+  "uid": "videodetect-system-overview",
+  "version": 1
+}

monitoring/grafana/provisioning/dashboards.yml

@@ -0,0 +1,11 @@
+apiVersion: 1
+
+providers:
+  - name: "default"
+    orgId: 1
+    folder: "VideoDetect"
+    type: file
+    disableDeletion: false
+    editable: true
+    options:
+      path: /etc/grafana/provisioning/dashboards

monitoring/grafana/provisioning/datasources.yml

@@ -0,0 +1,9 @@
+apiVersion: 1
+
+datasources:
+  - name: Prometheus
+    type: prometheus
+    access: proxy
+    url: http://prometheus:9090
+    isDefault: true
+    editable: false