monitoring/grafana-dashboard.json

{
  "dashboard": {
    "id": null,
    "title": "MediGuard AI Monitoring",
    "tags": ["mediguard", "ai", "medical"],
    "timezone": "browser",
    "panels": [
      {
        "id": 1,
        "title": "API Request Rate",
        "type": "graph",
        "targets": [
          {
            "expr": "rate(http_requests_total[5m])",
            "legendFormat": "{{method}} {{endpoint}}"
          }
        ],
        "yAxes": [
          {
            "label": "Requests/sec"
          }
        ]
      },
      {
        "id": 2,
        "title": "Response Time",
        "type": "graph",
        "targets": [
          {
            "expr": "histogram_quantile(0.95, rate(http_request_duration_seconds_bucket[5m]))",
            "legendFormat": "95th percentile"
          },
          {
            "expr": "histogram_quantile(0.50, rate(http_request_duration_seconds_bucket[5m]))",
            "legendFormat": "50th percentile"
          }
        ],
        "yAxes": [
          {
            "label": "Seconds"
          }
        ]
      },
      {
        "id": 3,
        "title": "Error Rate",
        "type": "singlestat",
        "targets": [
          {
            "expr": "rate(http_requests_total{status=~\"5..\"}[5m]) / rate(http_requests_total[5m])",
            "legendFormat": "Error Rate"
          }
        ],
        "valueMaps": [
          {
            "value": "null",
            "text": "N/A"
          }
        ],
        "thresholds": "0.01,0.05,0.1",
        "unit": "percentunit"
      },
      {
        "id": 4,
        "title": "Active Users",
        "type": "singlestat",
        "targets": [
          {
            "expr": "active_users_total",
            "legendFormat": "Active Users"
          }
        ]
      },
      {
        "id": 5,
        "title": "Workflow Execution Time",
        "type": "graph",
        "targets": [
          {
            "expr": "histogram_quantile(0.95, rate(workflow_duration_seconds_bucket[5m]))",
            "legendFormat": "95th percentile"
          }
        ],
        "yAxes": [
          {
            "label": "Seconds"
          }
        ]
      },
      {
        "id": 6,
        "title": "Database Connections",
        "type": "graph",
        "targets": [
          {
            "expr": "opensearch_connections_active",
            "legendFormat": "OpenSearch"
          },
          {
            "expr": "redis_connections_active",
            "legendFormat": "Redis"
          }
        ]
      },
      {
        "id": 7,
        "title": "Memory Usage",
        "type": "graph",
        "targets": [
          {
            "expr": "process_resident_memory_bytes",
            "legendFormat": "RSS"
          }
        ],
        "yAxes": [
          {
            "label": "Bytes"
          }
        ]
      },
      {
        "id": 8,
        "title": "CPU Usage",
        "type": "graph",
        "targets": [
          {
            "expr": "rate(process_cpu_seconds_total[5m])",
            "legendFormat": "CPU"
          }
        ],
        "yAxes": [
          {
            "label": "Cores"
          }
        ]
      },
      {
        "id": 9,
        "title": "LLM Request Rate",
        "type": "graph",
        "targets": [
          {
            "expr": "rate(llm_requests_total[5m])",
            "legendFormat": "{{provider}}"
          }
        ],
        "yAxes": [
          {
            "label": "Requests/sec"
          }
        ]
      },
      {
        "id": 10,
        "title": "Cache Hit Rate",
        "type": "singlestat",
        "targets": [
          {
            "expr": "rate(cache_hits_total[5m]) / (rate(cache_hits_total[5m]) + rate(cache_misses_total[5m]))",
            "legendFormat": "Hit Rate"
          }
        ],
        "unit": "percentunit",
        "thresholds": "0.8,0.9,0.95"
      },
      {
        "id": 11,
        "title": "Agent Performance",
        "type": "table",
        "targets": [
          {
            "expr": "agent_execution_duration_seconds",
            "legendFormat": "{{agent_name}}",
            "format": "table"
          }
        ],
        "columns": [
          {
            "text": "Agent",
            "value": "agent_name"
          },
          {
            "text": "Avg Duration",
            "value": "avg"
          },
          {
            "text": "Success Rate",
            "value": "success_rate"
          }
        ]
      },
      {
        "id": 12,
        "title": "System Health",
        "type": "row"
      },
      {
        "id": 13,
        "title": "Service Status",
        "type": "stat",
        "targets": [
          {
            "expr": "up{job=\"mediguard\"}",
            "legendFormat": "{{instance}}"
          }
        ]
      }
    ],
    "time": {
      "from": "now-1h",
      "to": "now"
    },
    "refresh": "30s"
  }
}