gigastt 1.0.2

{
  "__comment": "Starter Grafana dashboard for gigastt. Import via Dashboards > Import in the Grafana UI, or place under your provisioning/dashboards/ directory. Requires gigastt >= 0.8.0 with --metrics enabled and a Prometheus datasource named 'Prometheus'.",
  "title": "gigastt",
  "uid": "gigastt-overview",
  "schemaVersion": 39,
  "version": 1,
  "refresh": "30s",
  "time": { "from": "now-1h", "to": "now" },
  "timezone": "browser",
  "tags": ["gigastt", "speech-to-text"],
  "panels": [
    {
      "id": 1,
      "title": "Request Rate",
      "type": "timeseries",
      "gridPos": { "x": 0, "y": 0, "w": 12, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "reqps",
          "custom": { "lineWidth": 2 }
        }
      },
      "targets": [
        {
          "expr": "sum(rate(gigastt_http_requests_total[1m])) by (path, method)",
          "legendFormat": "{{method}} {{path}}",
          "refId": "A"
        }
      ]
    },
    {
      "id": 2,
      "title": "Request Duration (p50 / p95 / p99)",
      "type": "timeseries",
      "gridPos": { "x": 12, "y": 0, "w": 12, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "s",
          "custom": { "lineWidth": 2 }
        }
      },
      "targets": [
        {
          "expr": "histogram_quantile(0.50, sum(rate(gigastt_http_request_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p50",
          "refId": "A"
        },
        {
          "expr": "histogram_quantile(0.95, sum(rate(gigastt_http_request_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p95",
          "refId": "B"
        },
        {
          "expr": "histogram_quantile(0.99, sum(rate(gigastt_http_request_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p99",
          "refId": "C"
        }
      ]
    },
    {
      "id": 3,
      "title": "Error Rate (5xx)",
      "type": "timeseries",
      "gridPos": { "x": 0, "y": 8, "w": 12, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "percentunit",
          "custom": { "lineWidth": 2 },
          "thresholds": {
            "mode": "absolute",
            "steps": [
              { "color": "green", "value": null },
              { "color": "red", "value": 0.05 }
            ]
          }
        }
      },
      "targets": [
        {
          "expr": "sum(rate(gigastt_http_requests_total{status=~\"5..\"}[5m])) / sum(rate(gigastt_http_requests_total[5m]))",
          "legendFormat": "5xx rate",
          "refId": "A"
        }
      ]
    },
    {
      "id": 4,
      "title": "Pool Available Sessions",
      "type": "timeseries",
      "gridPos": { "x": 12, "y": 8, "w": 12, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "short",
          "custom": { "lineWidth": 2 },
          "thresholds": {
            "mode": "absolute",
            "steps": [
              { "color": "red", "value": 0 },
              { "color": "yellow", "value": 1 },
              { "color": "green", "value": 2 }
            ]
          }
        }
      },
      "targets": [
        {
          "expr": "gigastt_pool_available",
          "legendFormat": "available",
          "refId": "A"
        }
      ]
    },
    {
      "id": 5,
      "title": "WebSocket Active Connections",
      "type": "timeseries",
      "gridPos": { "x": 0, "y": 16, "w": 8, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "short",
          "custom": { "lineWidth": 2 }
        }
      },
      "targets": [
        {
          "expr": "gigastt_ws_active_connections",
          "legendFormat": "active",
          "refId": "A"
        }
      ]
    },
    {
      "id": 6,
      "title": "Inference Duration (p50 / p95 / p99)",
      "type": "timeseries",
      "gridPos": { "x": 8, "y": 16, "w": 8, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "s",
          "custom": { "lineWidth": 2 }
        }
      },
      "targets": [
        {
          "expr": "histogram_quantile(0.50, sum(rate(gigastt_inference_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p50",
          "refId": "A"
        },
        {
          "expr": "histogram_quantile(0.95, sum(rate(gigastt_inference_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p95",
          "refId": "B"
        },
        {
          "expr": "histogram_quantile(0.99, sum(rate(gigastt_inference_duration_seconds_bucket[5m])) by (le))",
          "legendFormat": "p99",
          "refId": "C"
        }
      ]
    },
    {
      "id": 7,
      "title": "Rate Limit Rejections",
      "type": "timeseries",
      "gridPos": { "x": 16, "y": 16, "w": 8, "h": 8 },
      "datasource": { "type": "prometheus", "uid": "prometheus" },
      "fieldConfig": {
        "defaults": {
          "unit": "reqps",
          "custom": { "lineWidth": 2 }
        }
      },
      "targets": [
        {
          "expr": "sum(rate(gigastt_rate_limit_rejections_total[1m]))",
          "legendFormat": "rejections/s",
          "refId": "A"
        }
      ]
    }
  ]
}