@senad-d/observme 0.1.0

package/dashboards/observme-overview.json

@@ -0,0 +1,2129 @@
+{
+  "__requires": [
+    {
+      "id": "grafana",
+      "name": "Grafana",
+      "type": "grafana",
+      "version": "10.0.0"
+    },
+    {
+      "id": "prometheus",
+      "name": "Prometheus",
+      "type": "datasource",
+      "version": "1.0.0"
+    },
+    {
+      "id": "row",
+      "name": "Row",
+      "type": "panel",
+      "version": ""
+    },
+    {
+      "id": "stat",
+      "name": "Stat",
+      "type": "panel",
+      "version": ""
+    },
+    {
+      "id": "timeseries",
+      "name": "Time series",
+      "type": "panel",
+      "version": ""
+    },
+    {
+      "id": "bargauge",
+      "name": "Bar gauge",
+      "type": "panel",
+      "version": ""
+    },
+    {
+      "id": "text",
+      "name": "Text",
+      "type": "panel",
+      "version": ""
+    }
+  ],
+  "annotations": {
+    "list": []
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "id": null,
+  "links": [
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Cost",
+      "tooltip": "Open the Cost dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-cost/observme-cost?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Models",
+      "tooltip": "Open the Models dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-models/observme-models?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Latency",
+      "tooltip": "Open the Latency dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Tools",
+      "tooltip": "Open the Tools dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-tools/observme-tools?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Agents",
+      "tooltip": "Open the Agents dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Agent Node Graphs",
+      "tooltip": "Open the aggregate Agent Node Graphs dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-agent-node-graphs/observme-agent-node-graphs?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Trace Journey",
+      "tooltip": "Open the Trace Journey dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Errors",
+      "tooltip": "Open the Errors dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-errors/observme-errors?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Export Health",
+      "tooltip": "Open the Export Health dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-export-health/observme-export-health?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "SLO Health",
+      "tooltip": "Open the SLO Health dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-slo-health/observme-slo-health?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "Logs and LLM I/O",
+      "tooltip": "Open the privacy-safe Logs and LLM I/O dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-logs-llm/observme-logs-llm?${__url_time_range}"
+    },
+    {
+      "icon": "external link",
+      "includeVars": true,
+      "keepTime": true,
+      "targetBlank": false,
+      "title": "LLM Conversations",
+      "tooltip": "Open the redacted opt-in LLM Conversations dashboard with the current time range.",
+      "type": "link",
+      "url": "/d/observme-llm-conversations/observme-llm-conversations?${__url_time_range}"
+    }
+  ],
+  "liveNow": false,
+  "panels": [
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 100,
+      "panels": [],
+      "title": "Health",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Selected-range Observability Export health against the canonical 99% objective. This compact landing-page chip intentionally renders no observed workload as healthy idle; open SLO Health to distinguish no-workload scorecard state from measured attainment.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "red",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 0.95
+              },
+              {
+                "color": "green",
+                "value": 0.99
+              }
+            ]
+          },
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 0,
+        "y": 1
+      },
+      "id": 1,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Export Health",
+          "url": "/d/observme-export-health/observme-export-health?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open SLO Health",
+          "url": "/d/observme-slo-health/observme-slo-health?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "clamp_max(clamp_min(1 - ((sum(rate(observme_telemetry_dropped_total[$__range])) or vector(0)) / clamp_min((sum(rate(observme_events_observed_total[$__range])) or vector(0)), 1e-9)), 0), 1)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Export health",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Selected-range Agent Lineage health against the canonical 99% objective. The failure numerator uses lineage-missing spawn failures, orphan agents, and trace-context propagation failures; no spawns intentionally render as healthy idle on this landing page.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "red",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 0.95
+              },
+              {
+                "color": "green",
+                "value": 0.99
+              }
+            ]
+          },
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 4,
+        "y": 1
+      },
+      "id": 2,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open SLO Health",
+          "url": "/d/observme-slo-health/observme-slo-health?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "clamp_min(1 - (((sum(rate(observme_subagent_spawn_failures_total{reason=\"lineage_missing\"}[$__range])) or vector(0)) + (sum(rate(observme_orphan_agents_total[$__range])) or vector(0)) + (sum(rate(observme_trace_context_propagation_failures_total[$__range])) or vector(0))) / clamp_min((sum(rate(observme_subagents_spawned_total[$__range])) or vector(0)), 1e-9)), 0)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Agent lineage health",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Selected-range Workflow Completion health against the canonical 99% objective. Completed plus explicitly failed terminal workflows are compared with starts; no starts intentionally render as healthy idle on this landing page.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "red",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 0.95
+              },
+              {
+                "color": "green",
+                "value": 0.99
+              }
+            ]
+          },
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 8,
+        "y": 1
+      },
+      "id": 3,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Errors",
+          "url": "/d/observme-errors/observme-errors?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open SLO Health",
+          "url": "/d/observme-slo-health/observme-slo-health?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "clamp_max((((sum(increase(observme_workflows_completed_total[$__range])) or vector(0)) + (sum(increase(observme_workflow_errors_total[$__range])) or vector(0))) / clamp_min((sum(increase(observme_workflows_started_total[$__range])) or vector(0)), 1)) + ((sum(increase(observme_workflows_started_total[$__range])) or vector(0)) == bool 0), 1)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Workflow completion health",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Current cross-domain failure pressure divided by the main workload counters. This is a landing-page index, not a canonical SLO because the failure families have different denominators; 2%/5% are tunable triage assumptions. Zero can mean healthy idle or no failures.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 0.02
+              },
+              {
+                "color": "red",
+                "value": 0.05
+              }
+            ]
+          },
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 12,
+        "y": 1
+      },
+      "id": 4,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Errors",
+          "url": "/d/observme-errors/observme-errors?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "(((sum(rate(observme_workflow_errors_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_agent_run_errors_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_llm_errors_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_tool_failures_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_handler_errors_total[$__rate_interval])) or vector(0))) / clamp_min(((sum(rate(observme_turns_completed_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_llm_requests_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_tool_calls_total[$__rate_interval])) or vector(0)) + (sum(rate(observme_agent_runs_total[$__rate_interval])) or vector(0))), 1e-9))",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Error pressure",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Estimated LLM cost burn per hour over the selected range. The red 50 USD/h reference matches ObservMeCostSpike; the yellow 10 USD/h early warning is a tunable budget assumption. Grafana's positive $__range_s scalar is used directly in arithmetic.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 10
+              },
+              {
+                "color": "red",
+                "value": 50
+              }
+            ]
+          },
+          "unit": "currencyUSD"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 16,
+        "y": 1
+      },
+      "id": 5,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Cost",
+          "url": "/d/observme-cost/observme-cost?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "(sum(increase(observme_llm_cost_usd_total[$__range])) or vector(0)) / ($__range_s / 3600)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Cost burn / hour",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 turn latency as the first user-facing latency check. The 30s/60s thresholds are tunable operational assumptions, not alert defaults. A zero fallback keeps the chip compact during idle ranges but can also indicate absent duration instrumentation; verify Latency volume and Export Health when unexpected.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 30000
+              },
+              {
+                "color": "red",
+                "value": 60000
+              }
+            ]
+          },
+          "unit": "ms"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 20,
+        "y": 1
+      },
+      "id": 6,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Latency",
+          "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_turn_duration_ms_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Latency health",
+      "type": "stat"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 5
+      },
+      "id": 101,
+      "panels": [],
+      "title": "Workload",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Estimated currently active sessions from started minus shutdown counters. This is an estimate and can reset with process restarts; zero can mean healthy idle.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 0,
+        "y": 6
+      },
+      "id": 7,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "clamp_min((sum(observme_sessions_started_total) or vector(0)) - (sum(observme_sessions_shutdown_total) or vector(0)), 0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Active sessions estimate",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Completed turns per minute, shown as a compact KPI with sparkline. Zero means no completed turns were observed in the interval and can be healthy idle.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 4,
+        "y": 6
+      },
+      "id": 8,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Latency",
+          "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "60 * (sum(rate(observme_turns_completed_total[$__rate_interval])) or vector(0))",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Turns/min",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "LLM requests per minute. Zero means no requests were observed in the interval and can be healthy idle; use Models or Cost for provider/model breakdowns.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 8,
+        "y": 6
+      },
+      "id": 9,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Models",
+          "url": "/d/observme-models/observme-models?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "60 * (sum(rate(observme_llm_requests_total[$__rate_interval])) or vector(0))",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "LLM requests/min",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Tool calls per minute. Zero means no calls were observed in the interval and can be healthy idle; open Tools for reliability and slow-tool breakdowns.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 12,
+        "y": 6
+      },
+      "id": 10,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Tools",
+          "url": "/d/observme-tools/observme-tools?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "60 * (sum(rate(observme_tool_calls_total[$__rate_interval])) or vector(0))",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Tool calls/min",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Subagent spawns per minute across all roles and depths. Zero means no spawns were observed in the interval and can be healthy idle; open Agents or Trace Journey for lineage details.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 16,
+        "y": 6
+      },
+      "id": 11,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "60 * (sum(rate(observme_subagents_spawned_total[$__rate_interval])) or vector(0))",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Subagent spawns/min",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Currently active ObservMe agent runtimes. Zero can mean healthy idle. Red at 100 references the tunable ObservMeActiveAgentsStuckHigh default; yellow at 50 is an early-warning assumption. Compare sustained values with fan-out and depth.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 50
+              },
+              {
+                "color": "red",
+                "value": 100
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 4,
+        "x": 20,
+        "y": 6
+      },
+      "id": 12,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(observme_active_agents) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Active agents",
+      "type": "stat"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 11
+      },
+      "id": 102,
+      "panels": [],
+      "title": "Cost",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Total LLM cost over the selected dashboard range. Open Cost for provider/model attribution and budget details.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 10
+              },
+              {
+                "color": "red",
+                "value": 50
+              }
+            ]
+          },
+          "unit": "currencyUSD"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 4,
+        "x": 0,
+        "y": 12
+      },
+      "id": 13,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Cost",
+          "url": "/d/observme-cost/observme-cost?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_cost_usd_total[$__range])) or vector(0)",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Total cost",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Selected-range LLM token mix. Detailed provider/model token, cache, and reasoning analysis belongs in Cost, Models, and Logs and LLM I/O.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 10,
+        "x": 4,
+        "y": 12
+      },
+      "id": 14,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Cost",
+          "url": "/d/observme-cost/observme-cost?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Models",
+          "url": "/d/observme-models/observme-models?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Logs and LLM I/O",
+          "url": "/d/observme-logs-llm/observme-logs-llm?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "displayMode": "gradient",
+        "maxVizHeight": 300,
+        "minVizHeight": 16,
+        "minVizWidth": 8,
+        "namePlacement": "auto",
+        "orientation": "horizontal",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showUnfilled": true,
+        "sizing": "auto",
+        "valueMode": "color"
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_input_tokens_total[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "input",
+          "range": false,
+          "refId": "A"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_output_tokens_total[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "output",
+          "range": false,
+          "refId": "B"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_reasoning_tokens_total[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "reasoning",
+          "range": false,
+          "refId": "C"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_cache_read_tokens_total[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "cache read",
+          "range": false,
+          "refId": "D"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_cache_write_tokens_total[$__range])) or vector(0)",
+          "instant": true,
+          "legendFormat": "cache write",
+          "range": false,
+          "refId": "E"
+        }
+      ],
+      "title": "Token mix",
+      "type": "bargauge"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Total LLM cost trend without provider/model detail. Open Cost or Models when this sparkline rises unexpectedly.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "USD",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 20,
+            "gradientMode": "opacity",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "smooth",
+            "lineWidth": 2,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "never",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "currencyUSD"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 10,
+        "x": 14,
+        "y": 12
+      },
+      "id": 15,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Cost",
+          "url": "/d/observme-cost/observme-cost?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(increase(observme_llm_cost_usd_total[$__interval])) or vector(0)",
+          "legendFormat": "cost/interval",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Cost trend",
+      "type": "timeseries"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 19
+      },
+      "id": 103,
+      "panels": [],
+      "title": "Latency",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 turn latency trend with a compact sparkline. Zero can mean idle or missing duration buckets, so confirm volume in Latency and export liveness when unexpected. The 30s/60s thresholds are tunable assumptions.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 30000
+              },
+              {
+                "color": "red",
+                "value": 60000
+              }
+            ]
+          },
+          "unit": "ms"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 0,
+        "y": 20
+      },
+      "id": 16,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Latency",
+          "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_turn_duration_ms_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "p95 turn latency",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 LLM request latency trend. Zero can mean idle or missing duration buckets; open Latency for volume context and Models for provider/model comparison. The 30s/60s thresholds are tunable assumptions.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 30000
+              },
+              {
+                "color": "red",
+                "value": 60000
+              }
+            ]
+          },
+          "unit": "ms"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 8,
+        "y": 20
+      },
+      "id": 17,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Models",
+          "url": "/d/observme-models/observme-models?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Latency",
+          "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_llm_request_duration_ms_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "p95 LLM latency",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 tool-call latency trend. Zero can mean idle or missing duration buckets; open Latency for volume context and Tools for slow or failing tools. The 10s/30s thresholds are tunable assumptions.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "yellow",
+                "value": 10000
+              },
+              {
+                "color": "red",
+                "value": 30000
+              }
+            ]
+          },
+          "unit": "ms"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 16,
+        "y": 20
+      },
+      "id": 18,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Tools",
+          "url": "/d/observme-tools/observme-tools?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Latency",
+          "url": "/d/observme-latency/observme-latency?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_tool_duration_ms_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "p95 tool latency",
+      "type": "stat"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 25
+      },
+      "id": 104,
+      "panels": [],
+      "title": "Agent lineage",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Selected-range lineage failures: spawn failures, orphan agents, and trace-context propagation failures. Empty/zero is healthy for the range.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 1
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 6,
+        "x": 0,
+        "y": 26
+      },
+      "id": 19,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "(sum(increase(observme_subagent_spawn_failures_total[$__range])) or vector(0)) + (sum(increase(observme_orphan_agents_total[$__range])) or vector(0)) + (sum(increase(observme_trace_context_propagation_failures_total[$__range])) or vector(0))",
+          "instant": true,
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Lineage failures",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 agent-tree depth. The red reference at 5 matches the tunable ObservMeExcessiveAgentTreeDepth default. Zero can mean idle or missing tree-depth buckets; use Agents for workload context.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 5
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 6,
+        "x": 6,
+        "y": 26
+      },
+      "id": 20,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Agent Node Graphs",
+          "url": "/d/observme-agent-node-graphs/observme-agent-node-graphs?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_agent_tree_depth_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "p95 tree depth",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "p95 fan-out per parent operation. The red reference at 20 matches the tunable ObservMeRunawayAgentFanOut default. Zero can mean idle or missing fan-out buckets; use Agents for workload context.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 20
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 6,
+        "x": 12,
+        "y": 26
+      },
+      "id": 21,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        },
+        {
+          "targetBlank": false,
+          "title": "Open Trace Journey",
+          "url": "/d/observme-trace-journey/observme-trace-journey?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showPercentChange": false,
+        "textMode": "auto",
+        "wideLayout": true
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "histogram_quantile(0.95, sum(rate(observme_agent_fanout_count_bucket[$__rate_interval])) by (le)) or vector(0)",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "p95 fan-out",
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "description": "Current active agents by bounded role and depth. No rows can mean healthy idle or an absent active-agent gauge; compare with the ungrouped Active agents chip and Export Health before treating it as healthy.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 6,
+        "x": 18,
+        "y": 26
+      },
+      "id": 22,
+      "links": [
+        {
+          "targetBlank": false,
+          "title": "Open Agents",
+          "url": "/d/observme-agents/observme-agents?${__url_time_range}"
+        }
+      ],
+      "options": {
+        "displayMode": "gradient",
+        "maxVizHeight": 300,
+        "minVizHeight": 16,
+        "minVizWidth": 8,
+        "namePlacement": "auto",
+        "orientation": "horizontal",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showUnfilled": true,
+        "sizing": "auto",
+        "valueMode": "color"
+      },
+      "pluginVersion": "11.1.0",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prometheus"
+          },
+          "editorMode": "code",
+          "expr": "sum(observme_active_agents) by (agent_role, subagent_depth)",
+          "instant": true,
+          "legendFormat": "{{agent_role}} depth={{subagent_depth}}",
+          "range": false,
+          "refId": "A"
+        }
+      ],
+      "title": "Active agents by role/depth",
+      "type": "bargauge"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 32
+      },
+      "id": 105,
+      "panels": [],
+      "title": "Links",
+      "type": "row"
+    },
+    {
+      "datasource": null,
+      "description": "Operator navigation map. These links intentionally preserve the current Grafana time range.",
+      "fieldConfig": {
+        "defaults": {},
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 24,
+        "x": 0,
+        "y": 33
+      },
+      "id": 23,
+      "options": {
+        "code": {
+          "language": "plaintext",
+          "showLineNumbers": false,
+          "showMiniMap": false
+        },
+        "content": "# Drill-down dashboards\n\nUse this overview to decide where to go next, then open the focused dashboard with the same time range:\n\n- [Cost](/d/observme-cost/observme-cost?${__url_time_range}) for spend, burn rate, token mix, and cache context.\n- [Models](/d/observme-models/observme-models?${__url_time_range}) for provider/model request, error, latency, and stop-reason context.\n- [Latency](/d/observme-latency/observme-latency?${__url_time_range}) for percentile trends, volume context, and slow-stage investigation.\n- [Tools](/d/observme-tools/observme-tools?${__url_time_range}) for tool volume, failures, latency, sizes, and offenders.\n- [Agents](/d/observme-agents/observme-agents?${__url_time_range}) for subagent reliability, fan-out, depth, orphan, wait, and join behavior.\n- [Agent Node Graphs](/d/observme-agent-node-graphs/observme-agent-node-graphs?${__url_time_range}) for bounded aggregate topology and health.\n- [Trace Journey](/d/observme-trace-journey/observme-trace-journey?${__url_time_range}) for per-execution metric-to-log-to-trace inspection.\n- [Errors](/d/observme-errors/observme-errors?${__url_time_range}) for error-family pressure and parsed failure context.\n- [Export Health](/d/observme-export-health/observme-export-health?${__url_time_range}) for telemetry liveness, drops, redaction, export failures, and handler health.\n- [SLO Health](/d/observme-slo-health/observme-slo-health?${__url_time_range}) for canonical scorecards, burn rates, and no-workload states.\n- [Logs and LLM I/O](/d/observme-logs-llm/observme-logs-llm?${__url_time_range}) for privacy-safe lifecycle logs and token/size trends.\n- [LLM Conversations](/d/observme-llm-conversations/observme-llm-conversations?${__url_time_range}) for redacted opt-in prompt, response, and thinking timelines.",
+        "mode": "markdown"
+      },
+      "pluginVersion": "11.1.0",
+      "title": "Dashboard links",
+      "type": "text"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "tags": [
+    "observme",
+    "pi",
+    "opentelemetry",
+    "overview"
+  ],
+  "time": {
+    "from": "now-6h",
+    "to": "now"
+  },
+  "timezone": "browser",
+  "title": "ObservMe Overview",
+  "uid": "observme-overview",
+  "version": 3,
+  "weekStart": ""
+}