mirror of
https://github.com/katanemo/plano.git
synced 2026-06-20 15:28:07 +02:00
grafana: add PlanoAI dashboard and Brightstaff row in Intelligent Inference
Made-with: Cursor
This commit is contained in:
Adil Hafeez
parent
938f9c4bdf
commit
433840bb82
6 changed files with 3462 additions and 0 deletions
2582
config/grafana/intelligent_inference.json
Normal file
2582
config/grafana/intelligent_inference.json
Normal file
File diff suppressed because it is too large
Load diff
661
config/grafana/planoai_dashboard.json
Normal file
661
config/grafana/planoai_dashboard.json
Normal file
|
|
@ -0,0 +1,661 @@
|
|||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": { "type": "grafana", "uid": "-- Grafana --" },
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
}
|
||||
]
|
||||
},
|
||||
"description": "PlanoAI control + dataplane: HTTP RED, routing decisions, LLM upstream, and build/runtime for the brightstaff orchestrator. Metrics arrive in DO Prometheus under the gen_ai_otel_brightstaff_* prefix.",
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": null,
|
||||
"links": [
|
||||
{
|
||||
"asDropdown": true,
|
||||
"icon": "external link",
|
||||
"includeVars": false,
|
||||
"keepTime": true,
|
||||
"tags": ["inf"],
|
||||
"targetBlank": false,
|
||||
"title": "Inference dashboards",
|
||||
"tooltip": "",
|
||||
"type": "dashboards",
|
||||
"url": ""
|
||||
}
|
||||
],
|
||||
"liveNow": false,
|
||||
"panels": [
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": { "h": 1, "w": 24, "x": 0, "y": 0 },
|
||||
"id": 100,
|
||||
"panels": [],
|
||||
"title": "HTTP RED",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Requests-per-second served by brightstaff, broken out by handler. The orchestrator handler is the routing decision endpoint.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "axisLabel": "req/s", "drawStyle": "line", "fillOpacity": 10, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "reqps"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 8, "w": 12, "x": 0, "y": 1 },
|
||||
"id": 1,
|
||||
"options": {
|
||||
"legend": { "calcs": ["mean", "lastNotNull"], "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi", "sort": "desc" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (handler) (rate(gen_ai_otel_brightstaff_http_requests_total[1m]))",
|
||||
"legendFormat": "{{handler}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "RPS by handler",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Fraction of brightstaff responses that are 5xx, computed over 5m. Page-worthy when sustained above ~1%.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "thresholds" },
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{ "color": "green", "value": null },
|
||||
{ "color": "yellow", "value": 0.01 },
|
||||
{ "color": "red", "value": 0.05 }
|
||||
]
|
||||
},
|
||||
"unit": "percentunit"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 8, "w": 6, "x": 12, "y": 1 },
|
||||
"id": 2,
|
||||
"options": {
|
||||
"colorMode": "background",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": { "calcs": ["lastNotNull"], "fields": "", "values": false },
|
||||
"textMode": "auto"
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum(rate(gen_ai_otel_brightstaff_http_requests_total{status_class=\"5xx\"}[5m])) / clamp_min(sum(rate(gen_ai_otel_brightstaff_http_requests_total[5m])), 1)",
|
||||
"legendFormat": "5xx",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "5xx rate (5m)",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "In-flight HTTP requests by handler. Climbs before latency does when brightstaff is saturated.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 10, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "short"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 8, "w": 6, "x": 18, "y": 1 },
|
||||
"id": 3,
|
||||
"options": {
|
||||
"legend": { "displayMode": "list", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (handler) (gen_ai_otel_brightstaff_http_in_flight_requests)",
|
||||
"legendFormat": "{{handler}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "In-flight by handler",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "p50 / p95 / p99 of brightstaff HTTP request duration by handler, computed from the histogram buckets over 5m.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 5, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "s"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 24, "x": 0, "y": 9 },
|
||||
"id": 4,
|
||||
"options": {
|
||||
"legend": { "calcs": ["mean", "max"], "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi", "sort": "desc" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.50, sum by (le, handler) (rate(gen_ai_otel_brightstaff_http_request_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p50 {{handler}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.95, sum by (le, handler) (rate(gen_ai_otel_brightstaff_http_request_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p95 {{handler}}",
|
||||
"range": true,
|
||||
"refId": "B"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.99, sum by (le, handler) (rate(gen_ai_otel_brightstaff_http_request_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p99 {{handler}}",
|
||||
"range": true,
|
||||
"refId": "C"
|
||||
}
|
||||
],
|
||||
"title": "HTTP duration p50 / p95 / p99 by handler",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": { "h": 1, "w": 24, "x": 0, "y": 18 },
|
||||
"id": 200,
|
||||
"panels": [],
|
||||
"title": "Routing decisions",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Outcome of every /routing/* call. decision_served = a real model picked. no_candidates = orchestrator returned the sentinel `none`. policy_error = orchestrator failed.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 30,
|
||||
"lineWidth": 1,
|
||||
"showPoints": "never",
|
||||
"stacking": { "group": "A", "mode": "normal" }
|
||||
},
|
||||
"unit": "reqps"
|
||||
},
|
||||
"overrides": [
|
||||
{
|
||||
"matcher": { "id": "byName", "options": "decision_served" },
|
||||
"properties": [{ "id": "color", "value": { "mode": "fixed", "fixedColor": "green" } }]
|
||||
},
|
||||
{
|
||||
"matcher": { "id": "byName", "options": "no_candidates" },
|
||||
"properties": [{ "id": "color", "value": { "mode": "fixed", "fixedColor": "yellow" } }]
|
||||
},
|
||||
{
|
||||
"matcher": { "id": "byName", "options": "policy_error" },
|
||||
"properties": [{ "id": "color", "value": { "mode": "fixed", "fixedColor": "red" } }]
|
||||
}
|
||||
]
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 0, "y": 19 },
|
||||
"id": 5,
|
||||
"options": {
|
||||
"legend": { "calcs": ["mean", "lastNotNull"], "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi", "sort": "desc" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (outcome) (rate(gen_ai_otel_brightstaff_routing_service_requests_total[5m]))",
|
||||
"legendFormat": "{{outcome}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "/routing/* outcomes",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Decision rate broken out by the model the orchestrator selected. The 'none' series is the fallback bucket — any non-zero value here means the router could not classify intent or had no eligible candidates.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 10, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "reqps"
|
||||
},
|
||||
"overrides": [
|
||||
{
|
||||
"matcher": { "id": "byName", "options": "none" },
|
||||
"properties": [
|
||||
{ "id": "color", "value": { "mode": "fixed", "fixedColor": "red" } },
|
||||
{ "id": "custom.lineStyle", "value": { "fill": "dash", "dash": [10, 10] } }
|
||||
]
|
||||
}
|
||||
]
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 12, "y": 19 },
|
||||
"id": 6,
|
||||
"options": {
|
||||
"legend": { "calcs": ["mean", "lastNotNull"], "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi", "sort": "desc" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (selected_model) (rate(gen_ai_otel_brightstaff_router_decisions_total[5m]))",
|
||||
"legendFormat": "{{selected_model}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Decisions/sec by selected model",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Total decisions per model over the last 15 minutes. The relative bar lengths are the routing mix the orchestrator is producing.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"mappings": [],
|
||||
"unit": "short"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 0, "y": 28 },
|
||||
"id": 7,
|
||||
"options": {
|
||||
"displayMode": "gradient",
|
||||
"minVizHeight": 10,
|
||||
"minVizWidth": 0,
|
||||
"orientation": "horizontal",
|
||||
"reduceOptions": { "calcs": ["lastNotNull"], "fields": "", "values": false },
|
||||
"showUnfilled": true,
|
||||
"valueMode": "color"
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (selected_model) (increase(gen_ai_otel_brightstaff_router_decisions_total[15m]))",
|
||||
"legendFormat": "{{selected_model}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Model selection distribution (last 15m)",
|
||||
"type": "bargauge"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Fraction of decisions that fell back (orchestrator returned `none` or errored). High = router cannot classify intent or no candidates configured.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "thresholds" },
|
||||
"mappings": [],
|
||||
"max": 1,
|
||||
"min": 0,
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{ "color": "green", "value": null },
|
||||
{ "color": "yellow", "value": 0.1 },
|
||||
{ "color": "red", "value": 0.5 }
|
||||
]
|
||||
},
|
||||
"unit": "percentunit"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 6, "x": 12, "y": 28 },
|
||||
"id": 8,
|
||||
"options": {
|
||||
"colorMode": "background",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": { "calcs": ["lastNotNull"], "fields": "", "values": false },
|
||||
"textMode": "auto"
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (route) (rate(gen_ai_otel_brightstaff_router_decisions_total{fallback=\"true\"}[5m])) / clamp_min(sum by (route) (rate(gen_ai_otel_brightstaff_router_decisions_total[5m])), 1)",
|
||||
"legendFormat": "{{route}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Fallback rate by route",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "p95 of router decision time by route. The orchestrator includes embedding + classification + candidate scoring; this is the primary tail-latency signal for routing.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 5, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "s"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 6, "x": 18, "y": 28 },
|
||||
"id": 9,
|
||||
"options": {
|
||||
"legend": { "displayMode": "list", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.95, sum by (le, route) (rate(gen_ai_otel_brightstaff_router_decision_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p95 {{route}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.99, sum by (le, route) (rate(gen_ai_otel_brightstaff_router_decision_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p99 {{route}}",
|
||||
"range": true,
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"title": "Router decision p95/p99",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"collapsed": true,
|
||||
"gridPos": { "h": 1, "w": 24, "x": 0, "y": 37 },
|
||||
"id": 300,
|
||||
"panels": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "p95 LLM upstream call duration by provider/model. Populated once brightstaff exports llm_upstream_duration_seconds.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 5, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "s"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 0, "y": 38 },
|
||||
"id": 10,
|
||||
"options": {
|
||||
"legend": { "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.95, sum by (le, provider, model) (rate(gen_ai_otel_brightstaff_llm_upstream_duration_seconds_bucket[5m])))",
|
||||
"legendFormat": "p95 {{provider}}/{{model}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "LLM upstream p95 by provider/model",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Non-success error classes by provider. timeout/connect = network, 5xx/429 = provider, parse = body shape mismatch, stream = mid-stream disconnect.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 30,
|
||||
"lineWidth": 1,
|
||||
"showPoints": "never",
|
||||
"stacking": { "group": "A", "mode": "normal" }
|
||||
},
|
||||
"unit": "reqps"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 12, "y": 38 },
|
||||
"id": 11,
|
||||
"options": {
|
||||
"legend": { "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (provider, error_class) (rate(gen_ai_otel_brightstaff_llm_upstream_requests_total{error_class!=\"none\"}[5m]))",
|
||||
"legendFormat": "{{provider}} / {{error_class}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "LLM upstream errors by provider / class",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Streaming time-to-first-token p95. Empty if the route never streams.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": { "drawStyle": "line", "fillOpacity": 5, "lineWidth": 1, "showPoints": "never" },
|
||||
"unit": "s"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 0, "y": 47 },
|
||||
"id": 12,
|
||||
"options": {
|
||||
"legend": { "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "histogram_quantile(0.95, sum by (le, provider, model) (rate(gen_ai_otel_brightstaff_llm_time_to_first_token_seconds_bucket[5m])))",
|
||||
"legendFormat": "p95 {{provider}}/{{model}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Time-to-first-token p95 (streaming)",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Tokens/sec by provider/model/kind — proxy for cost. Stacked.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 30,
|
||||
"lineWidth": 1,
|
||||
"showPoints": "never",
|
||||
"stacking": { "group": "A", "mode": "normal" }
|
||||
},
|
||||
"unit": "short"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 9, "w": 12, "x": 12, "y": 47 },
|
||||
"id": 13,
|
||||
"options": {
|
||||
"legend": { "displayMode": "table", "placement": "bottom", "showLegend": true },
|
||||
"tooltip": { "mode": "multi" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (provider, model, kind) (rate(gen_ai_otel_brightstaff_llm_tokens_total[5m]))",
|
||||
"legendFormat": "{{provider}}/{{model}} {{kind}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Token throughput by provider / model / kind",
|
||||
"type": "timeseries"
|
||||
}
|
||||
],
|
||||
"title": "LLM upstream (forward-compat)",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": { "h": 1, "w": 24, "x": 0, "y": 56 },
|
||||
"id": 400,
|
||||
"panels": [],
|
||||
"title": "Build & runtime",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Brightstaff build metadata — version + git_sha labels from the build_info gauge.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "thresholds" },
|
||||
"mappings": [],
|
||||
"thresholds": { "mode": "absolute", "steps": [{ "color": "blue", "value": null }] }
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 6, "w": 12, "x": 0, "y": 57 },
|
||||
"id": 14,
|
||||
"options": {
|
||||
"cellHeight": "sm",
|
||||
"footer": { "countRows": false, "fields": "", "reducer": ["sum"], "show": false },
|
||||
"showHeader": true
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "gen_ai_otel_brightstaff_build_info",
|
||||
"format": "table",
|
||||
"instant": true,
|
||||
"legendFormat": "__auto",
|
||||
"range": false,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Brightstaff build info",
|
||||
"transformations": [
|
||||
{
|
||||
"id": "organize",
|
||||
"options": {
|
||||
"excludeByName": { "Time": true, "Value": true, "__name__": true, "job": false, "instance": false },
|
||||
"indexByName": {},
|
||||
"renameByName": {}
|
||||
}
|
||||
}
|
||||
],
|
||||
"type": "table"
|
||||
},
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"description": "Cumulative routing decisions per model (last 24h). A rough cost-allocation view.",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": { "mode": "palette-classic" },
|
||||
"mappings": [],
|
||||
"unit": "short"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": { "h": 6, "w": 12, "x": 12, "y": 57 },
|
||||
"id": 15,
|
||||
"options": {
|
||||
"displayLabels": ["name", "percent"],
|
||||
"legend": { "displayMode": "list", "placement": "right", "showLegend": true, "values": ["value"] },
|
||||
"pieType": "donut",
|
||||
"reduceOptions": { "calcs": ["lastNotNull"], "fields": "", "values": false },
|
||||
"tooltip": { "mode": "single", "sort": "none" }
|
||||
},
|
||||
"pluginVersion": "11.6.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": { "type": "prometheus", "uid": "prod-pandora-dc-all-proxy" },
|
||||
"editorMode": "code",
|
||||
"expr": "sum by (selected_model) (increase(gen_ai_otel_brightstaff_router_decisions_total[24h]))",
|
||||
"legendFormat": "{{selected_model}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Routing mix (24h)",
|
||||
"type": "piechart"
|
||||
}
|
||||
],
|
||||
"preload": false,
|
||||
"refresh": "30s",
|
||||
"schemaVersion": 41,
|
||||
"tags": ["plano", "brightstaff", "llm", "inf"],
|
||||
"templating": {
|
||||
"list": []
|
||||
},
|
||||
"time": { "from": "now-6h", "to": "now" },
|
||||
"timepicker": {
|
||||
"refresh_intervals": ["1m", "5m", "15m", "30m", "1h", "2h", "1d"]
|
||||
},
|
||||
"timezone": "utc",
|
||||
"title": "PlanoAI — Routing & Brightstaff",
|
||||
"uid": "planoai-routing",
|
||||
"version": 1,
|
||||
"weekStart": ""
|
||||
}
|
||||
Loading…
Add table
Add a link
Reference in a new issue