80 lines
2.4 KiB
JSON
80 lines
2.4 KiB
JSON
|
{
|
||
|
|
"annotations": { "list": [] },
|
||
|
|
"editable": true,
|
||
|
|
"graphTooltip": 1,
|
||
|
|
"id": null,
|
||
|
|
"links": [],
|
||
|
|
"panels": [
|
||
|
|
{
|
||
|
|
"datasource": { "type": "prometheus", "uid": "Prometheus" },
|
||
|
|
"fieldConfig": { "defaults": { "unit": "ms" }, "overrides": [] },
|
||
|
|
"gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
|
||
|
|
"id": 1,
|
||
|
|
"targets": [
|
||
|
|
{
|
||
|
|
"expr": "histogram_quantile(0.95, sum(rate(llm_call_duration_milliseconds_bucket[5m])) by (le, agent, call_type))",
|
||
|
|
"legendFormat": "{{agent}} / {{call_type}} p95",
|
||
|
|
"refId": "A"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"title": "LLM duration p95 by agent / call_type",
|
||
|
|
"type": "timeseries"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"datasource": { "type": "prometheus", "uid": "Prometheus" },
|
||
|
|
"fieldConfig": { "defaults": { "unit": "ms" }, "overrides": [] },
|
||
|
|
"gridPos": { "h": 8, "w": 12, "x": 12, "y": 0 },
|
||
|
|
"id": 2,
|
||
|
|
"targets": [
|
||
|
|
{
|
||
|
|
"expr": "histogram_quantile(0.50, sum(rate(llm_call_duration_milliseconds_bucket[5m])) by (le, call_type))",
|
||
|
|
"legendFormat": "{{call_type}} p50",
|
||
|
|
"refId": "A"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"title": "LLM duration p50 by call_type (json vs chat vs stream)",
|
||
|
|
"type": "timeseries"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"datasource": { "type": "prometheus", "uid": "Prometheus" },
|
||
|
|
"gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
|
||
|
|
"id": 3,
|
||
|
|
"targets": [
|
||
|
|
{
|
||
|
|
"expr": "sum(rate(llm_call_total[5m])) by (outcome, call_type)",
|
||
|
|
"legendFormat": "{{outcome}} / {{call_type}}",
|
||
|
|
"refId": "A"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"title": "LLM calls by outcome",
|
||
|
|
"type": "timeseries"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"datasource": { "type": "prometheus", "uid": "Prometheus" },
|
||
|
|
"gridPos": { "h": 8, "w": 12, "x": 12, "y": 8 },
|
||
|
|
"id": 4,
|
||
|
|
"targets": [
|
||
|
|
{
|
||
|
|
"expr": "sum(rate(llm_tokens_input_total[5m])) by (agent)",
|
||
|
|
"legendFormat": "in {{agent}}",
|
||
|
|
"refId": "A"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"expr": "sum(rate(llm_tokens_output_total[5m])) by (agent)",
|
||
|
|
"legendFormat": "out {{agent}}",
|
||
|
|
"refId": "B"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"title": "LLM tokens/min",
|
||
|
|
"type": "timeseries"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"schemaVersion": 39,
|
||
|
|
"tags": ["life-echo", "llm"],
|
||
|
|
"templating": { "list": [] },
|
||
|
|
"time": { "from": "now-1h", "to": "now" },
|
||
|
|
"title": "Life Echo LLM",
|
||
|
|
"uid": "life-echo-llm",
|
||
|
|
"version": 1
|
||
|
|
}
|