toad-eye 2.7.0 → 2.9.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +80 -408
- package/dist/alerts/conditions.js +31 -6
- package/dist/alerts/grafana.js +6 -1
- package/dist/core/span-end-processor.d.ts +30 -0
- package/dist/core/span-end-processor.js +78 -0
- package/dist/core/tracer.js +35 -5
- package/dist/index.d.ts +1 -0
- package/dist/instrumentations/mcp.d.ts +12 -0
- package/dist/instrumentations/mcp.js +85 -0
- package/dist/mcp/client.d.ts +9 -0
- package/dist/mcp/client.js +137 -0
- package/dist/mcp/index.d.ts +3 -2
- package/dist/mcp/index.js +1 -0
- package/dist/mcp/metrics.d.ts +3 -1
- package/dist/mcp/metrics.js +15 -1
- package/dist/mcp/middleware.d.ts +5 -1
- package/dist/mcp/middleware.js +23 -8
- package/dist/mcp/spans.js +3 -3
- package/dist/mcp/types.d.ts +2 -21
- package/dist/types/config.d.ts +3 -0
- package/dist/types/index.d.ts +1 -0
- package/dist/types/metrics.d.ts +1 -0
- package/dist/types/metrics.js +1 -0
- package/dist/types/providers.d.ts +3 -1
- package/package.json +1 -1
- package/templates/alerts.example.yml +21 -0
- package/templates/grafana/dashboards/mcp-analytics.json +295 -0
- package/templates/grafana/dashboards/mcp-e2e.json +323 -0
- package/templates/grafana/dashboards/mcp-server.json +0 -19
|
@@ -0,0 +1,323 @@
|
|
|
1
|
+
{
|
|
2
|
+
"annotations": {
|
|
3
|
+
"list": []
|
|
4
|
+
},
|
|
5
|
+
"editable": true,
|
|
6
|
+
"fiscalYearStartMonth": 0,
|
|
7
|
+
"graphTooltip": 1,
|
|
8
|
+
"id": null,
|
|
9
|
+
"links": [],
|
|
10
|
+
"panels": [
|
|
11
|
+
{
|
|
12
|
+
"title": "Total Request Rate",
|
|
13
|
+
"description": "Combined rate of MCP client + server operations",
|
|
14
|
+
"type": "stat",
|
|
15
|
+
"gridPos": { "h": 4, "w": 6, "x": 0, "y": 0 },
|
|
16
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
17
|
+
"targets": [
|
|
18
|
+
{
|
|
19
|
+
"expr": "sum(rate(gen_ai_mcp_tool_calls_total[5m])) + sum(rate(gen_ai_client_requests_total[5m])) or vector(0)",
|
|
20
|
+
"refId": "A"
|
|
21
|
+
}
|
|
22
|
+
],
|
|
23
|
+
"fieldConfig": {
|
|
24
|
+
"defaults": {
|
|
25
|
+
"unit": "reqps",
|
|
26
|
+
"decimals": 2,
|
|
27
|
+
"thresholds": {
|
|
28
|
+
"steps": [
|
|
29
|
+
{ "color": "green", "value": null },
|
|
30
|
+
{ "color": "yellow", "value": 50 },
|
|
31
|
+
{ "color": "red", "value": 100 }
|
|
32
|
+
]
|
|
33
|
+
}
|
|
34
|
+
},
|
|
35
|
+
"overrides": []
|
|
36
|
+
}
|
|
37
|
+
},
|
|
38
|
+
{
|
|
39
|
+
"title": "LLM Avg Latency",
|
|
40
|
+
"description": "Average LLM call duration (client-side)",
|
|
41
|
+
"type": "stat",
|
|
42
|
+
"gridPos": { "h": 4, "w": 6, "x": 6, "y": 0 },
|
|
43
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
44
|
+
"targets": [
|
|
45
|
+
{
|
|
46
|
+
"expr": "sum(rate(gen_ai_client_operation_duration_sum[5m])) / sum(rate(gen_ai_client_operation_duration_count[5m])) or vector(0)",
|
|
47
|
+
"refId": "A"
|
|
48
|
+
}
|
|
49
|
+
],
|
|
50
|
+
"fieldConfig": {
|
|
51
|
+
"defaults": {
|
|
52
|
+
"unit": "ms",
|
|
53
|
+
"decimals": 0,
|
|
54
|
+
"thresholds": {
|
|
55
|
+
"steps": [
|
|
56
|
+
{ "color": "green", "value": null },
|
|
57
|
+
{ "color": "yellow", "value": 1000 },
|
|
58
|
+
{ "color": "red", "value": 5000 }
|
|
59
|
+
]
|
|
60
|
+
}
|
|
61
|
+
},
|
|
62
|
+
"overrides": []
|
|
63
|
+
}
|
|
64
|
+
},
|
|
65
|
+
{
|
|
66
|
+
"title": "Tool Avg Latency",
|
|
67
|
+
"description": "Average MCP tool execution duration (server-side)",
|
|
68
|
+
"type": "stat",
|
|
69
|
+
"gridPos": { "h": 4, "w": 6, "x": 12, "y": 0 },
|
|
70
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
71
|
+
"targets": [
|
|
72
|
+
{
|
|
73
|
+
"expr": "sum(rate(gen_ai_mcp_tool_duration_sum[5m])) / sum(rate(gen_ai_mcp_tool_duration_count[5m])) or vector(0)",
|
|
74
|
+
"refId": "A"
|
|
75
|
+
}
|
|
76
|
+
],
|
|
77
|
+
"fieldConfig": {
|
|
78
|
+
"defaults": {
|
|
79
|
+
"unit": "ms",
|
|
80
|
+
"decimals": 0,
|
|
81
|
+
"thresholds": {
|
|
82
|
+
"steps": [
|
|
83
|
+
{ "color": "green", "value": null },
|
|
84
|
+
{ "color": "yellow", "value": 500 },
|
|
85
|
+
{ "color": "red", "value": 2000 }
|
|
86
|
+
]
|
|
87
|
+
}
|
|
88
|
+
},
|
|
89
|
+
"overrides": []
|
|
90
|
+
}
|
|
91
|
+
},
|
|
92
|
+
{
|
|
93
|
+
"title": "End-to-End Error Rate",
|
|
94
|
+
"description": "Combined error rate across LLM + MCP",
|
|
95
|
+
"type": "stat",
|
|
96
|
+
"gridPos": { "h": 4, "w": 6, "x": 18, "y": 0 },
|
|
97
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
98
|
+
"targets": [
|
|
99
|
+
{
|
|
100
|
+
"expr": "(sum(rate(gen_ai_client_errors_total[5m])) + sum(rate(gen_ai_mcp_tool_errors_total[5m]))) / (sum(rate(gen_ai_client_requests_total[5m])) + sum(rate(gen_ai_mcp_tool_calls_total[5m]))) or vector(0)",
|
|
101
|
+
"refId": "A"
|
|
102
|
+
}
|
|
103
|
+
],
|
|
104
|
+
"fieldConfig": {
|
|
105
|
+
"defaults": {
|
|
106
|
+
"unit": "percentunit",
|
|
107
|
+
"decimals": 2,
|
|
108
|
+
"thresholds": {
|
|
109
|
+
"steps": [
|
|
110
|
+
{ "color": "green", "value": null },
|
|
111
|
+
{ "color": "yellow", "value": 0.05 },
|
|
112
|
+
{ "color": "red", "value": 0.1 }
|
|
113
|
+
]
|
|
114
|
+
}
|
|
115
|
+
},
|
|
116
|
+
"overrides": []
|
|
117
|
+
}
|
|
118
|
+
},
|
|
119
|
+
{
|
|
120
|
+
"title": "Latency Breakdown: LLM vs Tool",
|
|
121
|
+
"description": "Compare LLM call latency vs MCP tool execution latency over time",
|
|
122
|
+
"type": "timeseries",
|
|
123
|
+
"gridPos": { "h": 8, "w": 12, "x": 0, "y": 4 },
|
|
124
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
125
|
+
"targets": [
|
|
126
|
+
{
|
|
127
|
+
"expr": "sum(rate(gen_ai_client_operation_duration_sum[5m])) / sum(rate(gen_ai_client_operation_duration_count[5m]))",
|
|
128
|
+
"legendFormat": "LLM Avg Latency",
|
|
129
|
+
"refId": "A"
|
|
130
|
+
},
|
|
131
|
+
{
|
|
132
|
+
"expr": "sum(rate(gen_ai_mcp_tool_duration_sum[5m])) / sum(rate(gen_ai_mcp_tool_duration_count[5m]))",
|
|
133
|
+
"legendFormat": "Tool Avg Latency",
|
|
134
|
+
"refId": "B"
|
|
135
|
+
}
|
|
136
|
+
],
|
|
137
|
+
"fieldConfig": {
|
|
138
|
+
"defaults": {
|
|
139
|
+
"unit": "ms",
|
|
140
|
+
"custom": {
|
|
141
|
+
"drawStyle": "line",
|
|
142
|
+
"lineInterpolation": "smooth",
|
|
143
|
+
"fillOpacity": 20,
|
|
144
|
+
"pointSize": 5,
|
|
145
|
+
"showPoints": "auto"
|
|
146
|
+
}
|
|
147
|
+
},
|
|
148
|
+
"overrides": [
|
|
149
|
+
{
|
|
150
|
+
"matcher": { "id": "byName", "options": "LLM Avg Latency" },
|
|
151
|
+
"properties": [
|
|
152
|
+
{
|
|
153
|
+
"id": "color",
|
|
154
|
+
"value": { "fixedColor": "orange", "mode": "fixed" }
|
|
155
|
+
}
|
|
156
|
+
]
|
|
157
|
+
},
|
|
158
|
+
{
|
|
159
|
+
"matcher": { "id": "byName", "options": "Tool Avg Latency" },
|
|
160
|
+
"properties": [
|
|
161
|
+
{
|
|
162
|
+
"id": "color",
|
|
163
|
+
"value": { "fixedColor": "blue", "mode": "fixed" }
|
|
164
|
+
}
|
|
165
|
+
]
|
|
166
|
+
}
|
|
167
|
+
]
|
|
168
|
+
},
|
|
169
|
+
"options": {
|
|
170
|
+
"legend": {
|
|
171
|
+
"displayMode": "table",
|
|
172
|
+
"placement": "bottom",
|
|
173
|
+
"calcs": ["mean", "max"]
|
|
174
|
+
},
|
|
175
|
+
"tooltip": { "mode": "multi" }
|
|
176
|
+
}
|
|
177
|
+
},
|
|
178
|
+
{
|
|
179
|
+
"title": "Tool Call Waterfall by Tool",
|
|
180
|
+
"description": "Duration per tool over time — identify slow tools",
|
|
181
|
+
"type": "timeseries",
|
|
182
|
+
"gridPos": { "h": 8, "w": 12, "x": 12, "y": 4 },
|
|
183
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
184
|
+
"targets": [
|
|
185
|
+
{
|
|
186
|
+
"expr": "sum by (gen_ai_tool_name) (rate(gen_ai_mcp_tool_duration_sum[5m])) / sum by (gen_ai_tool_name) (rate(gen_ai_mcp_tool_duration_count[5m]))",
|
|
187
|
+
"legendFormat": "{{gen_ai_tool_name}}",
|
|
188
|
+
"refId": "A"
|
|
189
|
+
}
|
|
190
|
+
],
|
|
191
|
+
"fieldConfig": {
|
|
192
|
+
"defaults": {
|
|
193
|
+
"unit": "ms",
|
|
194
|
+
"custom": {
|
|
195
|
+
"drawStyle": "bars",
|
|
196
|
+
"fillOpacity": 60,
|
|
197
|
+
"stacking": { "mode": "normal" }
|
|
198
|
+
},
|
|
199
|
+
"color": { "mode": "palette-classic" }
|
|
200
|
+
},
|
|
201
|
+
"overrides": []
|
|
202
|
+
},
|
|
203
|
+
"options": {
|
|
204
|
+
"legend": {
|
|
205
|
+
"displayMode": "table",
|
|
206
|
+
"placement": "bottom",
|
|
207
|
+
"calcs": ["mean", "sum"]
|
|
208
|
+
},
|
|
209
|
+
"tooltip": { "mode": "multi" }
|
|
210
|
+
}
|
|
211
|
+
},
|
|
212
|
+
{
|
|
213
|
+
"title": "Error Propagation: LLM vs Tool",
|
|
214
|
+
"description": "Error rates split by LLM errors and MCP tool errors",
|
|
215
|
+
"type": "timeseries",
|
|
216
|
+
"gridPos": { "h": 8, "w": 12, "x": 0, "y": 12 },
|
|
217
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
218
|
+
"targets": [
|
|
219
|
+
{
|
|
220
|
+
"expr": "sum(rate(gen_ai_client_errors_total[5m]))",
|
|
221
|
+
"legendFormat": "LLM Errors",
|
|
222
|
+
"refId": "A"
|
|
223
|
+
},
|
|
224
|
+
{
|
|
225
|
+
"expr": "sum(rate(gen_ai_mcp_tool_errors_total[5m]))",
|
|
226
|
+
"legendFormat": "Tool Errors",
|
|
227
|
+
"refId": "B"
|
|
228
|
+
}
|
|
229
|
+
],
|
|
230
|
+
"fieldConfig": {
|
|
231
|
+
"defaults": {
|
|
232
|
+
"unit": "reqps",
|
|
233
|
+
"custom": {
|
|
234
|
+
"drawStyle": "line",
|
|
235
|
+
"lineInterpolation": "smooth",
|
|
236
|
+
"fillOpacity": 30,
|
|
237
|
+
"pointSize": 5,
|
|
238
|
+
"showPoints": "auto"
|
|
239
|
+
}
|
|
240
|
+
},
|
|
241
|
+
"overrides": [
|
|
242
|
+
{
|
|
243
|
+
"matcher": { "id": "byName", "options": "LLM Errors" },
|
|
244
|
+
"properties": [
|
|
245
|
+
{
|
|
246
|
+
"id": "color",
|
|
247
|
+
"value": { "fixedColor": "red", "mode": "fixed" }
|
|
248
|
+
}
|
|
249
|
+
]
|
|
250
|
+
},
|
|
251
|
+
{
|
|
252
|
+
"matcher": { "id": "byName", "options": "Tool Errors" },
|
|
253
|
+
"properties": [
|
|
254
|
+
{
|
|
255
|
+
"id": "color",
|
|
256
|
+
"value": { "fixedColor": "purple", "mode": "fixed" }
|
|
257
|
+
}
|
|
258
|
+
]
|
|
259
|
+
}
|
|
260
|
+
]
|
|
261
|
+
},
|
|
262
|
+
"options": {
|
|
263
|
+
"legend": {
|
|
264
|
+
"displayMode": "table",
|
|
265
|
+
"placement": "bottom",
|
|
266
|
+
"calcs": ["sum"]
|
|
267
|
+
},
|
|
268
|
+
"tooltip": { "mode": "multi" }
|
|
269
|
+
}
|
|
270
|
+
},
|
|
271
|
+
{
|
|
272
|
+
"title": "Cost: LLM Calls",
|
|
273
|
+
"description": "Total LLM cost over time — visible when tools call LLMs via sampling",
|
|
274
|
+
"type": "timeseries",
|
|
275
|
+
"gridPos": { "h": 8, "w": 12, "x": 12, "y": 12 },
|
|
276
|
+
"datasource": { "type": "prometheus", "uid": "prometheus" },
|
|
277
|
+
"targets": [
|
|
278
|
+
{
|
|
279
|
+
"expr": "sum by (gen_ai_request_model) (rate(gen_ai_client_request_cost_USD_sum[5m]))",
|
|
280
|
+
"legendFormat": "{{gen_ai_request_model}}",
|
|
281
|
+
"refId": "A"
|
|
282
|
+
}
|
|
283
|
+
],
|
|
284
|
+
"fieldConfig": {
|
|
285
|
+
"defaults": {
|
|
286
|
+
"unit": "currencyUSD",
|
|
287
|
+
"decimals": 6,
|
|
288
|
+
"custom": {
|
|
289
|
+
"drawStyle": "line",
|
|
290
|
+
"lineInterpolation": "smooth",
|
|
291
|
+
"fillOpacity": 15,
|
|
292
|
+
"pointSize": 5,
|
|
293
|
+
"showPoints": "auto"
|
|
294
|
+
},
|
|
295
|
+
"color": { "mode": "palette-classic" }
|
|
296
|
+
},
|
|
297
|
+
"overrides": []
|
|
298
|
+
},
|
|
299
|
+
"options": {
|
|
300
|
+
"legend": {
|
|
301
|
+
"displayMode": "table",
|
|
302
|
+
"placement": "bottom",
|
|
303
|
+
"calcs": ["sum"]
|
|
304
|
+
},
|
|
305
|
+
"tooltip": { "mode": "multi" }
|
|
306
|
+
}
|
|
307
|
+
}
|
|
308
|
+
],
|
|
309
|
+
"schemaVersion": 39,
|
|
310
|
+
"tags": ["toad-eye", "mcp", "e2e", "distributed-tracing"],
|
|
311
|
+
"templating": {
|
|
312
|
+
"list": []
|
|
313
|
+
},
|
|
314
|
+
"time": {
|
|
315
|
+
"from": "now-1h",
|
|
316
|
+
"to": "now"
|
|
317
|
+
},
|
|
318
|
+
"timepicker": {},
|
|
319
|
+
"timezone": "browser",
|
|
320
|
+
"title": "MCP End-to-End",
|
|
321
|
+
"uid": "toad-eye-mcp-e2e",
|
|
322
|
+
"version": 1
|
|
323
|
+
}
|
|
@@ -506,25 +506,6 @@
|
|
|
506
506
|
"tags": ["toad-eye", "mcp", "llm"],
|
|
507
507
|
"templating": {
|
|
508
508
|
"list": [
|
|
509
|
-
{
|
|
510
|
-
"name": "mcp_method",
|
|
511
|
-
"label": "MCP Method",
|
|
512
|
-
"type": "query",
|
|
513
|
-
"datasource": {
|
|
514
|
-
"type": "prometheus",
|
|
515
|
-
"uid": "prometheus"
|
|
516
|
-
},
|
|
517
|
-
"query": "label_values(gen_ai_mcp_tool_calls_total, mcp_method_name)",
|
|
518
|
-
"refresh": 2,
|
|
519
|
-
"includeAll": true,
|
|
520
|
-
"allValue": ".*",
|
|
521
|
-
"current": {
|
|
522
|
-
"text": "All",
|
|
523
|
-
"value": "$__all"
|
|
524
|
-
},
|
|
525
|
-
"multi": true,
|
|
526
|
-
"sort": 1
|
|
527
|
-
},
|
|
528
509
|
{
|
|
529
510
|
"name": "tool_name",
|
|
530
511
|
"label": "Tool",
|