fix: 修正 logo 中 N 和 G 字母造型
N 添加对角线笔画（█▄ █），G 添加内横杠（█ ▀█）， 避免与 O 字母造型雷同。同步更新 ui.ts 中的硬编码 wordmark。
This commit is contained in:
airlongdian
287
infra/monitoring.ts
Normal file
287
infra/monitoring.ts
Normal file
@@ -0,0 +1,287 @@
|
||||
import { SECRET } from "./secret"
|
||||
import { domain } from "./stage"
|
||||
|
||||
const description = "Managed by SST (Don't edit in Honeycomb UI)"
|
||||
const alertsDisabled = $app.stage !== "production"
|
||||
|
||||
const webhookRecipient = new honeycombio.WebhookRecipient("DiscordAlerts", {
|
||||
name: $app.stage === "production" ? "Discord Alerts" : `Discord Alerts (${$app.stage})`,
|
||||
url: `https://${domain}/honeycomb/webhook`,
|
||||
secret: SECRET.HoneycombWebhookSecret.result,
|
||||
templates: [
|
||||
{
|
||||
type: "trigger",
|
||||
body: `{
|
||||
"url": {{ .Result.URL | quote }},
|
||||
"type": {{ .Vars.type | quote }},
|
||||
"name": {{ .Name | quote }},
|
||||
"status": {{ .Alert.Status | quote }},
|
||||
"isTest": {{ .Alert.IsTest }},
|
||||
"groups": {{ .Result.GroupsTriggered | toJson }}
|
||||
}`,
|
||||
},
|
||||
],
|
||||
variables: [
|
||||
{
|
||||
name: "type",
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
// Honeycomb can keep stale query-local calculated fields when the name is unchanged,
|
||||
// so tie the field name to the expression while avoiding deploy-to-deploy churn.
|
||||
// https://github.com/honeycombio/terraform-provider-honeycombio/issues/852
|
||||
const calculatedField = (field: { name: string; expression: string }) => ({
|
||||
...field,
|
||||
name: `${field.name}_${(
|
||||
Array.from(field.expression).reduce((result, char) => Math.imul(31, result) + char.charCodeAt(0), 0) >>> 0
|
||||
).toString(36)}`,
|
||||
})
|
||||
|
||||
const modelHttpErrorsQuery = (product: "go" | "zen") => {
|
||||
const filters = [
|
||||
{ column: "model", op: "exists" },
|
||||
{ column: "event_type", op: "=", value: "completions" },
|
||||
{ column: "user_agent", op: "contains", value: "opencode" },
|
||||
{ column: "isGoTier", op: "=", value: product === "go" ? "true" : "false" },
|
||||
]
|
||||
const failedHttpStatus = calculatedField({
|
||||
name: "is_failed_http_status",
|
||||
expression: `
|
||||
IF(
|
||||
AND(
|
||||
GTE($status, "400"),
|
||||
NOT(EQUALS($status, "401")),
|
||||
NOT(
|
||||
AND(
|
||||
EQUALS($status, "429"),
|
||||
OR(
|
||||
EQUALS($error.type, "GoUsageLimitError"),
|
||||
EQUALS($error.type, "FreeUsageLimitError")
|
||||
)
|
||||
)
|
||||
)
|
||||
),
|
||||
1,
|
||||
0
|
||||
)`,
|
||||
})
|
||||
|
||||
return honeycombio.getQuerySpecificationOutput({
|
||||
breakdowns: ["model"],
|
||||
calculatedFields: [failedHttpStatus],
|
||||
calculations: [
|
||||
{ op: "COUNT", name: "TOTAL", filterCombination: "AND", filters },
|
||||
{
|
||||
op: "SUM",
|
||||
name: "FAILED",
|
||||
column: failedHttpStatus.name,
|
||||
filterCombination: "AND",
|
||||
filters,
|
||||
},
|
||||
],
|
||||
formulas: [{ name: "ERROR", expression: "IF(GTE($TOTAL, 150), DIV($FAILED, $TOTAL), 0)" }],
|
||||
timeRange: 900,
|
||||
}).json
|
||||
}
|
||||
|
||||
const providerHttpErrorsQuery = () => {
|
||||
const filters = [
|
||||
{ column: "provider", op: "exists" },
|
||||
{ column: "user_agent", op: "contains", value: "opencode" },
|
||||
]
|
||||
const successHttpStatus = calculatedField({
|
||||
name: "is_success_http_status",
|
||||
expression: `IF(AND(GTE($status, "200"), LT($status, "400")), 1, 0)`,
|
||||
})
|
||||
const failedProviderHttpStatus = calculatedField({
|
||||
name: "is_failed_provider_http_status",
|
||||
expression: `IF(GT($llm.error.code, "400"), 1, 0)`,
|
||||
})
|
||||
|
||||
return honeycombio.getQuerySpecificationOutput({
|
||||
breakdowns: ["provider"],
|
||||
calculatedFields: [successHttpStatus, failedProviderHttpStatus],
|
||||
calculations: [
|
||||
{
|
||||
op: "SUM",
|
||||
name: "SUCCESS",
|
||||
column: successHttpStatus.name,
|
||||
filterCombination: "AND",
|
||||
filters: [...filters, { column: "event_type", op: "=", value: "completions" }],
|
||||
},
|
||||
{
|
||||
op: "SUM",
|
||||
name: "FAILED",
|
||||
column: failedProviderHttpStatus.name,
|
||||
filterCombination: "AND",
|
||||
filters: [
|
||||
...filters,
|
||||
{ column: "event_type", op: "=", value: "llm.error" },
|
||||
{ column: "llm.error.code", op: "!=", value: "404" },
|
||||
],
|
||||
},
|
||||
],
|
||||
formulas: [
|
||||
{ name: "ERROR", expression: "IF(GTE(SUM($SUCCESS, $FAILED), 150), DIV($FAILED, SUM($SUCCESS, $FAILED)), 0)" },
|
||||
],
|
||||
timeRange: 900,
|
||||
}).json
|
||||
}
|
||||
|
||||
const modelLowTpsQuery = (product: "go" | "zen") => {
|
||||
const filters = [
|
||||
{ column: "model", op: "exists" },
|
||||
{ column: "event_type", op: "=", value: "completions" },
|
||||
{ column: "user_agent", op: "contains", value: "opencode" },
|
||||
{ column: "isGoTier", op: "=", value: product === "go" ? "true" : "false" },
|
||||
{ column: "status", op: ">=", value: "200" },
|
||||
{ column: "status", op: "<", value: "400" },
|
||||
{ column: "tps.output", op: "exists" },
|
||||
]
|
||||
|
||||
return honeycombio.getQuerySpecificationOutput({
|
||||
breakdowns: ["model"],
|
||||
calculations: [
|
||||
{ op: "COUNT", name: "TOTAL", filterCombination: "AND", filters },
|
||||
{
|
||||
op: "P50",
|
||||
name: "TPS",
|
||||
column: "tps.output",
|
||||
filterCombination: "AND",
|
||||
filters,
|
||||
},
|
||||
],
|
||||
formulas: [{ name: "LOW_TPS", expression: "IF(GTE($TOTAL, 100), $TPS, 999)" }],
|
||||
timeRange: 1800,
|
||||
}).json
|
||||
}
|
||||
|
||||
new honeycombio.Trigger("IncreasedModelHttpErrorsGo", {
|
||||
name: "Increased Model HTTP Errors [Go]",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: modelHttpErrorsQuery("go"),
|
||||
alertType: "on_change",
|
||||
frequency: 300,
|
||||
thresholds: [{ op: ">=", value: 0.7, exceededLimit: 1 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "model_http_errors" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
new honeycombio.Trigger("IncreasedModelHttpErrorsZen", {
|
||||
name: "Increased Model HTTP Errors [Zen]",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: modelHttpErrorsQuery("zen"),
|
||||
alertType: "on_change",
|
||||
frequency: 300,
|
||||
thresholds: [{ op: ">=", value: 0.7, exceededLimit: 1 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "model_http_errors" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
new honeycombio.Trigger("LowModelTpsGo", {
|
||||
name: "Low Model TPS [Go]",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: modelLowTpsQuery("go"),
|
||||
alertType: "on_change",
|
||||
frequency: 600,
|
||||
thresholds: [{ op: "<=", value: 10, exceededLimit: 1 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "model_low_tps" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
new honeycombio.Trigger("LowModelTpsZen", {
|
||||
name: "Low Model TPS [Zen]",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: modelLowTpsQuery("zen"),
|
||||
alertType: "on_change",
|
||||
frequency: 600,
|
||||
thresholds: [{ op: "<=", value: 10, exceededLimit: 1 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "model_low_tps" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
new honeycombio.Trigger("IncreasedProviderHttpErrors", {
|
||||
name: "Increased Provider HTTP Errors",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: providerHttpErrorsQuery(),
|
||||
alertType: "on_change",
|
||||
frequency: 300,
|
||||
thresholds: [{ op: ">=", value: 0.7, exceededLimit: 1 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "provider_http_errors" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
new honeycombio.Trigger("IncreasedFreeTierRequests", {
|
||||
name: "Increased Free Tier Requests",
|
||||
description,
|
||||
disabled: alertsDisabled,
|
||||
queryJson: honeycombio.getQuerySpecificationOutput({
|
||||
calculations: [{ op: "COUNT" }],
|
||||
filters: [
|
||||
{ column: "event_type", op: "=", value: "completions" },
|
||||
{ column: "user_agent", op: "contains", value: "opencode" },
|
||||
{ column: "isFreeTier", op: "=", value: "true" },
|
||||
],
|
||||
timeRange: 3600,
|
||||
}).json,
|
||||
alertType: "on_change",
|
||||
frequency: 900,
|
||||
thresholds: [{ op: ">=", value: 50, exceededLimit: 1 }],
|
||||
baselineDetails: [{ type: "percentage", offsetMinutes: 1440 }],
|
||||
recipients: [
|
||||
{
|
||||
id: webhookRecipient.id,
|
||||
notificationDetails: [
|
||||
{
|
||||
variables: [{ name: "type", value: "custom" }],
|
||||
},
|
||||
],
|
||||
},
|
||||
],
|
||||
})
|
||||
Reference in New Issue
Block a user