chore(git): 同步主分支发布与业务变更

# Conflicts:
#	apps/api/go.mod
#	apps/api/go.sum
This commit is contained in:
2026-07-28 10:47:41 +08:00
238 changed files with 25370 additions and 2079 deletions
@@ -44,6 +44,20 @@ type Metrics struct {
billingEstimateFailed atomic.Uint64
billingIdempotentReplay atomic.Uint64
billingPricingUnavailable atomic.Uint64
asyncWorkerCapacity atomic.Int64
asyncWorkerDesiredCapacity atomic.Int64
asyncWorkerHardLimit atomic.Int64
asyncWorkerCapacityCapped atomic.Int64
asyncWorkerResizeSuccess atomic.Uint64
asyncWorkerRefreshFailed atomic.Uint64
asyncWorkerCreateFailed atomic.Uint64
asyncWorkerStartFailed atomic.Uint64
leaseRenewalSuccess atomic.Uint64
leaseRenewalFailure atomic.Uint64
leaseRenewalLost atomic.Uint64
taskEventDuplicate atomic.Uint64
taskEventUnknownType atomic.Uint64
taskEventBudgetExceeded atomic.Uint64
}
var processingDurationBounds = [...]time.Duration{
@@ -123,6 +137,52 @@ func (m *Metrics) ObserveBillingEvent(event string) {
}
}
func (m *Metrics) SetAsyncWorkerCapacity(current, desired, hardLimit int, capped bool) {
m.asyncWorkerCapacity.Store(int64(current))
m.asyncWorkerDesiredCapacity.Store(int64(desired))
m.asyncWorkerHardLimit.Store(int64(hardLimit))
cappedValue := int64(0)
if capped {
cappedValue = 1
}
m.asyncWorkerCapacityCapped.Store(cappedValue)
}
func (m *Metrics) ObserveAsyncWorkerResize(outcome string) {
switch outcome {
case "success":
m.asyncWorkerResizeSuccess.Add(1)
case "refresh_failed":
m.asyncWorkerRefreshFailed.Add(1)
case "create_failed":
m.asyncWorkerCreateFailed.Add(1)
case "start_failed":
m.asyncWorkerStartFailed.Add(1)
}
}
func (m *Metrics) ObserveConcurrencyLeaseRenewal(outcome string) {
switch outcome {
case "success":
m.leaseRenewalSuccess.Add(1)
case "lost":
m.leaseRenewalLost.Add(1)
default:
m.leaseRenewalFailure.Add(1)
}
}
func (m *Metrics) ObserveTaskEventSkip(reason string) {
switch reason {
case "duplicate":
m.taskEventDuplicate.Add(1)
case "budget_exceeded":
m.taskEventBudgetExceeded.Add(1)
default:
m.taskEventUnknownType.Add(1)
}
}
func (m *Metrics) Handler(provider MetricsSnapshotProvider, issuer, audience string, enabled bool) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
mode := "disabled"
@@ -212,6 +272,26 @@ func (m *Metrics) Handler(provider MetricsSnapshotProvider, issuer, audience str
plainCounter(w, "easyai_gateway_billing_estimate_failures_total", "Pricing estimate requests that failed.", m.billingEstimateFailed.Load())
plainCounter(w, "easyai_gateway_billing_idempotent_replays_total", "Generation requests replayed idempotently.", m.billingIdempotentReplay.Load())
plainCounter(w, "easyai_gateway_billing_pricing_unavailable_total", "Pricing requests rejected because no effective price was available.", m.billingPricingUnavailable.Load())
plainGauge(w, "easyai_gateway_async_worker_capacity", "Current River asynchronous worker capacity.", m.asyncWorkerCapacity.Load())
plainGauge(w, "easyai_gateway_async_worker_desired_capacity", "Uncapped asynchronous worker capacity derived from model and user-group policies.", m.asyncWorkerDesiredCapacity.Load())
plainGauge(w, "easyai_gateway_async_worker_hard_limit", "Per-process asynchronous worker safety limit.", m.asyncWorkerHardLimit.Load())
plainGauge(w, "easyai_gateway_async_worker_capacity_capped", "Whether desired asynchronous worker capacity is capped by the hard limit.", m.asyncWorkerCapacityCapped.Load())
outcomeCounters(w, "easyai_gateway_async_worker_resizes_total", "Asynchronous worker resize attempts by bounded outcome.", []outcomeValue{
{"success", m.asyncWorkerResizeSuccess.Load()},
{"refresh_failed", m.asyncWorkerRefreshFailed.Load()},
{"create_failed", m.asyncWorkerCreateFailed.Load()},
{"start_failed", m.asyncWorkerStartFailed.Load()},
})
outcomeCounters(w, "easyai_gateway_concurrency_lease_renewals_total", "Concurrency lease renewals by bounded outcome.", []outcomeValue{
{"success", m.leaseRenewalSuccess.Load()},
{"failure", m.leaseRenewalFailure.Load()},
{"lost", m.leaseRenewalLost.Load()},
})
outcomeCounters(w, "easyai_gateway_task_events_skipped_total", "Task events skipped before persistence by bounded reason.", []outcomeValue{
{"duplicate", m.taskEventDuplicate.Load()},
{"unknown_type", m.taskEventUnknownType.Load()},
{"budget_exceeded", m.taskEventBudgetExceeded.Load()},
})
})
}
@@ -249,3 +329,7 @@ func outcomeCounters(w http.ResponseWriter, name, help string, values []outcomeV
func plainCounter(w http.ResponseWriter, name, help string, value uint64) {
fmt.Fprintf(w, "# HELP %s %s\n# TYPE %s counter\n%s %d\n", name, help, name, name, value)
}
func plainGauge(w http.ResponseWriter, name, help string, value int64) {
fmt.Fprintf(w, "# HELP %s %s\n# TYPE %s gauge\n%s %d\n", name, help, name, name, value)
}
@@ -26,6 +26,12 @@ func TestMetricsExposeBoundedOutcomesAndState(t *testing.T) {
metrics.ObserveHeartbeat("accepted")
metrics.ObserveIntrospection("failed")
metrics.ObserveJWKSRefreshFailure("ssf")
metrics.SetAsyncWorkerCapacity(96, 128, 96, true)
metrics.ObserveAsyncWorkerResize("success")
metrics.ObserveConcurrencyLeaseRenewal("success")
metrics.ObserveConcurrencyLeaseRenewal("lost")
metrics.ObserveTaskEventSkip("duplicate")
metrics.ObserveTaskEventSkip("budget_exceeded")
recorder := httptest.NewRecorder()
metrics.Handler(metricsSnapshot{mode: "push_healthy", last: time.Now().Add(-time.Minute)}, "issuer", "audience", true).
@@ -42,6 +48,14 @@ func TestMetricsExposeBoundedOutcomesAndState(t *testing.T) {
`easyai_gateway_ssf_mode{mode="push_healthy"} 1`,
`easyai_gateway_oidc_introspection_total{outcome="failed"} 1`,
`easyai_gateway_jwks_refresh_failures_total{outcome="ssf"} 1`,
`easyai_gateway_async_worker_capacity 96`,
`easyai_gateway_async_worker_desired_capacity 128`,
`easyai_gateway_async_worker_capacity_capped 1`,
`easyai_gateway_async_worker_resizes_total{outcome="success"} 1`,
`easyai_gateway_concurrency_lease_renewals_total{outcome="success"} 1`,
`easyai_gateway_concurrency_lease_renewals_total{outcome="lost"} 1`,
`easyai_gateway_task_events_skipped_total{outcome="duplicate"} 1`,
`easyai_gateway_task_events_skipped_total{outcome="budget_exceeded"} 1`,
} {
if !strings.Contains(recorder.Body.String(), expected) {
t.Fatalf("missing metric %q in:\n%s", expected, recorder.Body.String())