include db for gitea build

internal/features/features.go

@@ -7,6 +7,8 @@ package features
 import (
 	"sort"
 	"time"
+
+	"git.tornberg.me/go-otel/internal/uaclass"
 )
 
 // Event is the subset of a normalized event used by the detector. Text fields
@@ -31,6 +33,14 @@ type Event struct {
 	Route            string
 	StatusCode       int
 	ExceptionType    string
+
+	// Trace-resolved context (see internal/tracectx). All optional: zero
+	// values mean "not resolved", which behaves like the pre-resolution data.
+	ClientIP    string // real client behind proxy hops, when known
+	UAClass     string // internal/uaclass class of the User-Agent
+	IsDuplicate bool   // repeats an ancestor span's failure; not a new request
+	IsRootCause bool   // error span with no erroring child
+	HasUpstream bool   // an ingress span that had a child in another service
 }
 
 // Config controls feature and label windows. A zero Config receives the
@@ -94,6 +104,18 @@ type Row struct {
 	TimeoutCount      int `json:"timeoutCount"`
 	CancellationCount int `json:"cancellationCount"`
 
+	// Logical-request view: DuplicateErrors are hops repeating an ancestor's
+	// failure, so LogicalErrors = ErrorCount - DuplicateErrors is the number of
+	// failed requests. DistinctClients / TopClientShare separate one noisy
+	// client from broad impact; BotErrorShare is the share of errors whose
+	// User-Agent classifies as a self-declared non-human client.
+	LogicalErrors   int     `json:"logicalErrors"`
+	DuplicateErrors int     `json:"duplicateErrors"`
+	RootCauseErrors int     `json:"rootCauseErrors"`
+	DistinctClients int     `json:"distinctClients"`
+	TopClientShare  float64 `json:"topClientShare"`
+	BotErrorShare   float64 `json:"botErrorShare"`
+
 	BaselineErrorRate float64 `json:"baselineErrorRate"`
 	BaselineSamples   int     `json:"baselineSamples"`
 	FutureEventCount  int     `json:"futureEventCount"`
@@ -189,10 +211,13 @@ type accumulator struct {
 	durations []float64
 	traces    map[string]struct{}
 	issues    map[string]struct{}
+	clients   map[string]int
+	botErrors int
+	classed   int
 }
 
 func newAccumulator(start, end time.Time, service, release string) *accumulator {
-	return &accumulator{value: Row{WindowStart: start, WindowEnd: end, Service: service, Release: release}, traces: map[string]struct{}{}, issues: map[string]struct{}{}}
+	return &accumulator{value: Row{WindowStart: start, WindowEnd: end, Service: service, Release: release}, traces: map[string]struct{}{}, issues: map[string]struct{}{}, clients: map[string]int{}}
 }
 
 func (a *accumulator) add(e Event) {
@@ -223,6 +248,20 @@ func (a *accumulator) add(e Event) {
 		return
 	}
 	a.value.ErrorCount++
+	if e.IsDuplicate {
+		a.value.DuplicateErrors++
+	} else if e.ClientIP != "" {
+		a.clients[e.ClientIP]++
+	}
+	if e.IsRootCause {
+		a.value.RootCauseErrors++
+	}
+	if e.UAClass != "" && !e.IsDuplicate {
+		a.classed++
+		if uaclass.IsBot(e.UAClass) {
+			a.botErrors++
+		}
+	}
 	if e.Source == "frontend" {
 		a.value.FrontendErrors++
 	} else {
@@ -252,6 +291,17 @@ func issueKey(e Event) string {
 func (a *accumulator) row() Row {
 	a.value.ErrorRate = rate(a.value.ErrorCount, a.value.TotalEvents)
 	a.value.UniqueTraceCount, a.value.UniqueIssueCount = len(a.traces), len(a.issues)
+	a.value.LogicalErrors = a.value.ErrorCount - a.value.DuplicateErrors
+	a.value.DistinctClients = len(a.clients)
+	top, known := 0, 0
+	for _, n := range a.clients {
+		known += n
+		if n > top {
+			top = n
+		}
+	}
+	a.value.TopClientShare = rate(top, known)
+	a.value.BotErrorShare = rate(a.botErrors, a.classed)
 	if len(a.durations) != 0 {
 		sort.Float64s(a.durations)
 		a.value.P50Duration = percentile(a.durations, .50)