Files
anthropic-proxy/internal/telemetry/metrics.go
T
Alexander eda66ff7d4 refactor(ratelimit): remove in-memory per-window token tracking
Token counts per rate limit window are now derived in Grafana via
increase(counter[5h/168h]) on the existing cumulative OTel counters.
Removes TokensIn/Out from Window, RecordTokens, setResetTime, and
the window_tokens observable gauges.
2026-04-14 13:49:05 +02:00

86 lines
3.0 KiB
Go

package telemetry
import (
"context"
"go.opentelemetry.io/otel/attribute"
"go.opentelemetry.io/otel/metric"
"github.com/fujin/anthropic-proxy/internal/ratelimit"
)
var (
RequestCounter metric.Int64Counter
RequestDuration metric.Float64Histogram
RequestBodySize metric.Int64Histogram
UpstreamErrors metric.Int64Counter
TokensInput metric.Int64Counter
TokensOutput metric.Int64Counter
CredentialCooldowns metric.Int64Counter
ActiveCredentials metric.Int64UpDownCounter
StreamRequests metric.Int64Counter
)
// InitMetrics creates all metric instruments from the given meter.
// If tracker is non-nil, registers observable gauges for per-window usage.
func InitMetrics(meter metric.Meter, tracker *ratelimit.Tracker) {
RequestCounter, _ = meter.Int64Counter("proxy.request.count",
metric.WithDescription("Total proxy requests"),
)
RequestDuration, _ = meter.Float64Histogram("proxy.request.duration_ms",
metric.WithDescription("Request latency in milliseconds"),
metric.WithUnit("ms"),
)
RequestBodySize, _ = meter.Int64Histogram("proxy.request.body_size_bytes",
metric.WithDescription("Request body size in bytes"),
metric.WithUnit("By"),
)
UpstreamErrors, _ = meter.Int64Counter("proxy.upstream.errors",
metric.WithDescription("Upstream error count"),
)
TokensInput, _ = meter.Int64Counter("proxy.tokens.input",
metric.WithDescription("Input tokens consumed"),
)
TokensOutput, _ = meter.Int64Counter("proxy.tokens.output",
metric.WithDescription("Output tokens consumed"),
)
CredentialCooldowns, _ = meter.Int64Counter("proxy.credential.cooldowns",
metric.WithDescription("Credential cooldown activations"),
)
ActiveCredentials, _ = meter.Int64UpDownCounter("proxy.credential.active",
metric.WithDescription("Currently active (non-cooldown) credentials"),
)
StreamRequests, _ = meter.Int64Counter("proxy.stream.requests",
metric.WithDescription("Streaming request count"),
)
if tracker == nil {
return
}
attr5h := attribute.String("window", "5h")
attr7d := attribute.String("window", "7d")
attrSonnet := attribute.String("window", "7d_sonnet")
meter.Float64ObservableGauge("proxy.usage.utilization",
metric.WithDescription("Current utilization % from API"),
metric.WithFloat64Callback(func(_ context.Context, o metric.Float64Observer) error {
o.Observe(tracker.FiveHour().Utilization, metric.WithAttributes(attr5h))
o.Observe(tracker.SevenDay().Utilization, metric.WithAttributes(attr7d))
o.Observe(tracker.Sonnet().Utilization, metric.WithAttributes(attrSonnet))
return nil
}),
)
meter.Int64ObservableGauge("proxy.usage.resets_at",
metric.WithDescription("Unix seconds when window resets"),
metric.WithInt64Callback(func(_ context.Context, o metric.Int64Observer) error {
o.Observe(tracker.FiveHour().ResetsAt.Unix(), metric.WithAttributes(attr5h))
o.Observe(tracker.SevenDay().ResetsAt.Unix(), metric.WithAttributes(attr7d))
o.Observe(tracker.Sonnet().ResetsAt.Unix(), metric.WithAttributes(attrSonnet))
return nil
}),
)
}