Skip to content

add token metrics #41

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
whatattitude merged 1 commit into from
Jun 15, 2025
Merged

add token metrics #41

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
50 changes: 50 additions & 0 deletions metrics/metrics.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,11 @@ func RegisterMetrics(registry prometheus.Registerer) {
registry.MustRegister(relayRequestE2ESuccessCounter)
registry.MustRegister(relayRequestE2EFailedCounter)
registry.MustRegister(relayRequestE2EDurationObsever)
// token metrics
registry.MustRegister(inputTokensCounter)
registry.MustRegister(outputTokensCounter)
registry.MustRegister(cacheHitTokensCounter)
registry.MustRegister(inferenceTokensCounter)
}

var (
Expand Down Expand Up @@ -84,6 +89,34 @@ var (
},
[]string{"channel", "model", "group", "token_key", "token_name"},
)
// Token metrics
inputTokensCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: Namespace,
Name: "input_tokens_total",
Help: "Total number of input tokens processed",
}, []string{"channel", "model", "group", "user_id"})

outputTokensCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: Namespace,
Name: "output_tokens_total",
Help: "Total number of output tokens generated",
}, []string{"channel", "model", "group", "user_id"})

cacheHitTokensCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: Namespace,
Name: "cache_hit_tokens_total",
Help: "Total number of tokens served from cache",
}, []string{"channel", "model", "group", "user_id"})

inferenceTokensCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: Namespace,
Name: "inference_tokens_total",
Help: "Total number of tokens processed during inference",
}, []string{"channel", "model", "group", "user_id"})
)

func IncrementRelayRequestTotalCounter(channel, tag, baseURL, model, group string, add float64) {
Expand Down Expand Up @@ -121,3 +154,20 @@ func IncrementRelayRequestE2EFailedCounter(channel, model, group, code, tokenKey
func ObserveRelayRequestE2EDuration(channel, model, group, tokenKey, tokenName string, duration float64) {
relayRequestE2EDurationObsever.WithLabelValues(channel, model, group, tokenKey, tokenName).Observe(duration)
}

// Token metrics functions
func IncrementInputTokens(channel, model, group, userId string, add float64) {
inputTokensCounter.WithLabelValues(channel, model, group, userId).Add(add)
}

func IncrementOutputTokens(channel, model, group, userId string, add float64) {
outputTokensCounter.WithLabelValues(channel, model, group, userId).Add(add)
}

func IncrementCacheHitTokens(channel, model, group, userId string, add float64) {
cacheHitTokensCounter.WithLabelValues(channel, model, group, userId).Add(add)
}

func IncrementInferenceTokens(channel, model, group, userId string, add float64) {
inferenceTokensCounter.WithLabelValues(channel, model, group, userId).Add(add)
}
14 changes: 13 additions & 1 deletion relay/relay-text.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,6 +138,9 @@ func TextHelper(c *gin.Context, relayInfo *relaycommon.RelayInfo, textRequest *d
c.Set("prompt_tokens", promptTokens)
}

// Record input tokens metric
metrics.IncrementInputTokens(strconv.Itoa(relayInfo.ChannelId), textRequest.Model, relayInfo.Group, strconv.Itoa(relayInfo.UserId), float64(promptTokens))

priceData, err := helper.ModelPriceHelper(c, relayInfo, promptTokens, int(textRequest.MaxTokens))
if err != nil {
funcErr = service.OpenAIErrorWrapperLocal(err, "model_price_error", http.StatusInternalServerError)
Expand Down Expand Up @@ -291,6 +294,7 @@ func TextHelper(c *gin.Context, relayInfo *relaycommon.RelayInfo, textRequest *d
} else {
postConsumeQuota(c, relayInfo, usage.(*dto.Usage), preConsumedQuota, userQuota, priceData, "")
}

return nil
}

Expand Down Expand Up @@ -405,7 +409,6 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
promptTokens := usage.PromptTokens
cacheTokens := usage.PromptTokensDetails.CachedTokens
//

completionTokens := usage.CompletionTokens
thinkingTokens := usage.CompletionTokenDetails.ReasoningTokens
Expand Down Expand Up @@ -479,6 +482,15 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
logContent += ", " + extraContent
}

// Record token metrics
metrics.IncrementOutputTokens(strconv.Itoa(relayInfo.ChannelId), modelName, relayInfo.Group, strconv.Itoa(relayInfo.UserId), float64(completionTokens))

if cacheTokens > 0 {
metrics.IncrementCacheHitTokens(strconv.Itoa(relayInfo.ChannelId), modelName, relayInfo.Group, strconv.Itoa(relayInfo.UserId), float64(cacheTokens))
}

metrics.IncrementInferenceTokens(strconv.Itoa(relayInfo.ChannelId), modelName, relayInfo.Group, strconv.Itoa(relayInfo.UserId), float64(thinkingTokens))

other := service.GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, cacheTokens, cacheRatio, modelPrice)
model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, thinkingTokens, logModel,
tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
Expand Down