Files
new-api/service/responses_usage.go
T
CaIon bb28f2ccad feat(tokenkit): move token counting into a module calibrated on agent traffic
Token estimates badly undercounted agent requests. Claude tool
definitions, Responses function calls and their outputs, Chat assistant
tool calls, and Gemini system instructions, function calls, function
responses, and tools never reached the counter; Chat tool schemas were
counted as Go map syntax and Claude tool results as escaped JSON.
Measured against upstream usage on 54 real Claude Code and Codex
conversation windows, native prompt estimates were 31-93% low, and the
Claude and Gemini completion fallbacks about 90% low because streamed
tool calls were not counted.

All of these now reach the counter. The completion fallbacks count tool
names and arguments, plus 40 tokens per Claude tool_use block. The
estimator weights are refitted per tokenizer: Claude before Opus 4.7,
Claude 4.7 and later (about 30% more tokens for the same text), GPT, and
Gemini. Prompt estimates now land within 1% bias and 3-4% mean error for
each family.

Token counting moves into tokenkit, an independent module like
relaykit, so other gateways can use it. It also replaces the flat
520-token image estimate with each provider's sizing rules, checked
against upstream usage on sizes outside the test table: OpenAI patch and
tile rules per model and detail level (gpt-5.5 and later default to
original detail, and their high detail is billed as measured, not as
documented), Claude's two resolution tiers, and Gemini 3 media
resolution and Gemini 2.x tiles. Chat image_url parts without detail
no longer default to high in ParseContent, so they estimate as the
upstream default auto; no converter sent that value upstream.

tiktoken-go moves to v0.8.1, about twice as fast on ordinary text. It
requires Go 1.26, so the root module and tokenkit move to Go 1.26;
relaykit stays on 1.25.1. In every version BPE cost grows with the
square of a pre-tokenized piece, so 100,000 Chinese characters without
punctuation took 21.6 s to count during pre-consume. Count now cuts
single-class runs longer than 1024 bytes, which leaves ordinary text
unchanged and brings that input down to 106 ms. The encoder cache no
longer grows with client-supplied model names.
2026-10-07 21:26:35 +08:00

169 lines
6.2 KiB
Go

package service
import (
"fmt"
"strings"
"github.com/QuantumNous/new-api/common"
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/relaykit/relayconvert"
"github.com/QuantumNous/new-api/tokenkit"
)
// ResponsesUsageAccumulator owns the accounting facts for one Responses stream.
// HTTP SSE and WebSocket transports feed the same events into it, then settle
// Finish's usage through the normal text billing path, including interrupted
// streams. Observe and Finish must be called by the same stream owner.
type ResponsesUsageAccumulator struct {
info *relaycommon.RelayInfo
usage *dto.Usage
outputText strings.Builder
imageCounter relaycommon.ImageGenerationCallCounter
imageCommitted bool
started bool
finished bool
}
func NewResponsesUsageAccumulator(info *relaycommon.RelayInfo) *ResponsesUsageAccumulator {
return &ResponsesUsageAccumulator{info: info, usage: &dto.Usage{}}
}
func (a *ResponsesUsageAccumulator) Observe(event *dto.ResponsesStreamResponse) {
if a == nil || event == nil || a.finished {
return
}
if event.Response != nil {
a.info.ObserveResponseModel(event.Response.Model)
}
a.started = true
ObserveResponsesOutcome(a.info, event)
switch event.Type {
case "response.completed", "response.done", "response.failed", "response.incomplete", "response.cancelled", "response.canceled":
if event.Response != nil {
ApplyResponsesUsage(a.usage, event.Response.Usage)
if a.outputText.Len() == 0 {
// Some upstreams carry the output only on the terminal event.
a.outputText.WriteString(relayconvert.ExtractOutputTextFromResponses(event.Response))
}
}
if a.imageCommitted {
return
}
failed := event.Type != "response.completed" && event.Type != "response.done"
if failed || (event.Response != nil && relaycommon.IsNonBillableResponsesStatus(event.Response.Status)) {
a.imageCounter.Reset()
} else if event.Response != nil {
for i := range event.Response.Output {
a.imageCounter.Observe(&event.Response.Output[i], &i)
}
}
a.imageCounter.Commit(a.info)
a.imageCommitted = true
case "response.output_text.delta", "response.function_call_arguments.delta",
"response.reasoning_summary_text.delta", "response.reasoning_text.delta", "response.refusal.delta":
// Every delta kind here is generated output that upstream bills as
// output tokens, so all of them feed the missing-usage estimate.
a.outputText.WriteString(event.Delta)
case dto.ResponsesOutputTypeItemDone:
if event.Item == nil {
return
}
switch event.Item.Type {
case dto.BuildInCallWebSearchCall, dto.BuildInCallFileSearchCall, dto.BuildInCallFunctionCall:
a.info.CountBillableToolCall(event.Item.Type, event.Item.Name)
case dto.ResponsesOutputTypeImageGenerationCall:
if !a.imageCommitted {
a.imageCounter.Observe(event.Item, event.OutputIndex)
}
}
}
}
func (a *ResponsesUsageAccumulator) Finish() *dto.Usage {
if a.finished {
return a.usage
}
a.finished = true
// A final image item can already have reached the client before the stream
// disconnects. Explicit failed/incomplete terminals reset and commit zero in
// Observe; otherwise retain completed tool usage even without a terminal.
if !a.imageCommitted {
a.imageCounter.Commit(a.info)
a.imageCommitted = true
}
if a.usage.CompletionTokens == 0 {
if output := a.outputText.String(); output != "" {
a.usage.CompletionTokens = tokenkit.Count(a.info.GetUpstreamModelName(), output)
}
}
// Upstream bills the prompt as soon as it starts generating, so a stream
// that produced any event but no usage still owes its input tokens unless
// upstream reported an explicit failure.
billsPrompt := a.usage.CompletionTokens != 0 || (a.started && !a.info.StreamStatus.ResponseFailed())
if a.usage.PromptTokens == 0 && billsPrompt {
a.usage.PromptTokens = a.info.GetEstimatePromptTokens()
}
a.usage.TotalTokens = a.usage.PromptTokens + a.usage.CompletionTokens
if a.usage.BillingUsage != nil {
a.usage.BillingUsage = dto.CloneBillingUsageWithEstimatedCompletion(a.usage.BillingUsage, a.usage.CompletionTokens)
}
return a.usage
}
// ObserveResponsesOutcome records the protocol outcome of one Responses event
// on the stream status for health classification. Only codes and types are
// kept; messages never leave the event.
func ObserveResponsesOutcome(info *relaycommon.RelayInfo, event *dto.ResponsesStreamResponse) {
if info == nil || info.StreamStatus == nil || event == nil {
return
}
var responseStatus string
if event.Response != nil {
_ = common.Unmarshal(event.Response.Status, &responseStatus)
}
switch {
case event.Type == "error" || event.Type == "response.failed" || event.Type == "response.error" || responseStatus == "failed":
code, errorType := event.Code, ""
if event.Response != nil {
if oaiErr := event.Response.GetOpenAIError(); oaiErr != nil {
if oaiErr.Code != nil {
code = fmt.Sprint(oaiErr.Code)
}
errorType = oaiErr.Type
}
}
info.StreamStatus.MarkFailed(code, errorType, 0)
case event.Type == "response.incomplete" || responseStatus == "incomplete":
reason := ""
if event.Response != nil && event.Response.IncompleteDetails != nil {
reason = event.Response.IncompleteDetails.Reason
}
info.StreamStatus.MarkIncomplete(reason)
case event.Type == "response.cancelled" || event.Type == "response.canceled" || responseStatus == "cancelled":
info.StreamStatus.MarkCancelled()
case event.Type == "response.completed" || event.Type == "response.done" || responseStatus == "completed":
info.StreamStatus.MarkCompleted()
}
}
func ApplyResponsesUsage(dst *dto.Usage, src *dto.Usage) {
if dst == nil || src == nil {
return
}
incoming := relayconvert.NormalizeResponsesUsage(src)
if src.InputTokensDetails != nil {
inputDetails := *src.InputTokensDetails
incoming.InputTokensDetails = &inputDetails
}
if src.OutputTokensDetails != nil {
incoming.CompletionTokenDetails = *src.OutputTokensDetails
}
incoming.PromptCacheHitTokens = src.PromptCacheHitTokens
dto.MergeUsageNonZero(dst, incoming)
outputDetails := dst.CompletionTokenDetails
if outputDetails != (dto.OutputTokenDetails{}) {
dst.OutputTokensDetails = &outputDetails
}
}