feat: sync deployed v1.0.0-rc.38-custom.42
Sync production source including routing, billing, Astra detection and public uptime. Update custom build instructions and license packaging.
This commit is contained in:
1 parent
aa6c3f50f4
commit
ceddfd3d6b
828 files changed
+122253
-13702
No files matched your search
@@ -1,12 +1,15 @@
|
||||
package openai
|
||||
|
||||
import (
|
||||
"bufio"
|
||||
"bytes"
|
||||
"fmt"
|
||||
"io"
|
||||
"net/http"
|
||||
"strings"
|
||||
|
||||
"github.com/QuantumNous/new-api/common"
|
||||
"github.com/QuantumNous/new-api/constant"
|
||||
"github.com/QuantumNous/new-api/logger"
|
||||
relaycommon "github.com/QuantumNous/new-api/relay/common"
|
||||
"github.com/QuantumNous/new-api/relay/helper"
|
||||
@@ -16,10 +19,25 @@ import (
|
||||
"github.com/QuantumNous/new-api/service"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
)
|
||||
|
||||
func OaiResponsesHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
|
||||
defer service.CloseResponseBodyGracefully(resp)
|
||||
if strings.HasPrefix(strings.ToLower(resp.Header.Get("Content-Type")), "text/event-stream") {
|
||||
body, err := readBufferedResponsesSSE(resp.Body)
|
||||
if err != nil {
|
||||
return nil, types.NewOpenAIError(err, types.ErrorCodeBadResponseBody, http.StatusBadGateway)
|
||||
}
|
||||
converted := *resp
|
||||
converted.Header = resp.Header.Clone()
|
||||
converted.Header.Set("Content-Type", "application/json")
|
||||
converted.Header.Del("Content-Length")
|
||||
converted.Body = io.NopCloser(bytes.NewReader(body))
|
||||
converted.ContentLength = int64(len(body))
|
||||
return OaiResponsesHandler(c, info, &converted)
|
||||
}
|
||||
|
||||
// read response body
|
||||
var responsesResponse dto.OpenAIResponsesResponse
|
||||
@@ -35,11 +53,15 @@ func OaiResponsesHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
|
||||
return nil, types.WithOpenAIError(*oaiError, resp.StatusCode)
|
||||
}
|
||||
|
||||
info.ObserveResponseModel(responsesResponse.Model)
|
||||
responseBody = rewriteSGLangResponsesCreatedAt(info, responseBody, "created_at", responsesResponse.CreatedAt)
|
||||
|
||||
// 写入新的 response body
|
||||
service.IOCopyBytesGracefully(c, resp, responseBody)
|
||||
|
||||
// compute usage
|
||||
usage := relayconvert.NormalizeResponsesUsage(responsesResponse.Usage)
|
||||
usage := &dto.Usage{}
|
||||
service.ApplyResponsesUsage(usage, responsesResponse.Usage)
|
||||
// Count actual tool invocations from Output (not tool declarations).
|
||||
for _, output := range responsesResponse.Output {
|
||||
switch output.Type {
|
||||
@@ -64,6 +86,58 @@ func OaiResponsesHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
|
||||
return usage, nil
|
||||
}
|
||||
|
||||
// Some upstreams return SSE even for stream=false. Require a terminal response
|
||||
// and reuse JSON delivery and settlement so upstream usage is charged once.
|
||||
func readBufferedResponsesSSE(body io.Reader) ([]byte, error) {
|
||||
scanner := helper.NewStreamScanner(body)
|
||||
scanner.Split(bufio.ScanLines)
|
||||
accumulator := relayconvert.NewResponsesBufferedAccumulator()
|
||||
var data []string
|
||||
for scanner.Scan() {
|
||||
line := strings.TrimSuffix(scanner.Text(), "\r")
|
||||
if payload, ok := strings.CutPrefix(line, "data:"); ok {
|
||||
data = append(data, strings.TrimPrefix(payload, " "))
|
||||
}
|
||||
if line != "" || len(data) == 0 {
|
||||
continue
|
||||
}
|
||||
payload := strings.Join(data, "\n")
|
||||
data = nil
|
||||
if payload == "[DONE]" {
|
||||
break
|
||||
}
|
||||
var event dto.ResponsesStreamResponse
|
||||
if err := common.UnmarshalJsonStr(payload, &event); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
accumulator.ProcessEvent(&event)
|
||||
switch event.Type {
|
||||
case "error", "response.error", "response.failed", "response.cancelled", "response.canceled":
|
||||
return nil, fmt.Errorf("responses stream ended with %s", event.Type)
|
||||
case "response.completed", "response.done", "response.incomplete":
|
||||
if event.Response == nil {
|
||||
return nil, fmt.Errorf("responses terminal event has no response")
|
||||
}
|
||||
result := []byte(gjson.Get(payload, "response").Raw)
|
||||
if len(event.Response.Output) == 0 {
|
||||
accumulator.SupplementResponseOutput(event.Response)
|
||||
if len(event.Response.Output) != 0 {
|
||||
output, err := common.Marshal(event.Response.Output)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return sjson.SetRawBytes(result, "output", output)
|
||||
}
|
||||
}
|
||||
return result, nil
|
||||
}
|
||||
}
|
||||
if err := scanner.Err(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return nil, fmt.Errorf("responses stream ended without a terminal response")
|
||||
}
|
||||
|
||||
func OaiResponsesStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
|
||||
if resp == nil || resp.Body == nil {
|
||||
logger.LogError(c, "invalid response or response body")
|
||||
@@ -72,10 +146,7 @@ func OaiResponsesStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp
|
||||
|
||||
defer service.CloseResponseBodyGracefully(resp)
|
||||
|
||||
var usage = &dto.Usage{}
|
||||
var responseTextBuilder strings.Builder
|
||||
imageCounter := &relaycommon.ImageGenerationCallCounter{}
|
||||
imageCommitted := false
|
||||
accumulator := service.NewResponsesUsageAccumulator(info)
|
||||
|
||||
helper.StreamScannerHandler(c, resp, info, func(data string, sr *helper.StreamResult) {
|
||||
|
||||
@@ -86,77 +157,28 @@ func OaiResponsesStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp
|
||||
sr.Error(err)
|
||||
return
|
||||
}
|
||||
sendResponsesStreamData(c, streamResponse, data)
|
||||
switch streamResponse.Type {
|
||||
case "response.completed", "response.done":
|
||||
if streamResponse.Response != nil {
|
||||
if streamResponse.Response.Usage != nil {
|
||||
incomingUsage := relayconvert.NormalizeResponsesUsage(streamResponse.Response.Usage)
|
||||
usage = dto.MergeUsageNonZero(usage, incomingUsage)
|
||||
}
|
||||
if !imageCommitted {
|
||||
if relaycommon.IsNonBillableResponsesStatus(streamResponse.Response.Status) {
|
||||
imageCounter.Reset()
|
||||
imageCounter.Commit(info)
|
||||
imageCommitted = true
|
||||
} else {
|
||||
for i := range streamResponse.Response.Output {
|
||||
idx := i
|
||||
imageCounter.Observe(&streamResponse.Response.Output[i], &idx)
|
||||
}
|
||||
imageCounter.Commit(info)
|
||||
imageCommitted = true
|
||||
}
|
||||
}
|
||||
} else if !imageCommitted {
|
||||
imageCounter.Commit(info)
|
||||
imageCommitted = true
|
||||
}
|
||||
case "response.failed", "response.incomplete", "response.cancelled", "response.canceled":
|
||||
if !imageCommitted {
|
||||
imageCounter.Reset()
|
||||
imageCounter.Commit(info)
|
||||
imageCommitted = true
|
||||
}
|
||||
case "response.output_text.delta":
|
||||
// 处理输出文本
|
||||
responseTextBuilder.WriteString(streamResponse.Delta)
|
||||
case dto.ResponsesOutputTypeItemDone:
|
||||
if streamResponse.Item != nil {
|
||||
switch streamResponse.Item.Type {
|
||||
case dto.BuildInCallWebSearchCall:
|
||||
info.CountBillableToolCall(dto.BuildInCallWebSearchCall, "")
|
||||
case dto.BuildInCallFileSearchCall:
|
||||
info.CountBillableToolCall(dto.BuildInCallFileSearchCall, "")
|
||||
case dto.BuildInCallFunctionCall:
|
||||
info.CountBillableToolCall(dto.BuildInCallFunctionCall, streamResponse.Item.Name)
|
||||
case dto.ResponsesOutputTypeImageGenerationCall:
|
||||
if !imageCommitted {
|
||||
imageCounter.Observe(streamResponse.Item, streamResponse.OutputIndex)
|
||||
}
|
||||
}
|
||||
}
|
||||
if streamResponse.Response != nil {
|
||||
data = string(rewriteSGLangResponsesCreatedAt(info, []byte(data), "response.created_at", streamResponse.Response.CreatedAt))
|
||||
}
|
||||
sendResponsesStreamData(c, streamResponse, data)
|
||||
accumulator.Observe(&streamResponse)
|
||||
})
|
||||
|
||||
if usage.CompletionTokens == 0 {
|
||||
// 计算输出文本的 token 数量
|
||||
tempStr := responseTextBuilder.String()
|
||||
if len(tempStr) > 0 {
|
||||
// 非正常结束,使用输出文本的 token 数量
|
||||
completionTokens := service.CountTextToken(tempStr, info.UpstreamModelName)
|
||||
usage.CompletionTokens = completionTokens
|
||||
}
|
||||
common.SetContextKey(c, constant.ContextKeyResponseStreamStatus, info.StreamStatus)
|
||||
info.StreamStatus.RequireTerminal()
|
||||
return accumulator.Finish(), nil
|
||||
}
|
||||
|
||||
func rewriteSGLangResponsesCreatedAt(info *relaycommon.RelayInfo, payload []byte, path string, createdAt dto.IntValue) []byte {
|
||||
if info.GetChannelType() != constant.ChannelTypeSGLang {
|
||||
return payload
|
||||
}
|
||||
|
||||
if usage.PromptTokens == 0 && usage.CompletionTokens != 0 {
|
||||
usage.PromptTokens = info.GetEstimatePromptTokens()
|
||||
if !gjson.GetBytes(payload, path).Exists() {
|
||||
return payload
|
||||
}
|
||||
|
||||
usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
|
||||
if usage.BillingUsage != nil {
|
||||
usage.BillingUsage = dto.CloneBillingUsageWithEstimatedCompletion(usage.BillingUsage, usage.CompletionTokens)
|
||||
patched, err := sjson.SetBytes(payload, path, int(createdAt))
|
||||
if err != nil {
|
||||
return payload
|
||||
}
|
||||
|
||||
return usage, nil
|
||||
return patched
|
||||
}
|
||||
Reference in new issue
Block a user