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:
cneicy committed
1 parent
aa6c3f50f4
commit
ceddfd3d6b
828 files changed
+122253
-13702
No files matched your search
+93
-79
@@ -1,36 +1,40 @@
|
||||
package relay
|
||||
|
||||
import (
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"io"
|
||||
"net/http"
|
||||
"strings"
|
||||
|
||||
"github.com/QuantumNous/new-api/common"
|
||||
"github.com/QuantumNous/new-api/logger"
|
||||
relaycommon "github.com/QuantumNous/new-api/relay/common"
|
||||
relayconstant "github.com/QuantumNous/new-api/relay/constant"
|
||||
"github.com/QuantumNous/new-api/relay/helper"
|
||||
"github.com/QuantumNous/new-api/relaykit/dto"
|
||||
"github.com/QuantumNous/new-api/relaykit/types"
|
||||
"github.com/QuantumNous/new-api/service"
|
||||
"github.com/QuantumNous/new-api/setting/model_setting"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
)
|
||||
|
||||
func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types.NewAPIError) {
|
||||
info.InitChannelMeta(c)
|
||||
if info.RelayMode == relayconstant.RelayModeResponsesCompact &&
|
||||
!common.SupportsResponsesCompact(info.ChannelType, info.ApiType) {
|
||||
return types.NewErrorWithStatusCode(
|
||||
fmt.Errorf("unsupported endpoint %q for api type %d", "/v1/responses/compact", info.ApiType),
|
||||
types.ErrorCodeInvalidRequest,
|
||||
http.StatusBadRequest,
|
||||
types.ErrOptionWithSkipRetry(),
|
||||
)
|
||||
}
|
||||
// Independent cursors retain the payload metadata used by net/http redirects
|
||||
// and transport retries without closing the shared request storage.
|
||||
type responsesAttemptBody struct {
|
||||
common.ReplayableBody
|
||||
reader io.Reader
|
||||
}
|
||||
|
||||
func (b responsesAttemptBody) Read(p []byte) (int, error) { return b.reader.Read(p) }
|
||||
|
||||
func shouldEnableAstraResponseGuard(path string, info *relaycommon.RelayInfo) bool {
|
||||
return path == "/v1/responses" &&
|
||||
info.OriginModelName == "gpt-6-astra" &&
|
||||
info.AstraAutoSwitch && !info.AstraDetectionDisabled &&
|
||||
info.ResponseGuard == nil
|
||||
}
|
||||
|
||||
func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types.NewAPIError) {
|
||||
var responsesReq *dto.OpenAIResponsesRequest
|
||||
switch req := info.Request.(type) {
|
||||
case *dto.OpenAIResponsesRequest:
|
||||
@@ -61,77 +65,64 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
|
||||
)
|
||||
}
|
||||
|
||||
request, err := common.DeepCopy(responsesReq)
|
||||
if err != nil {
|
||||
return types.NewError(fmt.Errorf("failed to copy request to GeneralOpenAIRequest: %w", err), types.ErrorCodeInvalidRequest, types.ErrOptionWithSkipRetry())
|
||||
adaptor, requestBody, closer, apiErr := PrepareResponsesRequest(c, info, responsesReq)
|
||||
if apiErr != nil {
|
||||
return apiErr
|
||||
}
|
||||
|
||||
err = helper.ModelMappedHelper(c, info, request)
|
||||
if err != nil {
|
||||
return types.NewError(err, types.ErrorCodeChannelModelMappedError, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
if err := helper.ApplyReasoningModelSuffix(c, info, request); err != nil {
|
||||
return newConvertRequestFailedError(c, info, err)
|
||||
}
|
||||
|
||||
adaptor := GetAdaptor(info.ApiType)
|
||||
if adaptor == nil {
|
||||
return types.NewError(fmt.Errorf("invalid api type: %d", info.ApiType), types.ErrorCodeInvalidApiType, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
adaptor.Init(info)
|
||||
var requestBody io.Reader
|
||||
if model_setting.GetGlobalSettings().PassThroughRequestEnabled || info.ChannelSetting.PassThroughBodyEnabled {
|
||||
storage, err := common.GetBodyStorage(c)
|
||||
if err != nil {
|
||||
return types.NewError(err, types.ErrorCodeReadRequestBodyFailed, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
requestBody = common.NewReplayableBodyReader(storage)
|
||||
} else {
|
||||
convertedRequest, err := adaptor.ConvertOpenAIResponsesRequest(c, info, *request)
|
||||
if err != nil {
|
||||
return newConvertRequestFailedError(c, info, err)
|
||||
}
|
||||
relaycommon.AppendRequestConversionFromRequest(info, convertedRequest)
|
||||
jsonData, err := common.Marshal(convertedRequest)
|
||||
if err != nil {
|
||||
return types.NewError(err, types.ErrorCodeConvertRequestFailed, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
|
||||
// remove disabled fields for OpenAI Responses API
|
||||
jsonData, err = relaycommon.RemoveDisabledFields(jsonData, info.ChannelOtherSettings, info.ChannelSetting.PassThroughBodyEnabled)
|
||||
if err != nil {
|
||||
return types.NewError(err, types.ErrorCodeConvertRequestFailed, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
|
||||
// apply param override
|
||||
if len(info.ParamOverride) > 0 {
|
||||
jsonData, err = relaycommon.ApplyParamOverrideWithRelayInfo(jsonData, info)
|
||||
if err != nil {
|
||||
return newAPIErrorFromParamOverride(err)
|
||||
defer closer.Close()
|
||||
// Missing-usage estimation follows the actual prepared payload, including
|
||||
// parameter overrides, rather than the client's pre-transformation request.
|
||||
if info.DurableAttemptBilling {
|
||||
if cursor, err := requestBody.NewReader(); err == nil {
|
||||
var sent dto.OpenAIResponsesRequest
|
||||
decodeErr := json.NewDecoder(cursor).Decode(&sent)
|
||||
_ = cursor.Close()
|
||||
if decodeErr == nil {
|
||||
if tokens, err := service.CountRequestToken(c, sent.GetTokenCountMeta(), info); err == nil {
|
||||
info.SetEstimatePromptTokens(tokens)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
logger.LogDebug(c, "requestBody: %s", jsonData)
|
||||
body, closer, err := relaycommon.NewOutboundJSONBody(jsonData)
|
||||
if err != nil {
|
||||
return types.NewError(err, types.ErrorCodeConvertRequestFailed, types.ErrOptionWithSkipRetry())
|
||||
}
|
||||
defer closer.Close()
|
||||
jsonData = nil
|
||||
requestBody = body
|
||||
}
|
||||
if shouldEnableAstraResponseGuard(c.Request.URL.Path, info) {
|
||||
info.ResponseGuard = &relaycommon.ResponseModelGuard{Deadline: info.GuardDeadline}
|
||||
}
|
||||
|
||||
var httpResp *http.Response
|
||||
resp, err := adaptor.DoRequest(c, info, requestBody)
|
||||
guardEnabled := info.ResponseGuard != nil && info.AstraAutoSwitch
|
||||
|
||||
body, err := requestBody.NewReader()
|
||||
if err != nil {
|
||||
return types.NewOpenAIError(err, types.ErrorCodeDoRequestFailed, http.StatusInternalServerError)
|
||||
return types.NewOpenAIError(err, types.ErrorCodeReadRequestBodyFailed, http.StatusInternalServerError)
|
||||
}
|
||||
resp, requestErr := adaptor.DoRequest(c, info, responsesAttemptBody{ReplayableBody: requestBody, reader: body})
|
||||
_ = body.Close()
|
||||
if requestErr != nil {
|
||||
apiErr := types.NewError(requestErr, types.ErrorCodeDoRequestFailed)
|
||||
guard := info.ResponseGuard
|
||||
if guard == nil || apiErr.GetErrorCode() != relaycommon.ResponseGuardErrorCode {
|
||||
return apiErr
|
||||
}
|
||||
attempt := guard.Last()
|
||||
if attempt != nil {
|
||||
service.RequestPolicy(c).AddEvent(service.PolicyEvent{ChannelID: info.ChannelId, ErrorCode: string(relaycommon.ResponseGuardErrorCode), ErrorSource: "model_guard", Health: "unchanged", Decision: service.PolicyDecision{Action: "reject", Reason: attempt.Reason, Source: "model_guard"}})
|
||||
}
|
||||
return guard.APIError()
|
||||
}
|
||||
var httpResp *http.Response
|
||||
if resp != nil {
|
||||
httpResp = resp.(*http.Response)
|
||||
}
|
||||
if guardEnabled {
|
||||
defer func() {
|
||||
if attempt := info.ResponseGuard.Last(); attempt != nil {
|
||||
attempt.Finish(c.Writer.Written())
|
||||
}
|
||||
}()
|
||||
}
|
||||
|
||||
statusCodeMappingStr := c.GetString("status_code_mapping")
|
||||
|
||||
if resp != nil {
|
||||
httpResp = resp.(*http.Response)
|
||||
|
||||
if httpResp != nil {
|
||||
if httpResp.StatusCode != http.StatusOK {
|
||||
newAPIError = service.RelayErrorHandler(c.Request.Context(), httpResp, false)
|
||||
// reset status code 重置状态码
|
||||
@@ -141,6 +132,23 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
|
||||
}
|
||||
|
||||
usage, newAPIError := adaptor.DoResponse(c, httpResp, info)
|
||||
if guardEnabled {
|
||||
if attempt := info.ResponseGuard.Last(); attempt != nil {
|
||||
attempt.Finish(c.Writer.Written())
|
||||
if guardErr := attempt.APIError(); guardErr != nil {
|
||||
if info.StreamStatus != nil {
|
||||
info.StreamStatus.MarkFailed(string(relaycommon.ResponseGuardErrorCode), "api_error", http.StatusConflict)
|
||||
}
|
||||
if c.Writer.Written() && c.Request.Context().Err() == nil {
|
||||
message := common.MessageWithRequestId(attempt.PublicErrorMessage(), c.GetString(common.RequestIdKey))
|
||||
c.SSEvent("response.failed", gin.H{"type": "response.failed", "sequence_number": attempt.Sequence + 1, "response": gin.H{"id": attempt.ResponseID, "object": "response", "status": "failed", "error": gin.H{"code": relaycommon.ResponseGuardErrorCode, "message": message}}})
|
||||
c.Writer.Flush()
|
||||
c.Set("response_guard_error_sent", true)
|
||||
}
|
||||
return guardErr
|
||||
}
|
||||
}
|
||||
}
|
||||
if newAPIError != nil {
|
||||
// reset status code 重置状态码
|
||||
service.ResetStatusCode(newAPIError, statusCodeMappingStr)
|
||||
@@ -165,10 +173,16 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
|
||||
return nil
|
||||
}
|
||||
|
||||
if strings.HasPrefix(info.OriginModelName, "gpt-4o-audio") {
|
||||
service.PostAudioConsumeQuota(c, info, usageDto, "")
|
||||
} else {
|
||||
service.PostTextConsumeQuota(c, info, usageDto, nil)
|
||||
}
|
||||
ConsumeResponsesQuota(c, info, usageDto)
|
||||
return nil
|
||||
}
|
||||
|
||||
// ConsumeResponsesQuota applies the same settlement dispatch to HTTP and
|
||||
// WebSocket Responses usage. Compact requests keep their separate repricing.
|
||||
func ConsumeResponsesQuota(c *gin.Context, info *relaycommon.RelayInfo, usage *dto.Usage) {
|
||||
if strings.HasPrefix(info.OriginModelName, "gpt-4o-audio") {
|
||||
service.PostAudioConsumeQuota(c, info, usage, "")
|
||||
return
|
||||
}
|
||||
service.PostTextConsumeQuota(c, info, usage, nil)
|
||||
}
|
||||
Reference in new issue
Block a user