mirror of
https://github.com/songquanpeng/one-api.git
synced 2025-09-17 09:16:36 +08:00
Merge 790e1e9f5b into 6ded638f70
This commit is contained in:
Laisky.Cai
GitHub
commit
022988c007
@ -8,6 +8,7 @@ import (
|
||||
"strings"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/songquanpeng/one-api/common/config"
|
||||
"github.com/songquanpeng/one-api/relay/adaptor"
|
||||
"github.com/songquanpeng/one-api/relay/adaptor/doubao"
|
||||
"github.com/songquanpeng/one-api/relay/adaptor/minimax"
|
||||
@ -82,6 +83,33 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
|
||||
}
|
||||
request.StreamOptions.IncludeUsage = true
|
||||
}
|
||||
|
||||
// o1/o1-mini/o1-preview do not support system prompt/max_tokens/temperature
|
||||
if strings.HasPrefix(request.Model, "o1") ||
|
||||
strings.HasPrefix(request.Model, "o3") {
|
||||
temperature := float64(1)
|
||||
request.Temperature = &temperature // Only the default (1) value is supported
|
||||
|
||||
request.MaxTokens = 0
|
||||
request.Messages = func(raw []model.Message) (filtered []model.Message) {
|
||||
for i := range raw {
|
||||
if raw[i].Role != "system" {
|
||||
filtered = append(filtered, raw[i])
|
||||
}
|
||||
}
|
||||
|
||||
return
|
||||
}(request.Messages)
|
||||
}
|
||||
|
||||
if request.Stream && !config.EnforceIncludeUsage &&
|
||||
(strings.HasPrefix(request.Model, "gpt-4o-audio") ||
|
||||
strings.HasPrefix(request.Model, "gpt-4o-mini-audio")) {
|
||||
// TODO: Since it is not clear how to implement billing in stream mode,
|
||||
// it is temporarily not supported
|
||||
return nil, errors.New("set ENFORCE_INCLUDE_USAGE=true to enable stream mode for gpt-4o-audio")
|
||||
}
|
||||
|
||||
return request, nil
|
||||
}
|
||||
|
||||
|
||||
@ -12,6 +12,8 @@ var ModelList = []string{
|
||||
"gpt-4o-2024-11-20",
|
||||
"chatgpt-4o-latest",
|
||||
"gpt-4o-mini", "gpt-4o-mini-2024-07-18",
|
||||
"gpt-4o-mini-audio-preview", "gpt-4o-mini-audio-preview-2024-12-17",
|
||||
"gpt-4o-audio-preview", "gpt-4o-audio-preview-2024-12-17", "gpt-4o-audio-preview-2024-10-01",
|
||||
"gpt-4-vision-preview",
|
||||
"text-embedding-ada-002", "text-embedding-3-small", "text-embedding-3-large",
|
||||
"text-curie-001", "text-babbage-001", "text-ada-001", "text-davinci-002", "text-davinci-003",
|
||||
@ -24,4 +26,5 @@ var ModelList = []string{
|
||||
"o1", "o1-2024-12-17",
|
||||
"o1-preview", "o1-preview-2024-09-12",
|
||||
"o1-mini", "o1-mini-2024-09-12",
|
||||
"o3-mini", "o3-mini-2025-01-31",
|
||||
}
|
||||
|
||||
@ -42,6 +42,14 @@ var ModelRatio = map[string]float64{
|
||||
"gpt-4o-mini": 0.075, // $0.00015 / 1K tokens
|
||||
"gpt-4o-mini-2024-07-18": 0.075, // $0.00015 / 1K tokens
|
||||
"gpt-4-vision-preview": 5, // $0.01 / 1K tokens
|
||||
// Audio billing will mix text and audio tokens, the unit price is different.
|
||||
// Here records the cost of text, the cost multiplier of audio
|
||||
// relative to text is in AudioRatio
|
||||
"gpt-4o-audio-preview": 1.25, // $0.0025 / 1K tokens
|
||||
"gpt-4o-audio-preview-2024-12-17": 1.25, // $0.0025 / 1K tokens
|
||||
"gpt-4o-audio-preview-2024-10-01": 1.25, // $0.0025 / 1K tokens
|
||||
"gpt-4o-mini-audio-preview": 0.15 * MILLI_USD, // $0.15/1M tokens
|
||||
"gpt-4o-mini-audio-preview-2024-12-17": 0.15 * MILLI_USD, // $0.15/1M tokens
|
||||
"gpt-3.5-turbo": 0.25, // $0.0005 / 1K tokens
|
||||
"gpt-3.5-turbo-0301": 0.75,
|
||||
"gpt-3.5-turbo-0613": 0.75,
|
||||
@ -56,6 +64,8 @@ var ModelRatio = map[string]float64{
|
||||
"o1-preview-2024-09-12": 7.5,
|
||||
"o1-mini": 1.5, // $3.00 / 1M input tokens
|
||||
"o1-mini-2024-09-12": 1.5,
|
||||
"o3-mini": 1.1 * MILLI_USD,
|
||||
"o3-mini-2025-01-31": 1.1 * MILLI_USD,
|
||||
"davinci-002": 1, // $0.002 / 1K tokens
|
||||
"babbage-002": 0.2, // $0.0004 / 1K tokens
|
||||
"text-ada-001": 0.2,
|
||||
@ -65,7 +75,7 @@ var ModelRatio = map[string]float64{
|
||||
"text-davinci-003": 10,
|
||||
"text-davinci-edit-001": 10,
|
||||
"code-davinci-edit-001": 10,
|
||||
"whisper-1": 15, // $0.006 / minute -> $0.006 / 150 words -> $0.006 / 200 tokens -> $0.03 / 1k tokens
|
||||
"whisper-1": 15,
|
||||
"tts-1": 7.5, // $0.015 / 1K characters
|
||||
"tts-1-1106": 7.5,
|
||||
"tts-1-hd": 15, // $0.030 / 1K characters
|
||||
@ -359,6 +369,76 @@ var ModelRatio = map[string]float64{
|
||||
"mistralai/mixtral-8x7b-instruct-v0.1": 0.300 * USD,
|
||||
}
|
||||
|
||||
// AudioRatio represents the price ratio between audio tokens and text tokens
|
||||
var AudioRatio = map[string]float64{
|
||||
"gpt-4o-audio-preview": 16,
|
||||
"gpt-4o-audio-preview-2024-12-17": 16,
|
||||
"gpt-4o-audio-preview-2024-10-01": 40,
|
||||
"gpt-4o-mini-audio-preview": 10 / 0.15,
|
||||
"gpt-4o-mini-audio-preview-2024-12-17": 10 / 0.15,
|
||||
}
|
||||
|
||||
// GetAudioPromptRatio returns the audio prompt ratio for the given model.
|
||||
func GetAudioPromptRatio(actualModelName string) float64 {
|
||||
var v float64
|
||||
if ratio, ok := AudioRatio[actualModelName]; ok {
|
||||
v = ratio
|
||||
} else {
|
||||
v = 16
|
||||
}
|
||||
|
||||
return v
|
||||
}
|
||||
|
||||
// AudioCompletionRatio is the completion ratio for audio models.
|
||||
var AudioCompletionRatio = map[string]float64{
|
||||
"whisper-1": 0,
|
||||
"gpt-4o-audio-preview": 2,
|
||||
"gpt-4o-audio-preview-2024-12-17": 2,
|
||||
"gpt-4o-audio-preview-2024-10-01": 2,
|
||||
"gpt-4o-mini-audio-preview": 2,
|
||||
"gpt-4o-mini-audio-preview-2024-12-17": 2,
|
||||
}
|
||||
|
||||
// GetAudioCompletionRatio returns the completion ratio for audio models.
|
||||
func GetAudioCompletionRatio(actualModelName string) float64 {
|
||||
var v float64
|
||||
if ratio, ok := AudioCompletionRatio[actualModelName]; ok {
|
||||
v = ratio
|
||||
} else {
|
||||
v = 2
|
||||
}
|
||||
|
||||
return v
|
||||
}
|
||||
|
||||
// AudioTokensPerSecond is the number of audio tokens per second for each model.
|
||||
var AudioPromptTokensPerSecond = map[string]float64{
|
||||
// Whisper API price is $0.0001/sec. One-api's historical ratio is 15,
|
||||
// corresponding to $0.03/kilo_tokens.
|
||||
// After conversion, tokens per second should be 0.0001/0.03*1000 = 3.3333.
|
||||
"whisper-1": 0.0001 / 0.03 * 1000,
|
||||
// gpt-4o-audio series processes 10 tokens per second
|
||||
"gpt-4o-audio-preview": 10,
|
||||
"gpt-4o-audio-preview-2024-12-17": 10,
|
||||
"gpt-4o-audio-preview-2024-10-01": 10,
|
||||
"gpt-4o-mini-audio-preview": 10,
|
||||
"gpt-4o-mini-audio-preview-2024-12-17": 10,
|
||||
}
|
||||
|
||||
// GetAudioPromptTokensPerSecond returns the number of audio tokens per second
|
||||
// for the given model.
|
||||
func GetAudioPromptTokensPerSecond(actualModelName string) float64 {
|
||||
var v float64
|
||||
if tokensPerSecond, ok := AudioPromptTokensPerSecond[actualModelName]; ok {
|
||||
v = tokensPerSecond
|
||||
} else {
|
||||
v = 10
|
||||
}
|
||||
|
||||
return v
|
||||
}
|
||||
|
||||
var CompletionRatio = map[string]float64{
|
||||
// aws llama3
|
||||
"llama3-8b-8192(33)": 0.0006 / 0.0003,
|
||||
@ -497,8 +577,9 @@ func GetCompletionRatio(name string, channelType int) float64 {
|
||||
}
|
||||
return 2
|
||||
}
|
||||
// including o1, o1-preview, o1-mini
|
||||
if strings.HasPrefix(name, "o1") {
|
||||
// including o1/o1-preview/o1-mini
|
||||
if strings.HasPrefix(name, "o1") ||
|
||||
strings.HasPrefix(name, "o3") {
|
||||
return 4
|
||||
}
|
||||
if name == "chatgpt-4o-latest" {
|
||||
|
||||
@ -34,6 +34,9 @@ type GeneralOpenAIRequest struct {
|
||||
MaxTokens int `json:"max_tokens,omitempty"`
|
||||
MaxCompletionTokens *int `json:"max_completion_tokens,omitempty"`
|
||||
N int `json:"n,omitempty"`
|
||||
// ReasoningEffort constrains effort on reasoning for reasoning models, reasoning models only.
|
||||
ReasoningEffort string `json:"reasoning_effort,omitempty" binding:"omitempty,oneof=low medium high"`
|
||||
// Modalities currently the model only programmatically allows modalities = [“text”, “audio”]
|
||||
Modalities []string `json:"modalities,omitempty"`
|
||||
Prediction any `json:"prediction,omitempty"`
|
||||
Audio *Audio `json:"audio,omitempty"`
|
||||
|
||||
Loading…
Reference in New Issue
Block a user