update gpt-4-32k adapter

This commit is contained in:
Zhang Minghan 2023-09-23 22:38:26 +08:00
parent 5130d806a9
commit 653a90bd08
15 changed files with 273 additions and 82 deletions

View File

@ -3,23 +3,24 @@ package api
import (
"chat/auth"
"chat/types"
"chat/utils"
)
type Buffer struct {
Enable bool `json:"enable"`
Model string `json:"model"`
Quota float32 `json:"quota"`
Data string `json:"data"`
Cursor int `json:"cursor"`
Times int `json:"times"`
}
func NewBuffer(enable bool, history []types.ChatGPTMessage) *Buffer {
buffer := &Buffer{Data: "", Cursor: 0, Times: 0, Enable: enable}
if enable {
buffer.Quota = auth.CountInputToken(utils.CountTokenPrice(history))
func NewBuffer(model string, history []types.ChatGPTMessage) *Buffer {
return &Buffer{
Data: "",
Cursor: 0,
Times: 0,
Model: model,
Quota: auth.CountInputToken(model, history),
}
return buffer
}
func (b *Buffer) GetCursor() int {
@ -27,10 +28,7 @@ func (b *Buffer) GetCursor() int {
}
func (b *Buffer) GetQuota() float32 {
if !b.Enable {
return 0.
}
return b.Quota + auth.CountOutputToken(b.ReadTimes())
return b.Quota + auth.CountOutputToken(b.Model, b.ReadTimes())
}
func (b *Buffer) Write(data string) string {

View File

@ -29,15 +29,15 @@ func SendSegmentMessage(conn *websocket.Conn, message interface{}) {
_ = conn.WriteMessage(websocket.TextMessage, []byte(utils.ToJson(message)))
}
func GetErrorQuota(isGPT4 bool) float32 {
if isGPT4 {
func GetErrorQuota(model string) float32 {
if types.IsGPT4Model(model) {
return -0xe // special value for error
} else {
return 0
}
}
func TextChat(db *sql.DB, cache *redis.Client, user *auth.User, conn *websocket.Conn, instance *conversation.Conversation) string {
func GetTextSegment(instance *conversation.Conversation) (string, []types.ChatGPTMessage) {
var keyword string
var segment []types.ChatGPTMessage
@ -46,11 +46,16 @@ func TextChat(db *sql.DB, cache *redis.Client, user *auth.User, conn *websocket.
} else {
segment = conversation.CopyMessage(instance.GetMessageSegment(12))
}
return keyword, segment
}
func TextChat(db *sql.DB, cache *redis.Client, user *auth.User, conn *websocket.Conn, instance *conversation.Conversation) string {
keyword, segment := GetTextSegment(instance)
SendSegmentMessage(conn, types.ChatSegmentResponse{Keyword: keyword, End: false})
isProPlan := auth.CanEnableSubscription(db, cache, user)
if instance.IsEnableGPT4() && (!isProPlan) && (!auth.CanEnableGPT4(db, user)) {
model := instance.GetModel()
useReverse := auth.CanEnableSubscription(db, cache, user)
if !auth.CanEnableModelWithSubscription(db, user, model, useReverse) {
SendSegmentMessage(conn, types.ChatSegmentResponse{
Message: defaultQuotaMessage,
Quota: 0,
@ -59,9 +64,9 @@ func TextChat(db *sql.DB, cache *redis.Client, user *auth.User, conn *websocket.
return defaultQuotaMessage
}
buffer := NewBuffer(instance.IsEnableGPT4(), segment)
StreamRequest(instance.IsEnableGPT4(), isProPlan, segment,
utils.Multi(instance.IsEnableGPT4() || isProPlan, -1, 2000),
buffer := NewBuffer(model, segment)
StreamRequest(model, useReverse, segment,
utils.Multi(types.IsGPT4Model(model) || useReverse, -1, 2000),
func(resp string) {
SendSegmentMessage(conn, types.ChatSegmentResponse{
Message: buffer.Write(resp),
@ -70,19 +75,19 @@ func TextChat(db *sql.DB, cache *redis.Client, user *auth.User, conn *websocket.
})
})
if buffer.IsEmpty() {
if isProPlan {
if useReverse {
auth.DecreaseSubscriptionUsage(cache, user)
}
SendSegmentMessage(conn, types.ChatSegmentResponse{
Message: defaultErrorMessage,
Quota: GetErrorQuota(instance.IsEnableGPT4()),
Quota: GetErrorQuota(model),
End: true,
})
return defaultErrorMessage
}
// collect quota
if !isProPlan {
if !useReverse {
user.UseQuota(db, buffer.GetQuota())
}
SendSegmentMessage(conn, types.ChatSegmentResponse{End: true, Quota: buffer.GetQuota()})

View File

@ -105,14 +105,25 @@ func NativeStreamRequest(model string, endpoint string, apikeys string, messages
}
}
func StreamRequest(enableGPT4 bool, isProPlan bool, messages []types.ChatGPTMessage, token int, callback func(string)) {
if enableGPT4 {
if isProPlan {
func StreamRequest(model string, enableReverse bool, messages []types.ChatGPTMessage, token int, callback func(string)) {
switch model {
case types.GPT4,
types.GPT40314,
types.GPT40613:
if enableReverse {
NativeStreamRequest(viper.GetString("openai.reverse"), viper.GetString("openai.pro_endpoint"), viper.GetString("openai.pro"), messages, token, callback)
} else {
NativeStreamRequest("gpt-4", viper.GetString("openai.gpt4_endpoint"), viper.GetString("openai.gpt4"), messages, token, callback)
NativeStreamRequest(model, viper.GetString("openai.gpt4_endpoint"), viper.GetString("openai.gpt4"), messages, token, callback)
}
} else {
NativeStreamRequest("gpt-3.5-turbo-16k-0613", viper.GetString("openai.user_endpoint"), viper.GetString("openai.user"), messages, token, callback)
case types.GPT432k,
types.GPT432k0613,
types.GPT432k0314:
NativeStreamRequest(model, viper.GetString("openai.gpt4_endpoint"), viper.GetString("openai.gpt4"), messages, token, callback)
case types.GPT3Turbo16k,
types.GPT3Turbo16k0301,
types.GPT3Turbo16k0613:
NativeStreamRequest(types.GPT3Turbo16k, viper.GetString("openai.user_endpoint"), viper.GetString("openai.user"), messages, token, callback)
default:
NativeStreamRequest(types.GPT3Turbo, viper.GetString("openai.anonymous_endpoint"), viper.GetString("openai.anonymous"), messages, token, callback)
}
}

View File

@ -62,7 +62,6 @@ const resources = {
buy: "Buy {{amount}} points",
dalle: "DALL·E Image Generator",
"dalle-free": "5 free quotas per day",
gpt4: "GPT-4",
flex: "Flexible Billing",
input: "Input",
output: "Output",
@ -106,7 +105,7 @@ const resources = {
"free-web": "web searching feature",
"free-conversation": "conversation storage",
"free-api": "API calls",
"pro-gpt4": "GPT-4 10 requests per day",
"pro-gpt4": "GPT-4 50 requests per day",
"pro-dalle": "50 quotas per day",
"pro-service": "Priority Service Support",
"pro-thread": "Concurrency Increase",
@ -207,7 +206,6 @@ const resources = {
buy: "购买 {{amount}} 点数",
dalle: "DALL·E AI 绘图",
"dalle-free": "每天 5 次免费绘图配额",
gpt4: "GPT-4",
flex: "灵活计费",
input: "输入",
output: "输出",
@ -250,7 +248,7 @@ const resources = {
"free-web": "联网搜索功能",
"free-conversation": "对话存储记录",
"free-api": "API 调用",
"pro-gpt4": "GPT-4 每日请求 10 次",
"pro-gpt4": "GPT-4 每日请求 50 次",
"pro-dalle": "每日 50 次绘图",
"pro-service": "优先服务支持",
"pro-thread": "并发数提升",
@ -354,7 +352,6 @@ const resources = {
buy: "Купить {{amount}} очков",
dalle: "Генератор изображений DALL·E",
"dalle-free": "5 бесплатных квот в день",
gpt4: "GPT-4",
flex: "Гибкая тарификация",
input: "Вход",
output: "Выход",
@ -399,7 +396,7 @@ const resources = {
"free-web": "веб-поиск",
"free-conversation": "хранение разговоров",
"free-api": "API вызовы",
"pro-gpt4": "GPT-4 10 запросов в день",
"pro-gpt4": "GPT-4 50 запросов в день",
"pro-dalle": "50 квот в день",
"pro-service": "Приоритетная служба поддержки",
"pro-thread": "Увеличение параллелизма",

View File

@ -248,7 +248,7 @@ function Quota() {
<Separator orientation={`horizontal`} className={`my-2`} />
<div className={`product-item`}>
<div className={`row title`}>
<div>{t("buy.gpt4")}</div>
<div>GPT-4</div>
<div className={`grow`} />
<div className={`column`}>
<Cloud className={`h-4 w-4`} /> {t("buy.flex")}
@ -276,6 +276,37 @@ function Quota() {
4.3 / 1k token
</div>
</div>
</div>
<div className={`product-item`}>
<div className={`row title`}>
<div>GPT-4-32K</div>
<div className={`grow`} />
<div className={`column`}>
<Cloud className={`h-4 w-4`} /> {t("buy.flex")}
</div>
</div>
<div className={`row desc`}>
<div className={`column`}>
<HardDriveUpload className={`h-4 w-4`} />
{t("buy.input")}
</div>
<div className={`grow`} />
<div className={`column`}>
<Cloud className={`h-4 w-4`} />
4.2 / 1k token
</div>
</div>
<div className={`row desc`}>
<div className={`column`}>
<HardDriveDownload className={`h-4 w-4`} />
{t("buy.output")}
</div>
<div className={`grow`} />
<div className={`column`}>
<Cloud className={`h-4 w-4`} />
8.6 / 1k token
</div>
</div>
<div className={`row desc`}>
<div className={`column info`}>
<Info className={`h-4 w-4`} />

View File

@ -46,6 +46,8 @@ import { buySubscription } from "../conversation/addition.ts";
function calc_prize(month: number): number {
if (month >= 12) {
return 8 * month * 0.8;
} else if (month >= 6) {
return 8 * month * 0.9;
}
return 8 * month;
@ -101,17 +103,22 @@ function Upgrade({ children }: UpgradeProps) {
<SelectContent>
<SelectItem value={"1"}>{t(`sub.time.1`)}</SelectItem>
<SelectItem value={"3"}>{t(`sub.time.3`)}</SelectItem>
<SelectItem value={"6"}>{t(`sub.time.6`)}</SelectItem>
<SelectItem value={"6"}>
{t(`sub.time.6`)}
<Badge className={`ml-2 cent`}>
{t(`percent`, { cent: 9 })}
</Badge>
</SelectItem>
<SelectItem value={"12"}>
{t(`sub.time.12`)}
<Badge className={`ml-2 cent`}>
{t(`percent`, { cent: 9 })}
{t(`percent`, { cent: 8 })}
</Badge>
</SelectItem>
</SelectContent>
</Select>
<p className={`price`}>
{t("sub.price", { price: calc_prize(month) })}
{t("sub.price", { price: calc_prize(month).toFixed(2) })}
</p>
</div>
<DialogFooter>

View File

@ -29,7 +29,7 @@ func BuySubscription(db *sql.DB, user *User, month int) bool {
func IncreaseSubscriptionUsage(cache *redis.Client, user *User) bool {
today := time.Now().Format("2006-01-02")
return utils.IncrWithLimit(cache, fmt.Sprintf(":subscription-usage:%s:%d", today, user.ID), 1, 999, 60*60*24) // 1 day
return utils.IncrWithLimit(cache, fmt.Sprintf(":subscription-usage:%s:%d", today, user.ID), 1, 50, 60*60*24) // 1 day
}
func DecreaseSubscriptionUsage(cache *redis.Client, user *User) bool {

View File

@ -1,6 +1,8 @@
package auth
import (
"chat/types"
"chat/utils"
"database/sql"
)
@ -13,18 +15,46 @@ import (
// $0.03 / 1K tokens $0.06 / 1K tokens
// ¥0.21 / 1K tokens ¥0.43 / 1K tokens
// 2.1 nio / 1K tokens 4.3 nio / 1K tokens
//
// GPT-4 price (32k-context)
// Input Output
// $0.06 / 1K tokens $0.12 / 1K tokens
// ¥0.43 / 1K tokens ¥0.86 / 1K tokens
// 4.3 nio / 1K tokens 8.6 nio / 1K tokens
// Dalle price (512x512)
// $0.018 / per image
// ¥0.13 / per image
// 1 nio / per image
func CountInputToken(n int) float32 {
return float32(n) / 1000 * 2.1
func CountInputToken(model string, v []types.ChatGPTMessage) float32 {
switch model {
case types.GPT3Turbo:
return 0
case types.GPT3Turbo16k:
return 0
case types.GPT4:
return float32(utils.CountTokenPrice(v, model)) / 1000 * 2.1
case types.GPT432k:
return float32(utils.CountTokenPrice(v, model)) / 1000 * 2.1 * 2
default:
return 0
}
}
func CountOutputToken(n int) float32 {
return float32(n) / 1000 * 4.3
func CountOutputToken(model string, t int) float32 {
switch model {
case types.GPT3Turbo:
return 0
case types.GPT3Turbo16k:
return 0
case types.GPT4:
return float32(t*utils.GetWeightByModel(model)) / 1000 * 4.3
case types.GPT432k:
return float32(t*utils.GetWeightByModel(model)) / 1000 * 8.6
default:
return 0
}
}
func ReduceDalle(db *sql.DB, user *User) bool {
@ -34,8 +64,24 @@ func ReduceDalle(db *sql.DB, user *User) bool {
return user.UseQuota(db, 1)
}
func CanEnableGPT4(db *sql.DB, user *User) bool {
func CanEnableModel(db *sql.DB, user *User, model string) bool {
switch model {
case types.GPT4, types.GPT40613, types.GPT40314:
return user.GetQuota(db) >= 5
case types.GPT432k, types.GPT432k0613, types.GPT432k0314:
return user.GetQuota(db) >= 50
default:
return true
}
}
func CanEnableModelWithSubscription(db *sql.DB, user *User, model string, useReverse bool) bool {
if utils.Contains(model, types.GPT4Array) {
if useReverse {
return true
}
}
return CanEnableModel(db, user, model)
}
func BuyQuota(db *sql.DB, user *User, quota int) bool {

View File

@ -13,7 +13,7 @@ type Conversation struct {
Id int64 `json:"id"`
Name string `json:"name"`
Message []types.ChatGPTMessage `json:"message"`
EnableGPT4 bool `json:"enable_gpt4"`
Model string `json:"model"`
EnableWeb bool `json:"enable_web"`
}
@ -21,7 +21,7 @@ type FormMessage struct {
Type string `json:"type"` // ping
Message string `json:"message"`
Web bool `json:"web"`
GPT4 bool `json:"gpt4"`
Model string `json:"model"`
}
func NewConversation(db *sql.DB, id int64) *Conversation {
@ -30,21 +30,21 @@ func NewConversation(db *sql.DB, id int64) *Conversation {
Id: GetConversationLengthByUserID(db, id) + 1,
Name: "new chat",
Message: []types.ChatGPTMessage{},
EnableGPT4: false,
Model: types.GPT3Turbo,
EnableWeb: false,
}
}
func (c *Conversation) IsEnableGPT4() bool {
return c.EnableGPT4
func (c *Conversation) GetModel() string {
return c.Model
}
func (c *Conversation) IsEnableWeb() bool {
return c.EnableWeb
}
func (c *Conversation) SetEnableGPT4(enable bool) {
c.EnableGPT4 = enable
func (c *Conversation) SetModel(model string) {
c.Model = model
}
func (c *Conversation) SetEnableWeb(enable bool) {
@ -141,7 +141,7 @@ func (c *Conversation) AddMessageFromUserForm(data []byte) (string, error) {
}
c.AddMessageFromUser(form.Message)
c.SetEnableGPT4(form.GPT4)
c.SetModel(form.Model)
c.SetEnableWeb(form.Web)
return form.Message, nil
}

View File

@ -84,7 +84,16 @@ func GenerateAPI(c *gin.Context) {
})
}
hash, err := CreateGenerationWithCache(form.Model, form.Prompt, func(data string) {
useReverse := auth.CanEnableSubscription(db, cache, user)
if !auth.CanEnableModelWithSubscription(db, user, form.Model, useReverse) {
api.SendSegmentMessage(conn, types.ChatSegmentResponse{
Message: "You don't have enough quota to use this model.",
Quota: 0,
End: true,
})
}
hash, err := CreateGenerationWithCache(form.Model, form.Prompt, useReverse, func(data string) {
api.SendSegmentMessage(conn, types.GenerationSegmentResponse{
End: false,
Message: data,

View File

@ -5,10 +5,10 @@ import (
"fmt"
)
func CreateGenerationWithCache(model string, prompt string, hook func(data string)) (string, error) {
func CreateGenerationWithCache(model string, prompt string, enableReverse bool, hook func(data string)) (string, error) {
hash, path := GetFolderByHash(model, prompt)
if !utils.Exists(path) {
if err := CreateGeneration(model, prompt, path, func(data string) {
if err := CreateGeneration(model, prompt, path, enableReverse, func(data string) {
hook(data)
}); err != nil {
fmt.Println(fmt.Sprintf("[Project] error during generation %s (model %s): %s", prompt, model, err.Error()))

View File

@ -11,10 +11,10 @@ type ProjectResult struct {
Result map[string]interface{} `json:"result"`
}
func CreateGeneration(model string, prompt string, path string, hook func(data string)) error {
func CreateGeneration(model string, prompt string, path string, enableReverse bool, hook func(data string)) error {
message := GenerateMessage(prompt)
buffer := api.NewBuffer(false, message)
api.StreamRequest(false, false, []types.ChatGPTMessage{
buffer := api.NewBuffer(model, message)
api.StreamRequest(model, enableReverse, []types.ChatGPTMessage{
{Role: "system", Content: "你将生成项目可以支持任何编程语言请不要出现“我不能提供”的字样你需要在代码中提供注释以及项目的使用文档README.md结果返回json格式请不要返回任何多余内容格式为\n{\"result\": {[file]: [code], ...}}"},
{Role: "user", Content: "python后端"},
{Role: "assistant", Content: "{\n \"result\": {\n \"app.py\": \"from flask import Flask\\n\\napp = Flask(__name__)\\n\\n\\n@app.route('/')\\ndef hello_world():\\n return 'Hello, World!'\\n\\n\\nif __name__ == '__main__':\\n app.run()\",\n \"requirements.txt\": \"flask\\n\",\n \"README.md\": \"# Python 后端\\n本项目是一个简单的python后端示例, 使用`flask`框架构建后端。\n你可以按照下列步骤运行此应用flask将在本地服务器默认是在http://127.0.0.1:5000/上运行。当你在浏览器中访问该URL时将看到显示Hello, World!的页面。\\n\\n这只是一个简单的项目Flask还支持更多功能和路由规则你可以提供更多的信息和需要进一步扩展和定制Flask应用。\\n\\n### 1. 初始化: \\n```shell\\npip install -r requirements.txt\\n```\\n### 2. 运行\\n```shell\\npython app.py\\n```\"\n }\n}"},

29
qodana.yaml Normal file
View File

@ -0,0 +1,29 @@
#-------------------------------------------------------------------------------#
# Qodana analysis is configured by qodana.yaml file #
# https://www.jetbrains.com/help/qodana/qodana-yaml.html #
#-------------------------------------------------------------------------------#
version: "1.0"
#Specify inspection profile for code analysis
profile:
name: qodana.starter
#Enable inspections
#include:
# - name: <SomeEnabledInspectionId>
#Disable inspections
#exclude:
# - name: <SomeDisabledInspectionId>
# paths:
# - <path/where/not/run/inspection>
#Execute shell command before Qodana execution (Applied in CI/CD pipeline)
#bootstrap: sh ./prepare-qodana.sh
#Install IDE plugins before Qodana execution (Applied in CI/CD pipeline)
#plugins:
# - id: <plugin.id> #(plugin id can be found at https://plugins.jetbrains.com)
#Specify Qodana linter for analysis (Applied in CI/CD pipeline)
linter: jetbrains/qodana-go:latest

51
types/globals.go Normal file
View File

@ -0,0 +1,51 @@
package types
import "chat/utils"
const (
GPT3Turbo = "gpt-3.5-turbo"
GPT3Turbo0613 = "gpt-3.5-turbo-0613"
GPT3Turbo0301 = "gpt-3.5-turbo-0301"
GPT3Turbo16k = "gpt-3.5-turbo-16k"
GPT3Turbo16k0613 = "gpt-3.5-turbo-16k-0613"
GPT3Turbo16k0301 = "gpt-3.5-turbo-16k-0301"
GPT4 = "gpt-4"
GPT40314 = "gpt-4-0314"
GPT40613 = "gpt-4-0613"
GPT432k = "gpt-4-32k"
GPT432k0314 = "gpt-4-32k-0314"
GPT432k0613 = "gpt-4-32k-0613"
Dalle = "dalle"
)
var GPT3TurboArray = []string{
GPT3Turbo,
GPT3Turbo0613,
GPT3Turbo0301,
}
var GPT3Turbo16kArray = []string{
GPT3Turbo16k,
GPT3Turbo16k0613,
GPT3Turbo16k0301,
}
var GPT4Array = []string{
GPT4,
GPT40314,
GPT40613,
}
var GPT432kArray = []string{
GPT432k,
GPT432k0314,
GPT432k0613,
}
func IsGPT4Model(model string) bool {
return utils.Contains(model, GPT4Array) || utils.Contains(model, GPT432kArray)
}
func IsGPT3TurboModel(model string) bool {
return utils.Contains(model, GPT3TurboArray) || utils.Contains(model, GPT3Turbo16kArray)
}

View File

@ -13,22 +13,29 @@ import (
func GetWeightByModel(model string) int {
switch model {
case "gpt-3.5-turbo-0613",
"gpt-3.5-turbo-16k-0613",
"gpt-4-0314",
"gpt-4-32k-0314",
"gpt-4-0613",
"gpt-4-32k-0613":
case types.GPT432k,
types.GPT432k0613,
types.GPT432k0314:
return 3 * 10
case types.GPT3Turbo,
types.GPT3Turbo0613,
types.GPT3Turbo16k,
types.GPT3Turbo16k0613,
types.GPT4,
types.GPT40314,
types.GPT40613:
return 3
case "gpt-3.5-turbo-0301":
case types.GPT3Turbo0301, types.GPT3Turbo16k0301:
return 4 // every message follows <|start|>{role/name}\n{content}<|end|>\n
default:
if strings.Contains(model, "gpt-3.5-turbo") {
if strings.Contains(model, types.GPT3Turbo) {
// warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.
return GetWeightByModel("gpt-3.5-turbo-0613")
} else if strings.Contains(model, "gpt-4") {
return GetWeightByModel(types.GPT3Turbo0613)
} else if strings.Contains(model, types.GPT4) {
// warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.
return GetWeightByModel("gpt-4-0613")
return GetWeightByModel(types.GPT40613)
} else {
// not implemented: See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens
panic(fmt.Errorf("not implemented for model %s", model))
@ -55,6 +62,6 @@ func NumTokensFromMessages(messages []types.ChatGPTMessage, model string) (token
return tokens
}
func CountTokenPrice(messages []types.ChatGPTMessage) int {
return NumTokensFromMessages(messages, "gpt-4")
func CountTokenPrice(messages []types.ChatGPTMessage, model string) int {
return NumTokensFromMessages(messages, model)
}