GithubProxy/coai
1
0
Fork 0
mirror of https://github.com/coaidev/coai.git synced 2025-05-23 23:10:13 +09:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix subscription usage increasing when request is error, update claude models

Browse Source
This commit is contained in:
Zhang Minghan 2023-11-03 17:37:57 +08:00
parent ebd8a6f1c6
commit a22f14a123
26 changed files with 183 additions and 287 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
README.md
View File

@ -64,8 +64,8 @@
- GPT-3.5-Turbo-Instruct
- GPT-4 (_0314_, _0613_)
- GPT-4-32k (_0314_, _0613_)
- GPT-4-Reverse (_gpt-4_, _**gpt-4v**_)
- DALL-E
- GPT-4-Reverse (_gpt-4_, _**gpt-4v**_, _**dalle3**_)
- DALL-E 2
- Claude
- Slack-Claude (unstable)
- Claude-2
@ -83,6 +83,13 @@
- Creative
- Balanced
- Precise
- ChatGLM
- Pro
- Std
- Lite
- DashScope Tongyi
- Qwen Plus (net)
- Qwen Turbo (net)
- More models are under development...

6
adapter/adapter.go
View File

@ -4,6 +4,7 @@ import (
"chat/adapter/bing"
"chat/adapter/claude"
"chat/adapter/dashscope"
"chat/adapter/oneapi"
"chat/adapter/palm2"
"chat/adapter/slack"
"chat/adapter/zhipuai"
@ -20,7 +21,10 @@ type ChatProps struct {
}
func NewChatRequest(props *ChatProps, hook globals.Hook) error {
if globals.IsChatGPTModel(props.Model) {
if oneapi.IsHit(props.Model) {
return oneapi.Handle(props, hook)
} else if globals.IsChatGPTModel(props.Model) {
return createRetryChatGPTPool(props, hook)
} else if globals.IsClaudeModel(props.Model) {

39
adapter/chatgpt/chat.go
View File

@ -5,7 +5,6 @@ import (
"chat/globals"
"chat/utils"
"fmt"
"github.com/spf13/viper"
"strings"
)
@ -135,41 +134,3 @@ func (c *ChatInstance) CreateStreamChatRequest(props *ChatProps, callback global
},
)
}
func (c *ChatInstance) Test() bool {
result, err := c.CreateChatRequest(&ChatProps{
Model: globals.GPT3Turbo,
Message: []globals.Message{{Role: "user", Content: "hi"}},
Token: 1,
})
if err != nil {
fmt.Println(fmt.Sprintf("%s: test failed (%s)", c.GetApiKey(), err.Error()))
}
return err == nil && len(result) > 0
}
func FilterKeys(v string) []string {
endpoint := viper.GetString(fmt.Sprintf("openai.%s.endpoint", v))
keys := strings.Split(viper.GetString(fmt.Sprintf("openai.%s.apikey", v)), "|")
return FilterKeysNative(endpoint, keys)
}
func FilterKeysNative(endpoint string, keys []string) []string {
stack := make(chan string, len(keys))
for _, key := range keys {
go func(key string) {
instance := NewChatInstance(endpoint, key)
stack <- utils.Multi[string](instance.Test(), key, "")
}(key)
}
var result []string
for i := 0; i < len(keys); i++ {
if res := <-stack; res != "" {
result = append(result, res)
}
}
return result
}

6
adapter/chatgpt/struct.go
View File

@ -48,12 +48,10 @@ func NewChatInstanceFromConfig(v string) *ChatInstance {
func NewChatInstanceFromModel(props *InstanceProps) *ChatInstance {
switch props.Model {
case globals.GPT4, globals.GPT40314, globals.GPT40613:
case globals.GPT4, globals.GPT40314, globals.GPT40613,
globals.GPT432k, globals.GPT432k0613, globals.GPT432k0314:
return NewChatInstanceFromConfig("gpt4")
case globals.GPT432k, globals.GPT432k0613, globals.GPT432k0314:
return NewChatInstanceFromConfig("32k")
case globals.GPT4Vision, globals.Dalle3:
return NewChatInstanceFromConfig("reverse")

47
adapter/chatgpt/test.go Normal file
View File

@ -0,0 +1,47 @@
package chatgpt
import (
"chat/globals"
"chat/utils"
"fmt"
"github.com/spf13/viper"
"strings"
)
func (c *ChatInstance) Test() bool {
result, err := c.CreateChatRequest(&ChatProps{
Model: globals.GPT3Turbo,
Message: []globals.Message{{Role: "user", Content: "hi"}},
Token: 1,
})
if err != nil {
fmt.Println(fmt.Sprintf("%s: test failed (%s)", c.GetApiKey(), err.Error()))
}
return err == nil && len(result) > 0
}
func FilterKeys(v string) []string {
endpoint := viper.GetString(fmt.Sprintf("openai.%s.endpoint", v))
keys := strings.Split(viper.GetString(fmt.Sprintf("openai.%s.apikey", v)), "|")
return FilterKeysNative(endpoint, keys)
}
func FilterKeysNative(endpoint string, keys []string) []string {
stack := make(chan string, len(keys))
for _, key := range keys {
go func(key string) {
instance := NewChatInstance(endpoint, key)
stack <- utils.Multi[string](instance.Test(), key, "")
}(key)
}
var result []string
for i := 0; i < len(keys); i++ {
if res := <-stack; res != "" {
result = append(result, res)
}
}
return result
}

87
adapter/oneapi/chat.go
View File

@ -1,11 +1,10 @@
package chatgpt
package oneapi
import "C"
import (
"chat/globals"
"chat/utils"
"fmt"
"github.com/spf13/viper"
"strings"
)
@ -15,36 +14,11 @@ type ChatProps struct {
Token int
}
func (c *ChatInstance) GetChatEndpoint(props *ChatProps) string {
if props.Model == globals.GPT3TurboInstruct {
return fmt.Sprintf("%s/v1/completions", c.GetEndpoint())
}
func (c *ChatInstance) GetChatEndpoint() string {
return fmt.Sprintf("%s/v1/chat/completions", c.GetEndpoint())
}
func (c *ChatInstance) GetCompletionPrompt(messages []globals.Message) string {
result := ""
for _, message := range messages {
result += fmt.Sprintf("%s: %s\n", message.Role, message.Content)
}
return result
}
func (c *ChatInstance) GetChatBody(props *ChatProps, stream bool) interface{} {
if props.Model == globals.GPT3TurboInstruct {
// for completions
return utils.Multi[interface{}](props.Token != -1, CompletionRequest{
Model: props.Model,
Prompt: c.GetCompletionPrompt(props.Message),
MaxToken: props.Token,
Stream: stream,
}, CompletionWithInfinity{
Model: props.Model,
Prompt: c.GetCompletionPrompt(props.Message),
Stream: stream,
})
}
if props.Token != -1 {
return ChatRequest{
Model: props.Model,
@ -61,42 +35,41 @@ func (c *ChatInstance) GetChatBody(props *ChatProps, stream bool) interface{} {
}
}
// CreateChatRequest is the native http request body for chatgpt
// CreateChatRequest is the native http request body for oneapi
func (c *ChatInstance) CreateChatRequest(props *ChatProps) (string, error) {
res, err := utils.Post(
c.GetChatEndpoint(props),
c.GetChatEndpoint(),
c.GetHeader(),
c.GetChatBody(props, false),
)
if err != nil || res == nil {
return "", fmt.Errorf("chatgpt error: %s", err.Error())
return "", fmt.Errorf("oneapi error: %s", err.Error())
}
data := utils.MapToStruct[ChatResponse](res)
if data == nil {
return "", fmt.Errorf("chatgpt error: cannot parse response")
return "", fmt.Errorf("oneapi error: cannot parse response")
} else if data.Error.Message != "" {
return "", fmt.Errorf("chatgpt error: %s", data.Error.Message)
return "", fmt.Errorf("oneapi error: %s", data.Error.Message)
}
return data.Choices[0].Message.Content, nil
}
// CreateStreamChatRequest is the stream response body for chatgpt
// CreateStreamChatRequest is the stream response body for oneapi
func (c *ChatInstance) CreateStreamChatRequest(props *ChatProps, callback globals.Hook) error {
buf := ""
instruct := props.Model == globals.GPT3TurboInstruct
return utils.EventSource(
"POST",
c.GetChatEndpoint(props),
c.GetChatEndpoint(),
c.GetHeader(),
c.GetChatBody(props, true),
func(data string) error {
data, err := c.ProcessLine(instruct, buf, data)
data, err := c.ProcessLine(buf, data)
if err != nil {
if strings.HasPrefix(err.Error(), "chatgpt error") {
if strings.HasPrefix(err.Error(), "oneapi error") {
return err
}
@ -115,41 +88,3 @@ func (c *ChatInstance) CreateStreamChatRequest(props *ChatProps, callback global
},
)
}
func (c *ChatInstance) Test() bool {
result, err := c.CreateChatRequest(&ChatProps{
Model: globals.GPT3Turbo,
Message: []globals.Message{{Role: "user", Content: "hi"}},
Token: 1,
})
if err != nil {
fmt.Println(fmt.Sprintf("%s: test failed (%s)", c.GetApiKey(), err.Error()))
}
return err == nil && len(result) > 0
}
func FilterKeys(v string) []string {
endpoint := viper.GetString(fmt.Sprintf("openai.%s.endpoint", v))
keys := strings.Split(viper.GetString(fmt.Sprintf("openai.%s.apikey", v)), "|")
return FilterKeysNative(endpoint, keys)
}
func FilterKeysNative(endpoint string, keys []string) []string {
stack := make(chan string, len(keys))
for _, key := range keys {
go func(key string) {
instance := NewChatInstance(endpoint, key)
stack <- utils.Multi[string](instance.Test(), key, "")
}(key)
}
var result []string
for i := 0; i < len(keys); i++ {
if res := <-stack; res != "" {
result = append(result, res)
}
}
return result
}

32
adapter/oneapi/globals.go Normal file
View File

@ -0,0 +1,32 @@
package oneapi
import (
"chat/globals"
)
var HitModels = []string{
globals.Claude2, globals.Claude2100k,
}
func (c *ChatInstance) Process(data string) string {
return data
}
func (c *ChatInstance) FormatMessage(message []globals.Message) []globals.Message {
return message
}
func (c *ChatInstance) FormatModel(model string) string {
return model
}
func (c *ChatInstance) GetToken(model string) int {
switch model {
case globals.Claude2:
return 5000
case globals.Claude2100k:
return 50000
default:
return 2500
}
}

30
adapter/oneapi/handler.go Normal file
View File

@ -0,0 +1,30 @@
package oneapi
import (
"chat/globals"
"chat/utils"
)
type AdapterProps struct {
Model string
Plan bool
Infinity bool
Message []globals.Message
Token int
}
func HandleRequest(props *AdapterProps, hook globals.Hook) error {
instance := NewChatInstanceFromConfig()
return instance.CreateStreamChatRequest(&ChatProps{
Model: instance.FormatModel(props.Model),
Message: instance.FormatMessage(props.Message),
Token: utils.Multi(props.Token == 0, instance.GetToken(props.Model), props.Token),
}, func(data string) error {
return hook(instance.Process(data))
})
}
func Handle(props interface{}, hook globals.Hook) error {
conv := utils.MapToStruct[AdapterProps](props)
return HandleRequest(conv, hook)
}

40
adapter/oneapi/processor.go
View File

@ -1,4 +1,4 @@
package chatgpt
package oneapi
import (
"chat/globals"
@ -54,21 +54,6 @@ func processChatResponse(data string) *ChatStreamResponse {
return nil
}
func processCompletionResponse(data string) *CompletionResponse {
if strings.HasPrefix(data, "{") {
var form *CompletionResponse
if form = utils.UnmarshalForm[CompletionResponse](data); form != nil {
return form
}
if form = utils.UnmarshalForm[CompletionResponse](data[:len(data)-1]); form != nil {
return form
}
}
return nil
}
func processChatErrorResponse(data string) *ChatStreamErrorResponse {
if strings.HasPrefix(data, "{") {
var form *ChatStreamErrorResponse
@ -98,38 +83,23 @@ func getChoices(form *ChatStreamResponse) string {
return form.Data.Choices[0].Delta.Content
}
func getCompletionChoices(form *CompletionResponse) string {
if len(form.Data.Choices) == 0 {
return ""
}
return form.Data.Choices[0].Text
}
func (c *ChatInstance) ProcessLine(instruct bool, buf, data string) (string, error) {
func (c *ChatInstance) ProcessLine(buf, data string) (string, error) {
item := processFormat(buf + data)
if isDone(item) {
return "", nil
}
if form := processChatResponse(item); form == nil {
if instruct {
// legacy support
if completion := processCompletionResponse(item); completion != nil {
return getCompletionChoices(completion), nil
}
}
// recursive call
if len(buf) > 0 {
return c.ProcessLine(instruct, "", buf+item)
return c.ProcessLine("", buf+item)
}
if err := processChatErrorResponse(item); err == nil {
globals.Warn(fmt.Sprintf("chatgpt error: cannot parse response: %s", item))
globals.Warn(fmt.Sprintf("oneapi error: cannot parse response: %s", item))
return data, errors.New("parser error: cannot parse response")
} else {
return "", fmt.Errorf("chatgpt error: %s (type: %s)", err.Data.Error.Message, err.Data.Error.Type)
return "", fmt.Errorf("oneapi error: %s (type: %s)", err.Data.Error.Message, err.Data.Error.Type)
}
} else {

31
adapter/oneapi/struct.go
View File

@ -1,7 +1,6 @@
package chatgpt
package oneapi
import (
"chat/globals"
"chat/utils"
"fmt"
"github.com/spf13/viper"
@ -39,31 +38,13 @@ func NewChatInstance(endpoint, apiKey string) *ChatInstance {
}
}
func NewChatInstanceFromConfig(v string) *ChatInstance {
func NewChatInstanceFromConfig() *ChatInstance {
return NewChatInstance(
viper.GetString(fmt.Sprintf("openai.%s.endpoint", v)),
utils.GetRandomKey(viper.GetString(fmt.Sprintf("openai.%s.apikey", v))),
viper.GetString("oneapi.endpoint"),
viper.GetString("oneapi.apikey"),
)
}
func NewChatInstanceFromModel(props *InstanceProps) *ChatInstance {
switch props.Model {
case globals.GPT4, globals.GPT40314, globals.GPT40613:
return NewChatInstanceFromConfig("gpt4")
case globals.GPT432k, globals.GPT432k0613, globals.GPT432k0314:
return NewChatInstanceFromConfig("32k")
case globals.GPT4Vision, globals.Dalle3:
return NewChatInstanceFromConfig("reverse")
case globals.GPT3Turbo, globals.GPT3TurboInstruct, globals.GPT3Turbo0613, globals.GPT3Turbo0301,
globals.GPT3Turbo16k, globals.GPT3Turbo16k0301, globals.GPT3Turbo16k0613:
if props.Plan {
return NewChatInstanceFromConfig("subscribe")
}
return NewChatInstanceFromConfig("gpt3")
default:
return NewChatInstanceFromConfig("gpt3")
}
func IsHit(model string) bool {
return utils.Contains[string](model, HitModels)
}

60
adapter/oneapi/types.go
View File

@ -1,8 +1,8 @@
package chatgpt
package oneapi
import "chat/globals"
// ChatRequest is the request body for chatgpt
// ChatRequest is the request body for oneapi
type ChatRequest struct {
Model string `json:"model"`
Messages []globals.Message `json:"messages"`
@ -16,21 +16,7 @@ type ChatRequestWithInfinity struct {
Stream bool `json:"stream"`
}
// CompletionRequest ChatRequest is the request body for chatgpt completion
type CompletionRequest struct {
Model string `json:"model"`
Prompt string `json:"prompt"`
MaxToken int `json:"max_tokens"`
Stream bool `json:"stream"`
}
type CompletionWithInfinity struct {
Model string `json:"model"`
Prompt string `json:"prompt"`
Stream bool `json:"stream"`
}
// ChatResponse is the native http request body for chatgpt
// ChatResponse is the native http request body for oneapi
type ChatResponse struct {
ID string `json:"id"`
Object string `json:"object"`
@ -46,7 +32,7 @@ type ChatResponse struct {
} `json:"error"`
}
// ChatStreamResponse is the stream response body for chatgpt
// ChatStreamResponse is the stream response body for oneapi
type ChatStreamResponse struct {
ID string `json:"id"`
Object string `json:"object"`
@ -62,20 +48,6 @@ type ChatStreamResponse struct {
} `json:"data"`
}
// CompletionResponse is the native http request body / stream response body for chatgpt completion
type CompletionResponse struct {
ID string `json:"id"`
Object string `json:"object"`
Created int64 `json:"created"`
Model string `json:"model"`
Data struct {
Choices []struct {
Text string `json:"text"`
Index int `json:"index"`
} `json:"choices"`
} `json:"data"`
}
type ChatStreamErrorResponse struct {
Data struct {
Error struct {
@ -84,27 +56,3 @@ type ChatStreamErrorResponse struct {
} `json:"error"`
} `json:"data"`
}
type ImageSize string
// ImageRequest is the request body for chatgpt dalle image generation
type ImageRequest struct {
Prompt string `json:"prompt"`
Size ImageSize `json:"size"`
N int `json:"n"`
}
type ImageResponse struct {
Data []struct {
Url string `json:"url"`
} `json:"data"`
Error struct {
Message string `json:"message"`
} `json:"error"`
}
var (
ImageSize256 ImageSize = "256x256"
ImageSize512 ImageSize = "512x512"
ImageSize1024 ImageSize = "1024x1024"
)

15
adapter/request.go
View File

@ -32,21 +32,6 @@ func retryChatGPTPool(props *ChatProps, hook globals.Hook, retry int) error {
),
}, hook)
if globals.IsGPT4NativeModel(props.Model) && IsAvailableError(err) {
if !strings.Contains(err.Error(), "429") {
// not rate limited
return chatgpt.NewChatInstanceFromConfig("32k").CreateStreamChatRequest(&chatgpt.ChatProps{
Model: props.Model,
Message: props.Message,
Token: utils.Multi(
props.Token == 0,
utils.Multi(globals.IsGPT4Model(props.Model) || props.Plan || props.Infinity, -1, 2500),
props.Token,
),
}, hook)
}
}
if IsAvailableError(err) && retry < MaxRetries {
fmt.Println(fmt.Sprintf("retrying chatgpt pool (times: %d, error: %s)", retry+1, err.Error()))
return retryChatGPTPool(props, hook, retry+1)

1
addition/generation/api.go
View File

@ -86,6 +86,7 @@ func GenerateAPI(c *gin.Context) {
}
if err != nil {
auth.RevertSubscriptionUsage(cache, user, form.Model, plan)
conn.Send(globals.GenerationSegmentResponse{
End: true,
Error: err.Error(),

4
app/src/components/home/ModelSelector.tsx
View File

@ -38,14 +38,14 @@ function ModelSelector(props: ModelSelectorProps) {
});
const list = supportModels.map((model: Model): SelectItemProps => {
const array = ["gpt-4-0613", "gpt-4v", "gpt-4-dalle", "claude-2"];
const array = ["gpt-4-0613", "gpt-4v", "gpt-4-dalle", "claude-2-100k"];
if (subscription && array.includes(model.id)) {
return {
name: model.id,
value: model.name,
badge: { variant: "gold", name: "plus" },
} as SelectItemProps;
} else if (student && model.id === "claude-2") {
} else if (student && model.id === "claude-2-100k") {
return {
name: model.id,
value: model.name,

10
app/src/conf.ts
View File

@ -8,7 +8,7 @@ import {
} from "@/utils/env.ts";
import { getMemory } from "@/utils/memory.ts";
export const version = "3.6.9";
export const version = "3.6.9rc";
export const dev: boolean = getDev();
export const deploy: boolean = true;
export let rest_api: string = getRestApi(deploy);
@ -24,15 +24,15 @@ export const supportModels: Model[] = [
{ id: "gpt-4v", name: "GPT-4V", free: false, auth: true },
{ id: "gpt-4-dalle", name: "DALLE3", free: false, auth: true },
// anthropic models
{ id: "claude-2", name: "Claude-2", free: true, auth: false },
{ id: "claude-2-100k", name: "Claude-2-100k", free: false, auth: true },
// spark desk
{ id: "spark-desk-v3", name: "讯飞星火 V3", free: true, auth: true },
{ id: "spark-desk-v2", name: "讯飞星火 V2", free: true, auth: true },
{ id: "spark-desk-v2", name: "讯飞星火 V1.5", free: true, auth: true },
// anthropic models
{ id: "claude-1", name: "Claude-2", free: true, auth: false },
{ id: "claude-2", name: "Claude-2-100k", free: false, auth: true }, // not claude-2-100k
// dashscope models
{ id: "qwen-plus-net", name: "通义千问 Plus X", free: false, auth: true },
{ id: "qwen-plus", name: "通义千问 Plus", free: false, auth: true },

5
app/src/conversation/addition.ts
View File

@ -18,7 +18,6 @@ type SubscriptionResponse = {
enterprise?: boolean;
usage: {
gpt4: number;
dalle: number;
};
};
@ -67,7 +66,7 @@ export async function getSubscription(): Promise<SubscriptionResponse> {
status: false,
is_subscribed: false,
expired: 0,
usage: { gpt4: 0, dalle: 0 },
usage: { gpt4: 0 },
};
}
return resp.data as SubscriptionResponse;
@ -77,7 +76,7 @@ export async function getSubscription(): Promise<SubscriptionResponse> {
status: false,
is_subscribed: false,
expired: 0,
usage: { gpt4: 0, dalle: 0 },
usage: { gpt4: 0 },
};
}
}

9
app/src/dialogs/Subscription.tsx
View File

@ -31,7 +31,6 @@ import {
FolderGit2,
Globe,
Image,
ImagePlus,
LifeBuoy,
MessageSquare,
MessagesSquare,
@ -200,14 +199,6 @@ function Subscription() {
<p>{usage?.gpt4}</p> / <p> 50 </p>
</div>
</div>
<div className={`sub-column`}>
<ImagePlus className={`h-4 w-4 mr-1`} />
DALL-E
<div className={`grow`} />
<div className={`sub-value`}>
<p>{usage?.dalle}</p> / <p> 2000 </p>
</div>
</div>
</>
)}
</div>

1
app/src/store/subscription.ts
View File

@ -11,7 +11,6 @@ export const subscriptionSlice = createSlice({
expired: 0,
usage: {
gpt4: 0,
dalle: 0,
},
},
reducers: {

11
auth/rule.go
View File

@ -50,8 +50,15 @@ func HandleSubscriptionUsage(db *sql.DB, cache *redis.Client, user *User, model
return false
}
// CanEnableModelWithSubscription returns (canEnable, usePlan)
func CanEnableModelWithSubscription(db *sql.DB, cache *redis.Client, user *User, model string) (bool, bool) {
func RevertSubscriptionUsage(cache *redis.Client, user *User, model string, plan bool) {
if globals.IsGPT4NativeModel(model) && plan {
DecreaseSubscriptionUsage(cache, user, globals.GPT4)
} else if model == globals.Claude2100k && !plan {
DecreaseSubscriptionUsage(cache, user, globals.Claude2100k)
}
}
func CanEnableModelWithSubscription(db *sql.DB, cache *redis.Client, user *User, model string) (canEnable bool, usePlan bool) {
// use subscription quota first
if user != nil && HandleSubscriptionUsage(db, cache, user, model) {
return true, true

4
auth/subscription.go
View File

@ -34,3 +34,7 @@ func BuySubscription(db *sql.DB, user *User, month int) bool {
func IncreaseSubscriptionUsage(cache *redis.Client, user *User, t string, limit int64) bool {
return utils.IncrWithLimit(cache, globals.GetSubscriptionLimitFormat(t, user.ID), 1, limit, 60*60*24) // 1 day
}
func DecreaseSubscriptionUsage(cache *redis.Client, user *User, t string) bool {
return utils.DecrInt(cache, globals.GetSubscriptionLimitFormat(t, user.ID), 1)
}

6
auth/user.go
View File

@ -4,7 +4,6 @@ import (
"chat/globals"
"chat/utils"
"database/sql"
"errors"
"fmt"
"github.com/dgrijalva/jwt-go"
"github.com/gin-gonic/gin"
@ -219,10 +218,7 @@ func (u *User) CreateApiKey(db *sql.DB) string {
func (u *User) GetApiKey(db *sql.DB) string {
var key string
if err := db.QueryRow("SELECT api_key FROM apikey WHERE user_id = ?", u.GetID(db)).Scan(&key); err != nil {
if errors.Is(err, sql.ErrNoRows) {
return u.CreateApiKey(db)
}
return ""
return u.CreateApiKey(db)
}
return key
}

4
globals/variables.go
View File

@ -54,8 +54,8 @@ const (
GPT432k0613 = "gpt-4-32k-0613"
Dalle2 = "dalle"
Dalle3 = "gpt-4-dalle"
Claude2 = "claude-1" // claude v1.3
Claude2100k = "claude-2"
Claude2 = "claude-2"
Claude2100k = "claude-2-100k"
ClaudeSlack = "claude-slack"
SparkDesk = "spark-desk-v1.5"
SparkDeskV2 = "spark-desk-v2"

1
manager/chat.go
View File

@ -106,6 +106,7 @@ func ChatHandler(conn *Connection, user *auth.User, instance *conversation.Conve
if err != nil && err.Error() != "signal" {
globals.Warn(fmt.Sprintf("caught error from chat handler: %s (instance: %s, client: %s)", err, model, conn.GetCtx().ClientIP()))
auth.RevertSubscriptionUsage(cache, user, model, plan)
CollectQuota(conn.GetCtx(), user, buffer, plan)
conn.Send(globals.ChatSegmentResponse{
Message: err.Error(),

1
manager/completions.go
View File

@ -46,6 +46,7 @@ func NativeChatHandler(c *gin.Context, user *auth.User, model string, message []
buffer.Write(resp)
return nil
}); err != nil {
auth.RevertSubscriptionUsage(cache, user, model, plan)
CollectQuota(c, user, buffer, plan)
return keyword, err.Error(), GetErrorQuota(model)
}

5
middleware/throttle.go
View File

@ -28,9 +28,7 @@ func (l *Limiter) RateLimit(ctx *gin.Context, rds *redis.Client, ip string, path
var limits = map[string]Limiter{
"/login": {Duration: 10, Count: 5},
"/anonymous": {Duration: 60, Count: 15},
"/card": {Duration: 1, Count: 5},
"/user": {Duration: 1, Count: 1},
"/apikey": {Duration: 1, Count: 2},
"/package": {Duration: 1, Count: 2},
"/quota": {Duration: 1, Count: 2},
"/buy": {Duration: 1, Count: 2},
@ -41,6 +39,7 @@ var limits = map[string]Limiter{
"/invite": {Duration: 7200, Count: 20},
"/v1": {Duration: 1, Count: 600},
"/card": {Duration: 1, Count: 5},
"/generation": {Duration: 1, Count: 5},
"/article": {Duration: 1, Count: 5},
}

4
utils/tokenizer.go
View File

@ -88,7 +88,7 @@ func CountInputToken(model string, v []globals.Message) float32 {
case globals.Claude2:
return 0
case globals.Claude2100k:
return float32(CountTokenPrice(v, model)) / 1000 * 0.008
return float32(CountTokenPrice(v, model)) / 1000 * 0.05
case globals.ZhiPuChatGLMPro:
return float32(CountTokenPrice(v, model)) / 1000 * 0.1
case globals.ZhiPuChatGLMStd:
@ -118,7 +118,7 @@ func CountOutputToken(model string, t int) float32 {
case globals.Claude2:
return 0
case globals.Claude2100k:
return float32(t*GetWeightByModel(model)) / 1000 * 0.008
return float32(t*GetWeightByModel(model)) / 1000 * 0.05
case globals.ZhiPuChatGLMPro:
return float32(t*GetWeightByModel(model)) / 1000 * 0.1
case globals.ZhiPuChatGLMStd: