fix: fix PaLM2 not billed

README.md

@@ -61,6 +61,9 @@ _✨ All in one 的 OpenAI 接口，整合各种 API 访问方式，开箱即用
 1. 支持多种 API 访问渠道：
    + [x] OpenAI 官方通道（支持配置镜像）
    + [x] **Azure OpenAI API**
+   + [x] [Anthropic Claude 系列模型](https://anthropic.com)
+   + [x] [Google PaLM2 系列模型](https://developers.generativeai.google)
+   + [x] [百度文心一言系列模型](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [API Distribute](https://api.gptjk.top/register?aff=QGxj)
    + [x] [OpenAI-SB](https://openai-sb.com)
    + [x] [API2D](https://api2d.com/r/197971)

common/constants.go

@@ -151,6 +151,8 @@ const (
 	ChannelTypePaLM      = 11
 	ChannelTypeAPI2GPT   = 12
 	ChannelTypeAIGC2D    = 13
+	ChannelTypeAnthropic = 14
+	ChannelTypeBaidu     = 15
 )
 
 var ChannelBaseURLs = []string{
@@ -168,4 +170,6 @@ var ChannelBaseURLs = []string{
 	"",                              // 11
 	"https://api.api2gpt.com",       // 12
 	"https://api.aigc2d.com",        // 13
+	"https://api.anthropic.com",     // 14
+	"https://aip.baidubce.com",      // 15
 }

common/model-ratio.go

@@ -4,6 +4,7 @@ import "encoding/json"
 
 // ModelRatio
 // https://platform.openai.com/docs/models/model-endpoint-compatibility
+// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Blfmc9dlf
 // https://openai.com/pricing
 // TODO: when a new api is enabled, check the pricing here
 // 1 === $0.002 / 1K tokens
@@ -36,6 +37,11 @@ var ModelRatio = map[string]float64{
 	"text-moderation-stable":  0.1,
 	"text-moderation-latest":  0.1,
 	"dall-e":                  8,
+	"claude-instant-1":        0.75,
+	"claude-2":                30,
+	"ERNIE-Bot":               1,    // 0.012元/千tokens
+	"ERNIE-Bot-turbo":         0.67, // 0.008元/千tokens
+	"PaLM-2":                  1,
 }
 
 func ModelRatio2JSONString() string {

controller/billing.go

@@ -7,16 +7,19 @@ import (
 )
 
 func GetSubscription(c *gin.Context) {
-	var quota int
+	var remainQuota int
+	var usedQuota int
 	var err error
 	var token *model.Token
 	if common.DisplayTokenStatEnabled {
 		tokenId := c.GetInt("token_id")
 		token, err = model.GetTokenById(tokenId)
-		quota = token.RemainQuota
+		remainQuota = token.RemainQuota
+		usedQuota = token.UsedQuota
 	} else {
 		userId := c.GetInt("id")
-		quota, err = model.GetUserQuota(userId)
+		remainQuota, err = model.GetUserQuota(userId)
+		usedQuota, err = model.GetUserUsedQuota(userId)
 	}
 	if err != nil {
 		openAIError := OpenAIError{
@@ -28,6 +31,7 @@ func GetSubscription(c *gin.Context) {
 		})
 		return
 	}
+	quota := remainQuota + usedQuota
 	amount := float64(quota)
 	if common.DisplayInCurrencyEnabled {
 		amount /= common.QuotaPerUnit

controller/channel-test.go

@@ -14,7 +14,7 @@ import (
 	"time"
 )
 
-func testChannel(channel *model.Channel, request ChatRequest) error {
+func testChannel(channel *model.Channel, request ChatRequest) (error, *OpenAIError) {
 	switch channel.Type {
 	case common.ChannelTypeAzure:
 		request.Model = "gpt-35-turbo"
@@ -33,11 +33,11 @@ func testChannel(channel *model.Channel, request ChatRequest) error {
 
 	jsonData, err := json.Marshal(request)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	req, err := http.NewRequest("POST", requestURL, bytes.NewBuffer(jsonData))
 	if err != nil {
-		return err
+		return err, nil
 	}
 	if channel.Type == common.ChannelTypeAzure {
 		req.Header.Set("api-key", channel.Key)
@@ -48,18 +48,18 @@ func testChannel(channel *model.Channel, request ChatRequest) error {
 	client := &http.Client{}
 	resp, err := client.Do(req)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	defer resp.Body.Close()
 	var response TextResponse
 	err = json.NewDecoder(resp.Body).Decode(&response)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	if response.Usage.CompletionTokens == 0 {
-		return errors.New(fmt.Sprintf("type %s, code %v, message %s", response.Error.Type, response.Error.Code, response.Error.Message))
+		return errors.New(fmt.Sprintf("type %s, code %v, message %s", response.Error.Type, response.Error.Code, response.Error.Message)), &response.Error
 	}
-	return nil
+	return nil, nil
 }
 
 func buildTestRequest() *ChatRequest {
@@ -94,7 +94,7 @@ func TestChannel(c *gin.Context) {
 	}
 	testRequest := buildTestRequest()
 	tik := time.Now()
-	err = testChannel(channel, *testRequest)
+	err, _ = testChannel(channel, *testRequest)
 	tok := time.Now()
 	milliseconds := tok.Sub(tik).Milliseconds()
 	go channel.UpdateResponseTime(milliseconds)
@@ -158,13 +158,14 @@ func testAllChannels(notify bool) error {
 				continue
 			}
 			tik := time.Now()
-			err := testChannel(channel, *testRequest)
+			err, openaiErr := testChannel(channel, *testRequest)
 			tok := time.Now()
 			milliseconds := tok.Sub(tik).Milliseconds()
-			if err != nil || milliseconds > disableThreshold {
-				if milliseconds > disableThreshold {
-					err = errors.New(fmt.Sprintf("响应时间 %.2fs 超过阈值 %.2fs", float64(milliseconds)/1000.0, float64(disableThreshold)/1000.0))
-				}
+			if milliseconds > disableThreshold {
+				err = errors.New(fmt.Sprintf("响应时间 %.2fs 超过阈值 %.2fs", float64(milliseconds)/1000.0, float64(disableThreshold)/1000.0))
+				disableChannel(channel.Id, channel.Name, err.Error())
+			}
+			if shouldDisableChannel(openaiErr) {
 				disableChannel(channel.Id, channel.Name, err.Error())
 			}
 			channel.UpdateResponseTime(milliseconds)

controller/model.go

@@ -270,6 +270,51 @@ func init() {
 			Root:       "ChatGLM2",
 			Parent:     nil,
 		},
+		{
+			Id:         "claude-instant-1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anturopic",
+			Permission: permission,
+			Root:       "claude-instant-1",
+			Parent:     nil,
+		},
+		{
+			Id:         "claude-2",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anturopic",
+			Permission: permission,
+			Root:       "claude-2",
+			Parent:     nil,
+		},
+		{
+			Id:         "ERNIE-Bot",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "baidu",
+			Permission: permission,
+			Root:       "ERNIE-Bot",
+			Parent:     nil,
+		},
+		{
+			Id:         "ERNIE-Bot-turbo",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "baidu",
+			Permission: permission,
+			Root:       "ERNIE-Bot-turbo",
+			Parent:     nil,
+		},
+		{
+			Id:         "PaLM-2",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "google",
+			Permission: permission,
+			Root:       "PaLM-2",
+			Parent:     nil,
+		},
 	}
 	openAIModelsMap = make(map[string]OpenAIModels)
 	for _, model := range openAIModels {

controller/relay-baidu.go

@@ -0,0 +1,203 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/flfmc9do2
+
+type BaiduTokenResponse struct {
+	RefreshToken  string `json:"refresh_token"`
+	ExpiresIn     int    `json:"expires_in"`
+	SessionKey    string `json:"session_key"`
+	AccessToken   string `json:"access_token"`
+	Scope         string `json:"scope"`
+	SessionSecret string `json:"session_secret"`
+}
+
+type BaiduMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type BaiduChatRequest struct {
+	Messages []BaiduMessage `json:"messages"`
+	Stream   bool           `json:"stream"`
+	UserId   string         `json:"user_id,omitempty"`
+}
+
+type BaiduError struct {
+	ErrorCode int    `json:"error_code"`
+	ErrorMsg  string `json:"error_msg"`
+}
+
+type BaiduChatResponse struct {
+	Id               string `json:"id"`
+	Object           string `json:"object"`
+	Created          int64  `json:"created"`
+	Result           string `json:"result"`
+	IsTruncated      bool   `json:"is_truncated"`
+	NeedClearHistory bool   `json:"need_clear_history"`
+	Usage            Usage  `json:"usage"`
+	BaiduError
+}
+
+type BaiduChatStreamResponse struct {
+	BaiduChatResponse
+	SentenceId int  `json:"sentence_id"`
+	IsEnd      bool `json:"is_end"`
+}
+
+func requestOpenAI2Baidu(request GeneralOpenAIRequest) *BaiduChatRequest {
+	messages := make([]BaiduMessage, 0, len(request.Messages))
+	for _, message := range request.Messages {
+		messages = append(messages, BaiduMessage{
+			Role:    message.Role,
+			Content: message.Content,
+		})
+	}
+	return &BaiduChatRequest{
+		Messages: messages,
+		Stream:   request.Stream,
+	}
+}
+
+func responseBaidu2OpenAI(response *BaiduChatResponse) *OpenAITextResponse {
+	choice := OpenAITextResponseChoice{
+		Index: 0,
+		Message: Message{
+			Role:    "assistant",
+			Content: response.Result,
+		},
+		FinishReason: "stop",
+	}
+	fullTextResponse := OpenAITextResponse{
+		Id:      response.Id,
+		Object:  "chat.completion",
+		Created: response.Created,
+		Choices: []OpenAITextResponseChoice{choice},
+		Usage:   response.Usage,
+	}
+	return &fullTextResponse
+}
+
+func streamResponseBaidu2OpenAI(baiduResponse *BaiduChatStreamResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = baiduResponse.Result
+	choice.FinishReason = "stop"
+	response := ChatCompletionsStreamResponse{
+		Id:      baiduResponse.Id,
+		Object:  "chat.completion.chunk",
+		Created: baiduResponse.Created,
+		Model:   "ernie-bot",
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+	return &response
+}
+
+func baiduStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var usage Usage
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+		if atEOF && len(data) == 0 {
+			return 0, nil, nil
+		}
+		if i := strings.Index(string(data), "\n"); i >= 0 {
+			return i + 1, data[0:i], nil
+		}
+		if atEOF {
+			return len(data), data, nil
+		}
+		return 0, nil, nil
+	})
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			if len(data) < 6 { // ignore blank line or wrong format
+				continue
+			}
+			data = data[6:]
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			var baiduResponse BaiduChatStreamResponse
+			err := json.Unmarshal([]byte(data), &baiduResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			usage.PromptTokens += baiduResponse.Usage.PromptTokens
+			usage.CompletionTokens += baiduResponse.Usage.CompletionTokens
+			usage.TotalTokens += baiduResponse.Usage.TotalTokens
+			response := streamResponseBaidu2OpenAI(&baiduResponse)
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, &usage
+}
+
+func baiduHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var baiduResponse BaiduChatResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &baiduResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if baiduResponse.ErrorMsg != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: baiduResponse.ErrorMsg,
+				Type:    "baidu_error",
+				Param:   "",
+				Code:    baiduResponse.ErrorCode,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseBaidu2OpenAI(&baiduResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &fullTextResponse.Usage
+}

controller/relay-claude.go

@@ -0,0 +1,221 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+type ClaudeMetadata struct {
+	UserId string `json:"user_id"`
+}
+
+type ClaudeRequest struct {
+	Model             string   `json:"model"`
+	Prompt            string   `json:"prompt"`
+	MaxTokensToSample int      `json:"max_tokens_to_sample"`
+	StopSequences     []string `json:"stop_sequences,omitempty"`
+	Temperature       float64  `json:"temperature,omitempty"`
+	TopP              float64  `json:"top_p,omitempty"`
+	TopK              int      `json:"top_k,omitempty"`
+	//ClaudeMetadata    `json:"metadata,omitempty"`
+	Stream bool `json:"stream,omitempty"`
+}
+
+type ClaudeError struct {
+	Type    string `json:"type"`
+	Message string `json:"message"`
+}
+
+type ClaudeResponse struct {
+	Completion string      `json:"completion"`
+	StopReason string      `json:"stop_reason"`
+	Model      string      `json:"model"`
+	Error      ClaudeError `json:"error"`
+}
+
+func stopReasonClaude2OpenAI(reason string) string {
+	switch reason {
+	case "stop_sequence":
+		return "stop"
+	case "max_tokens":
+		return "length"
+	default:
+		return reason
+	}
+}
+
+func requestOpenAI2Claude(textRequest GeneralOpenAIRequest) *ClaudeRequest {
+	claudeRequest := ClaudeRequest{
+		Model:             textRequest.Model,
+		Prompt:            "",
+		MaxTokensToSample: textRequest.MaxTokens,
+		StopSequences:     nil,
+		Temperature:       textRequest.Temperature,
+		TopP:              textRequest.TopP,
+		Stream:            textRequest.Stream,
+	}
+	if claudeRequest.MaxTokensToSample == 0 {
+		claudeRequest.MaxTokensToSample = 1000000
+	}
+	prompt := ""
+	for _, message := range textRequest.Messages {
+		if message.Role == "user" {
+			prompt += fmt.Sprintf("\n\nHuman: %s", message.Content)
+		} else if message.Role == "assistant" {
+			prompt += fmt.Sprintf("\n\nAssistant: %s", message.Content)
+		} else {
+			// ignore other roles
+		}
+		prompt += "\n\nAssistant:"
+	}
+	claudeRequest.Prompt = prompt
+	return &claudeRequest
+}
+
+func streamResponseClaude2OpenAI(claudeResponse *ClaudeResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = claudeResponse.Completion
+	choice.FinishReason = stopReasonClaude2OpenAI(claudeResponse.StopReason)
+	var response ChatCompletionsStreamResponse
+	response.Object = "chat.completion.chunk"
+	response.Model = claudeResponse.Model
+	response.Choices = []ChatCompletionsStreamResponseChoice{choice}
+	return &response
+}
+
+func responseClaude2OpenAI(claudeResponse *ClaudeResponse) *OpenAITextResponse {
+	choice := OpenAITextResponseChoice{
+		Index: 0,
+		Message: Message{
+			Role:    "assistant",
+			Content: strings.TrimPrefix(claudeResponse.Completion, " "),
+			Name:    nil,
+		},
+		FinishReason: stopReasonClaude2OpenAI(claudeResponse.StopReason),
+	}
+	fullTextResponse := OpenAITextResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Choices: []OpenAITextResponseChoice{choice},
+	}
+	return &fullTextResponse
+}
+
+func claudeStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createdTime := common.GetTimestamp()
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+		if atEOF && len(data) == 0 {
+			return 0, nil, nil
+		}
+		if i := strings.Index(string(data), "\r\n\r\n"); i >= 0 {
+			return i + 4, data[0:i], nil
+		}
+		if atEOF {
+			return len(data), data, nil
+		}
+		return 0, nil, nil
+	})
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			if !strings.HasPrefix(data, "event: completion") {
+				continue
+			}
+			data = strings.TrimPrefix(data, "event: completion\r\ndata: ")
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			// some implementations may add \r at the end of data
+			data = strings.TrimSuffix(data, "\r")
+			var claudeResponse ClaudeResponse
+			err := json.Unmarshal([]byte(data), &claudeResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			responseText += claudeResponse.Completion
+			response := streamResponseClaude2OpenAI(&claudeResponse)
+			response.Id = responseId
+			response.Created = createdTime
+			jsonStr, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonStr)})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func claudeHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var claudeResponse ClaudeResponse
+	err = json.Unmarshal(responseBody, &claudeResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if claudeResponse.Error.Type != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: claudeResponse.Error.Message,
+				Type:    claudeResponse.Error.Type,
+				Param:   "",
+				Code:    claudeResponse.Error.Type,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseClaude2OpenAI(&claudeResponse)
+	completionTokens := countTokenText(claudeResponse.Completion, model)
+	usage := Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: completionTokens,
+		TotalTokens:      promptTokens + completionTokens,
+	}
+	fullTextResponse.Usage = usage
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &usage
+}

controller/relay-openai.go

@@ -0,0 +1,133 @@
+package controller
+
+import (
+	"bufio"
+	"bytes"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+func openaiStreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+		if atEOF && len(data) == 0 {
+			return 0, nil, nil
+		}
+		if i := strings.Index(string(data), "\n"); i >= 0 {
+			return i + 1, data[0:i], nil
+		}
+		if atEOF {
+			return len(data), data, nil
+		}
+		return 0, nil, nil
+	})
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			if len(data) < 6 { // ignore blank line or wrong format
+				continue
+			}
+			dataChan <- data
+			data = data[6:]
+			if !strings.HasPrefix(data, "[DONE]") {
+				switch relayMode {
+				case RelayModeChatCompletions:
+					var streamResponse ChatCompletionsStreamResponse
+					err := json.Unmarshal([]byte(data), &streamResponse)
+					if err != nil {
+						common.SysError("error unmarshalling stream response: " + err.Error())
+						return
+					}
+					for _, choice := range streamResponse.Choices {
+						responseText += choice.Delta.Content
+					}
+				case RelayModeCompletions:
+					var streamResponse CompletionsStreamResponse
+					err := json.Unmarshal([]byte(data), &streamResponse)
+					if err != nil {
+						common.SysError("error unmarshalling stream response: " + err.Error())
+						return
+					}
+					for _, choice := range streamResponse.Choices {
+						responseText += choice.Text
+					}
+				}
+			}
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			if strings.HasPrefix(data, "data: [DONE]") {
+				data = data[:12]
+			}
+			// some implementations may add \r at the end of data
+			data = strings.TrimSuffix(data, "\r")
+			c.Render(-1, common.CustomEvent{Data: data})
+			return true
+		case <-stopChan:
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func openaiHandler(c *gin.Context, resp *http.Response, consumeQuota bool) (*OpenAIErrorWithStatusCode, *Usage) {
+	var textResponse TextResponse
+	if consumeQuota {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+		}
+		err = json.Unmarshal(responseBody, &textResponse)
+		if err != nil {
+			return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+		}
+		if textResponse.Error.Type != "" {
+			return &OpenAIErrorWithStatusCode{
+				OpenAIError: textResponse.Error,
+				StatusCode:  resp.StatusCode,
+			}, nil
+		}
+		// Reset response body
+		resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	}
+	// We shouldn't set the header before we parse the response body, because the parse part may fail.
+	// And then we will have to send an error response, but in this case, the header has already been set.
+	// So the client will be confused by the response.
+	// For example, Postman will report error, and we cannot check the response at all.
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err := io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, &textResponse.Usage
+}

controller/relay-palm.go

@@ -1,10 +1,17 @@
 package controller
 
 import (
+	"encoding/json"
 	"fmt"
 	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
 )
 
+// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#request-body
+// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#response-body
+
 type PaLMChatMessage struct {
 	Author  string `json:"author"`
 	Content string `json:"content"`
@@ -15,45 +22,188 @@ type PaLMFilter struct {
 	Message string `json:"message"`
 }
 
-// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#request-body
+type PaLMPrompt struct {
+	Messages []PaLMChatMessage `json:"messages"`
+}
+
 type PaLMChatRequest struct {
-	Prompt         []Message `json:"prompt"`
-	Temperature    float64   `json:"temperature"`
-	CandidateCount int       `json:"candidateCount"`
-	TopP           float64   `json:"topP"`
-	TopK           int       `json:"topK"`
+	Prompt         PaLMPrompt `json:"prompt"`
+	Temperature    float64    `json:"temperature,omitempty"`
+	CandidateCount int        `json:"candidateCount,omitempty"`
+	TopP           float64    `json:"topP,omitempty"`
+	TopK           int        `json:"topK,omitempty"`
+}
+
+type PaLMError struct {
+	Code    int    `json:"code"`
+	Message string `json:"message"`
+	Status  string `json:"status"`
 }
 
-// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#response-body
 type PaLMChatResponse struct {
-	Candidates []Message    `json:"candidates"`
-	Messages   []Message    `json:"messages"`
-	Filters    []PaLMFilter `json:"filters"`
+	Candidates []PaLMChatMessage `json:"candidates"`
+	Messages   []Message         `json:"messages"`
+	Filters    []PaLMFilter      `json:"filters"`
+	Error      PaLMError         `json:"error"`
 }
 
-func relayPaLM(openAIRequest GeneralOpenAIRequest, c *gin.Context) *OpenAIErrorWithStatusCode {
-	// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage
-	messages := make([]PaLMChatMessage, 0, len(openAIRequest.Messages))
-	for _, message := range openAIRequest.Messages {
-		var author string
-		if message.Role == "user" {
-			author = "0"
-		} else {
-			author = "1"
-		}
-		messages = append(messages, PaLMChatMessage{
-			Author:  author,
+func requestOpenAI2PaLM(textRequest GeneralOpenAIRequest) *PaLMChatRequest {
+	palmRequest := PaLMChatRequest{
+		Prompt: PaLMPrompt{
+			Messages: make([]PaLMChatMessage, 0, len(textRequest.Messages)),
+		},
+		Temperature:    textRequest.Temperature,
+		CandidateCount: textRequest.N,
+		TopP:           textRequest.TopP,
+		TopK:           textRequest.MaxTokens,
+	}
+	for _, message := range textRequest.Messages {
+		palmMessage := PaLMChatMessage{
 			Content: message.Content,
-		})
+		}
+		if message.Role == "user" {
+			palmMessage.Author = "0"
+		} else {
+			palmMessage.Author = "1"
+		}
+		palmRequest.Prompt.Messages = append(palmRequest.Prompt.Messages, palmMessage)
 	}
-	request := PaLMChatRequest{
-		Prompt:         nil,
-		Temperature:    openAIRequest.Temperature,
-		CandidateCount: openAIRequest.N,
-		TopP:           openAIRequest.TopP,
-		TopK:           openAIRequest.MaxTokens,
-	}
-	// TODO: forward request to PaLM & convert response
-	fmt.Print(request)
-	return nil
+	return &palmRequest
+}
+
+func responsePaLM2OpenAI(response *PaLMChatResponse) *OpenAITextResponse {
+	fullTextResponse := OpenAITextResponse{
+		Choices: make([]OpenAITextResponseChoice, 0, len(response.Candidates)),
+	}
+	for i, candidate := range response.Candidates {
+		choice := OpenAITextResponseChoice{
+			Index: i,
+			Message: Message{
+				Role:    "assistant",
+				Content: candidate.Content,
+			},
+			FinishReason: "stop",
+		}
+		fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
+	}
+	return &fullTextResponse
+}
+
+func streamResponsePaLM2OpenAI(palmResponse *PaLMChatResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	if len(palmResponse.Candidates) > 0 {
+		choice.Delta.Content = palmResponse.Candidates[0].Content
+	}
+	choice.FinishReason = "stop"
+	var response ChatCompletionsStreamResponse
+	response.Object = "chat.completion.chunk"
+	response.Model = "palm2"
+	response.Choices = []ChatCompletionsStreamResponseChoice{choice}
+	return &response
+}
+
+func palmStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createdTime := common.GetTimestamp()
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			common.SysError("error reading stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			common.SysError("error closing stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		var palmResponse PaLMChatResponse
+		err = json.Unmarshal(responseBody, &palmResponse)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		fullTextResponse := streamResponsePaLM2OpenAI(&palmResponse)
+		fullTextResponse.Id = responseId
+		fullTextResponse.Created = createdTime
+		if len(palmResponse.Candidates) > 0 {
+			responseText = palmResponse.Candidates[0].Content
+		}
+		jsonResponse, err := json.Marshal(fullTextResponse)
+		if err != nil {
+			common.SysError("error marshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		dataChan <- string(jsonResponse)
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			c.Render(-1, common.CustomEvent{Data: "data: " + data})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func palmHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var palmResponse PaLMChatResponse
+	err = json.Unmarshal(responseBody, &palmResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if palmResponse.Error.Code != 0 || len(palmResponse.Candidates) == 0 {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: palmResponse.Error.Message,
+				Type:    palmResponse.Error.Status,
+				Param:   "",
+				Code:    palmResponse.Error.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responsePaLM2OpenAI(&palmResponse)
+	completionTokens := countTokenText(palmResponse.Candidates[0].Content, model)
+	usage := Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: completionTokens,
+		TotalTokens:      promptTokens + completionTokens,
+	}
+	fullTextResponse.Usage = usage
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &usage
 }

controller/relay-text.go

@@ -1,7 +1,6 @@
 package controller
 
 import (
-	"bufio"
 	"bytes"
 	"encoding/json"
 	"errors"
@@ -15,6 +14,13 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
+const (
+	APITypeOpenAI = iota
+	APITypeClaude
+	APITypePaLM
+	APITypeBaidu
+)
+
 func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	channelType := c.GetInt("channel")
 	tokenId := c.GetInt("token_id")
@@ -71,33 +77,63 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			isModelMapped = true
 		}
 	}
+	apiType := APITypeOpenAI
+	if strings.HasPrefix(textRequest.Model, "claude") {
+		apiType = APITypeClaude
+	} else if strings.HasPrefix(textRequest.Model, "ERNIE") {
+		apiType = APITypeBaidu
+	} else if strings.HasPrefix(textRequest.Model, "PaLM") {
+		apiType = APITypePaLM
+	}
 	baseURL := common.ChannelBaseURLs[channelType]
 	requestURL := c.Request.URL.String()
 	if c.GetString("base_url") != "" {
 		baseURL = c.GetString("base_url")
 	}
 	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
-	if channelType == common.ChannelTypeAzure {
-		// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
-		query := c.Request.URL.Query()
-		apiVersion := query.Get("api-version")
-		if apiVersion == "" {
-			apiVersion = c.GetString("api_version")
+	switch apiType {
+	case APITypeOpenAI:
+		if channelType == common.ChannelTypeAzure {
+			// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
+			query := c.Request.URL.Query()
+			apiVersion := query.Get("api-version")
+			if apiVersion == "" {
+				apiVersion = c.GetString("api_version")
+			}
+			requestURL := strings.Split(requestURL, "?")[0]
+			requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
+			baseURL = c.GetString("base_url")
+			task := strings.TrimPrefix(requestURL, "/v1/")
+			model_ := textRequest.Model
+			model_ = strings.Replace(model_, ".", "", -1)
+			// https://github.com/songquanpeng/one-api/issues/67
+			model_ = strings.TrimSuffix(model_, "-0301")
+			model_ = strings.TrimSuffix(model_, "-0314")
+			model_ = strings.TrimSuffix(model_, "-0613")
+			fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/%s", baseURL, model_, task)
 		}
-		requestURL := strings.Split(requestURL, "?")[0]
-		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
-		baseURL = c.GetString("base_url")
-		task := strings.TrimPrefix(requestURL, "/v1/")
-		model_ := textRequest.Model
-		model_ = strings.Replace(model_, ".", "", -1)
-		// https://github.com/songquanpeng/one-api/issues/67
-		model_ = strings.TrimSuffix(model_, "-0301")
-		model_ = strings.TrimSuffix(model_, "-0314")
-		model_ = strings.TrimSuffix(model_, "-0613")
-		fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/%s", baseURL, model_, task)
-	} else if channelType == common.ChannelTypePaLM {
-		err := relayPaLM(textRequest, c)
-		return err
+	case APITypeClaude:
+		fullRequestURL = "https://api.anthropic.com/v1/complete"
+		if baseURL != "" {
+			fullRequestURL = fmt.Sprintf("%s/v1/complete", baseURL)
+		}
+	case APITypeBaidu:
+		switch textRequest.Model {
+		case "ERNIE-Bot":
+			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions"
+		case "ERNIE-Bot-turbo":
+			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant"
+		case "BLOOMZ-7B":
+			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/bloomz_7b1"
+		}
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		fullRequestURL += "?access_token=" + apiKey // TODO: access token expire in 30 days
+	case APITypePaLM:
+		fullRequestURL = "https://generativelanguage.googleapis.com/v1beta2/models/chat-bison-001:generateMessage"
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		fullRequestURL += "?key=" + apiKey
 	}
 	var promptTokens int
 	var completionTokens int
@@ -142,16 +178,49 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	} else {
 		requestBody = c.Request.Body
 	}
+	switch apiType {
+	case APITypeClaude:
+		claudeRequest := requestOpenAI2Claude(textRequest)
+		jsonStr, err := json.Marshal(claudeRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeBaidu:
+		baiduRequest := requestOpenAI2Baidu(textRequest)
+		jsonStr, err := json.Marshal(baiduRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypePaLM:
+		palmRequest := requestOpenAI2PaLM(textRequest)
+		jsonStr, err := json.Marshal(palmRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	}
 	req, err := http.NewRequest(c.Request.Method, fullRequestURL, requestBody)
 	if err != nil {
 		return errorWrapper(err, "new_request_failed", http.StatusInternalServerError)
 	}
-	if channelType == common.ChannelTypeAzure {
-		key := c.Request.Header.Get("Authorization")
-		key = strings.TrimPrefix(key, "Bearer ")
-		req.Header.Set("api-key", key)
-	} else {
-		req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+	apiKey := c.Request.Header.Get("Authorization")
+	apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+	switch apiType {
+	case APITypeOpenAI:
+		if channelType == common.ChannelTypeAzure {
+			req.Header.Set("api-key", apiKey)
+		} else {
+			req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+		}
+	case APITypeClaude:
+		req.Header.Set("x-api-key", apiKey)
+		anthropicVersion := c.Request.Header.Get("anthropic-version")
+		if anthropicVersion == "" {
+			anthropicVersion = "2023-06-01"
+		}
+		req.Header.Set("anthropic-version", anthropicVersion)
 	}
 	req.Header.Set("Content-Type", c.Request.Header.Get("Content-Type"))
 	req.Header.Set("Accept", c.Request.Header.Get("Accept"))
@@ -183,7 +252,7 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			if strings.HasPrefix(textRequest.Model, "gpt-4") {
 				completionRatio = 2
 			}
-			if isStream {
+			if isStream && apiType != APITypeBaidu {
 				completionTokens = countTokenText(streamResponseText, textRequest.Model)
 			} else {
 				promptTokens = textResponse.Usage.PromptTokens
@@ -219,124 +288,72 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			}
 		}
 	}()
-
-	if isStream {
-		scanner := bufio.NewScanner(resp.Body)
-		scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
-			if atEOF && len(data) == 0 {
-				return 0, nil, nil
-			}
-
-			if i := strings.Index(string(data), "\n"); i >= 0 {
-				return i + 1, data[0:i], nil
-			}
-
-			if atEOF {
-				return len(data), data, nil
-			}
-
-			return 0, nil, nil
-		})
-		dataChan := make(chan string)
-		stopChan := make(chan bool)
-		go func() {
-			for scanner.Scan() {
-				data := scanner.Text()
-				if len(data) < 6 { // ignore blank line or wrong format
-					continue
-				}
-				dataChan <- data
-				data = data[6:]
-				if !strings.HasPrefix(data, "[DONE]") {
-					switch relayMode {
-					case RelayModeChatCompletions:
-						var streamResponse ChatCompletionsStreamResponse
-						err = json.Unmarshal([]byte(data), &streamResponse)
-						if err != nil {
-							common.SysError("error unmarshalling stream response: " + err.Error())
-							return
-						}
-						for _, choice := range streamResponse.Choices {
-							streamResponseText += choice.Delta.Content
-						}
-					case RelayModeCompletions:
-						var streamResponse CompletionsStreamResponse
-						err = json.Unmarshal([]byte(data), &streamResponse)
-						if err != nil {
-							common.SysError("error unmarshalling stream response: " + err.Error())
-							return
-						}
-						for _, choice := range streamResponse.Choices {
-							streamResponseText += choice.Text
-						}
-					}
-				}
-			}
-			stopChan <- true
-		}()
-		c.Writer.Header().Set("Content-Type", "text/event-stream")
-		c.Writer.Header().Set("Cache-Control", "no-cache")
-		c.Writer.Header().Set("Connection", "keep-alive")
-		c.Writer.Header().Set("Transfer-Encoding", "chunked")
-		c.Writer.Header().Set("X-Accel-Buffering", "no")
-		c.Stream(func(w io.Writer) bool {
-			select {
-			case data := <-dataChan:
-				if strings.HasPrefix(data, "data: [DONE]") {
-					data = data[:12]
-				}
-				// some implementations may add \r at the end of data
-				data = strings.TrimSuffix(data, "\r")
-				c.Render(-1, common.CustomEvent{Data: data})
-				return true
-			case <-stopChan:
-				return false
-			}
-		})
-		err = resp.Body.Close()
-		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
-		}
-		return nil
-	} else {
-		if consumeQuota {
-			responseBody, err := io.ReadAll(resp.Body)
+	switch apiType {
+	case APITypeOpenAI:
+		if isStream {
+			err, responseText := openaiStreamHandler(c, resp, relayMode)
 			if err != nil {
-				return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			err = resp.Body.Close()
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := openaiHandler(c, resp, consumeQuota)
 			if err != nil {
-				return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			err = json.Unmarshal(responseBody, &textResponse)
+			textResponse.Usage = *usage
+			return nil
+		}
+	case APITypeClaude:
+		if isStream {
+			err, responseText := claudeStreamHandler(c, resp)
 			if err != nil {
-				return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			if textResponse.Error.Type != "" {
-				return &OpenAIErrorWithStatusCode{
-					OpenAIError: textResponse.Error,
-					StatusCode:  resp.StatusCode,
-				}
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := claudeHandler(c, resp, promptTokens, textRequest.Model)
+			if err != nil {
+				return err
 			}
-			// Reset response body
-			resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+			textResponse.Usage = *usage
+			return nil
 		}
-		// We shouldn't set the header before we parse the response body, because the parse part may fail.
-		// And then we will have to send an error response, but in this case, the header has already been set.
-		// So the client will be confused by the response.
-		// For example, Postman will report error, and we cannot check the response at all.
-		for k, v := range resp.Header {
-			c.Writer.Header().Set(k, v[0])
+	case APITypeBaidu:
+		if isStream {
+			err, usage := baiduStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			textResponse.Usage = *usage
+			return nil
+		} else {
+			err, usage := baiduHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			textResponse.Usage = *usage
+			return nil
 		}
-		c.Writer.WriteHeader(resp.StatusCode)
-		_, err = io.Copy(c.Writer, resp.Body)
-		if err != nil {
-			return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError)
+	case APITypePaLM:
+		if textRequest.Stream { // PaLM2 API does not support stream
+			err, responseText := palmStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := palmHandler(c, resp, promptTokens, textRequest.Model)
+			if err != nil {
+				return err
+			}
+			textResponse.Usage = *usage
+			return nil
 		}
-		err = resp.Body.Close()
-		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
-		}
-		return nil
+	default:
+		return errorWrapper(errors.New("unknown api type"), "unknown_api_type", http.StatusInternalServerError)
 	}
 }

controller/relay-utils.go

@@ -91,3 +91,16 @@ func errorWrapper(err error, code string, statusCode int) *OpenAIErrorWithStatus
 		StatusCode:  statusCode,
 	}
 }
+
+func shouldDisableChannel(err *OpenAIError) bool {
+	if !common.AutomaticDisableChannelEnabled {
+		return false
+	}
+	if err == nil {
+		return false
+	}
+	if err.Type == "insufficient_quota" || err.Code == "invalid_api_key" || err.Code == "account_deactivated" {
+		return true
+	}
+	return false
+}

controller/relay.go

@@ -85,6 +85,20 @@ type TextResponse struct {
 	Error OpenAIError `json:"error"`
 }
 
+type OpenAITextResponseChoice struct {
+	Index        int `json:"index"`
+	Message      `json:"message"`
+	FinishReason string `json:"finish_reason"`
+}
+
+type OpenAITextResponse struct {
+	Id      string                     `json:"id"`
+	Object  string                     `json:"object"`
+	Created int64                      `json:"created"`
+	Choices []OpenAITextResponseChoice `json:"choices"`
+	Usage   `json:"usage"`
+}
+
 type ImageResponse struct {
 	Created int `json:"created"`
 	Data    []struct {
@@ -92,13 +106,19 @@ type ImageResponse struct {
 	}
 }
 
+type ChatCompletionsStreamResponseChoice struct {
+	Delta struct {
+		Content string `json:"content"`
+	} `json:"delta"`
+	FinishReason string `json:"finish_reason,omitempty"`
+}
+
 type ChatCompletionsStreamResponse struct {
-	Choices []struct {
-		Delta struct {
-			Content string `json:"content"`
-		} `json:"delta"`
-		FinishReason string `json:"finish_reason"`
-	} `json:"choices"`
+	Id      string                                `json:"id"`
+	Object  string                                `json:"object"`
+	Created int64                                 `json:"created"`
+	Model   string                                `json:"model"`
+	Choices []ChatCompletionsStreamResponseChoice `json:"choices"`
 }
 
 type CompletionsStreamResponse struct {
@@ -151,7 +171,7 @@ func Relay(c *gin.Context) {
 		channelId := c.GetInt("channel_id")
 		common.SysError(fmt.Sprintf("relay error (channel #%d): %s", channelId, err.Message))
 		// https://platform.openai.com/docs/guides/error-codes/api-errors
-		if common.AutomaticDisableChannelEnabled && (err.Type == "insufficient_quota" || err.Code == "invalid_api_key" || err.Code == "account_deactivated") {
+		if shouldDisableChannel(&err.OpenAIError) {
 			channelId := c.GetInt("channel_id")
 			channelName := c.GetString("channel_name")
 			disableChannel(channelId, channelName, err.Message)

web/src/constants/channel.constants.js

@@ -1,7 +1,10 @@
 export const CHANNEL_OPTIONS = [
   { key: 1, text: 'OpenAI', value: 1, color: 'green' },
+  { key: 14, text: 'Anthropic', value: 14, color: 'black' },
   { key: 8, text: '自定义', value: 8, color: 'pink' },
   { key: 3, text: 'Azure', value: 3, color: 'olive' },
+  { key: 11, text: 'PaLM', value: 11, color: 'orange' },
+  { key: 15, text: 'Baidu', value: 15, color: 'blue' },
   { key: 2, text: 'API2D', value: 2, color: 'blue' },
   { key: 4, text: 'CloseAI', value: 4, color: 'teal' },
   { key: 5, text: 'OpenAI-SB', value: 5, color: 'brown' },

web/src/pages/Channel/EditChannel.js

@@ -27,6 +27,7 @@ const EditChannel = () => {
   };
   const [batch, setBatch] = useState(false);
   const [inputs, setInputs] = useState(originInputs);
+  const [originModelOptions, setOriginModelOptions] = useState([]);
   const [modelOptions, setModelOptions] = useState([]);
   const [groupOptions, setGroupOptions] = useState([]);
   const [basicModels, setBasicModels] = useState([]);
@@ -44,19 +45,6 @@ const EditChannel = () => {
         data.models = [];
       } else {
         data.models = data.models.split(',');
-        setTimeout(() => {
-          let localModelOptions = [...modelOptions];
-          data.models.forEach((model) => {
-            if (!localModelOptions.find((option) => option.key === model)) {
-              localModelOptions.push({
-                key: model,
-                text: model,
-                value: model
-              });
-            }
-          });
-          setModelOptions(localModelOptions);
-        }, 1000);
       }
       if (data.group === '') {
         data.groups = [];
@@ -76,13 +64,16 @@ const EditChannel = () => {
   const fetchModels = async () => {
     try {
       let res = await API.get(`/api/channel/models`);
-      setModelOptions(res.data.data.map((model) => ({
+      let localModelOptions = res.data.data.map((model) => ({
         key: model.id,
         text: model.id,
         value: model.id
-      })));
+      }));
+      setOriginModelOptions(localModelOptions);
       setFullModels(res.data.data.map((model) => model.id));
-      setBasicModels(res.data.data.filter((model) => !model.id.startsWith('gpt-4')).map((model) => model.id));
+      setBasicModels(res.data.data.filter((model) => {
+        return model.id.startsWith('gpt-3') || model.id.startsWith('text-');
+      }).map((model) => model.id));
     } catch (error) {
       showError(error.message);
     }
@@ -101,6 +92,20 @@ const EditChannel = () => {
     }
   };
 
+  useEffect(() => {
+    let localModelOptions = [...originModelOptions];
+    inputs.models.forEach((model) => {
+      if (!localModelOptions.find((option) => option.key === model)) {
+        localModelOptions.push({
+          key: model,
+          text: model,
+          value: model
+        });
+      }
+    });
+    setModelOptions(localModelOptions);
+  }, [originModelOptions, inputs.models]);
+
   useEffect(() => {
     if (isEdit) {
       loadChannel().then();
@@ -280,15 +285,20 @@ const EditChannel = () => {
             <Input
               action={
                 <Button type={'button'} onClick={()=>{
+                  if (customModel.trim() === "") return;
+                  if (inputs.models.includes(customModel)) return;
                   let localModels = [...inputs.models];
                   localModels.push(customModel);
-                  let localModelOptions = [...modelOptions];
+                  let localModelOptions = [];
                   localModelOptions.push({
                     key: customModel,
                     text: customModel,
                     value: customModel,
                   });
-                  setModelOptions(localModelOptions);
+                  setModelOptions(modelOptions=>{
+                    return [...modelOptions, ...localModelOptions];
+                  });
+                  setCustomModel('');
                   handleInputChange(null, { name: 'models', value: localModels });
                 }}>填入</Button>
               }
@@ -327,7 +337,7 @@ const EditChannel = () => {
                 label='密钥'
                 name='key'
                 required
-                placeholder={'请输入密钥'}
+                placeholder={inputs.type === 15 ? "请输入 access token，当前版本暂不支持自动刷新，请每 30 天更新一次" : '请输入密钥'}
                 onChange={handleInputChange}
                 value={inputs.key}
                 autoComplete='new-password'