The 'stop generate' and 'regenerate response' function is ready

server/chat_handler.go

@@ -3,6 +3,7 @@ package server
 import (
 	"bufio"
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -59,12 +60,13 @@ func (s *Server) ChatHandle(c *gin.Context) {
 				delete(s.ChatClients, sessionId)
 				return
 			}
-
 			logger.Info("Receive a message: ", string(message))
 			//replyMessage(client, "当前 TOKEN 无效，请使用合法的 TOKEN 登录！", false)
 			//replyMessage(client, "![](images/wx.png)", true)
-			// TODO: 当前只保持当前会话的上下文，部保存用户的所有的聊天历史记录，后期要考虑保存所有的历史记录
-			err = s.sendMessage(session, chatRole, string(message), client, false)
+			ctx, cancel := context.WithCancel(context.Background())
+			s.ReqCancelFunc[sessionId] = cancel
+			// 回复消息
+			err = s.sendMessage(ctx, session, chatRole, string(message), client, false)
 			if err != nil {
 				logger.Error(err)
 			}
@@ -73,7 +75,13 @@ func (s *Server) ChatHandle(c *gin.Context) {
 }
 
 // 将消息发送给 ChatGPT 并获取结果，通过 WebSocket 推送到客户端
-func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, prompt string, ws Client, resetContext bool) error {
+func (s *Server) sendMessage(ctx context.Context, session types.ChatSession, role types.ChatRole, prompt string, ws Client, resetContext bool) error {
+	cancel := s.ReqCancelFunc[session.SessionId]
+	defer func() {
+		cancel()
+		delete(s.ReqCancelFunc, session.SessionId)
+	}()
+
 	user, err := GetUser(session.Username)
 	if err != nil {
 		replyMessage(ws, "当前 TOKEN 无效，请使用合法的 TOKEN 登录！", false)
@@ -98,38 +106,38 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 		replyMessage(ws, "![](images/start.png)", true)
 		return nil
 	}
-	var r = types.ApiRequest{
+	var req = types.ApiRequest{
 		Model:       s.Config.Chat.Model,
 		Temperature: s.Config.Chat.Temperature,
 		MaxTokens:   s.Config.Chat.MaxTokens,
 		Stream:      true,
 	}
-	var context []types.Message
+	var chatCtx []types.Message
 	var ctxKey = fmt.Sprintf("%s-%s", session.SessionId, role.Key)
 	if v, ok := s.ChatContexts[ctxKey]; ok && s.Config.Chat.EnableContext {
-		context = v.Messages
+		chatCtx = v.Messages
 	} else {
-		context = role.Context
+		chatCtx = role.Context
 	}
 
 	if s.DebugMode {
-		logger.Infof("会话上下文：%+v", context)
+		logger.Infof("会话上下文：%+v", chatCtx)
 	}
 
+	req.Messages = append(chatCtx, types.Message{
+		Role:    "user",
+		Content: prompt,
+	})
+
 	// 创建 HttpClient 请求对象
 	var client *http.Client
-	var retryCount = 5
+	var retryCount = 5 // 重试次数
 	var response *http.Response
 	var apiKey string
 	var failedKey = ""
 	var failedProxyURL = ""
 	for retryCount > 0 {
-		r.Messages = append(context, types.Message{
-			Role:    "user",
-			Content: prompt,
-		})
-
-		requestBody, err := json.Marshal(r)
+		requestBody, err := json.Marshal(req)
 		if err != nil {
 			return err
 		}
@@ -139,6 +147,7 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 			return err
 		}
 
+		request = request.WithContext(ctx)
 		request.Header.Add("Content-Type", "application/json")
 
 		proxyURL := s.getProxyURL(failedProxyURL)
@@ -164,7 +173,10 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 		response, err = client.Do(request)
 		if err == nil {
 			break
+		} else if strings.Contains(err.Error(), "context canceled") {
+			return errors.New("用户取消了请求：" + prompt)
 		} else {
+			logger.Error("HTTP API 请求失败：" + err.Error())
 			failedKey = apiKey
 			failedProxyURL = proxyURL
 		}
@@ -205,14 +217,14 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 			_ = utils.SaveConfig(s.Config, s.ConfigPath)
 
 			// 重发当前消息
-			return s.sendMessage(session, role, prompt, ws, false)
+			return s.sendMessage(ctx, session, role, prompt, ws, false)
 
 			// 上下文超出长度了
 		} else if strings.Contains(line, "This model's maximum context length is 4097 tokens") {
 			logger.Infof("会话上下文长度超出限制, Username: %s", user.Name)
 			// 重置上下文，重发当前消息
 			delete(s.ChatContexts, ctxKey)
-			return s.sendMessage(session, role, prompt, ws, true)
+			return s.sendMessage(ctx, session, role, prompt, ws, true)
 		} else if !strings.Contains(line, "data:") {
 			continue
 		}
@@ -246,7 +258,18 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 				IsHelloMsg: false,
 			})
 		}
-	}
+
+		// 监控取消信号
+		select {
+		case <-ctx.Done():
+			// 结束输出
+			replyChunkMessage(ws, types.WsMessage{Type: types.WsEnd, IsHelloMsg: false})
+			_ = response.Body.Close()
+			return errors.New("用户取消了请求：" + prompt)
+		default:
+			continue
+		}
+	} // end for
 	_ = response.Body.Close() // 关闭资源
 
 	// 消息发送成功
@@ -260,27 +283,26 @@ func (s *Server) sendMessage(session types.ChatSession, role types.ChatRole, pro
 		if message.Role == "" {
 			message.Role = "assistant"
 		}
-		// 追加上下文消息
-		useMsg := types.Message{Role: "user", Content: prompt}
-		context = append(context, useMsg)
 		message.Content = strings.Join(contents, "")
+		useMsg := types.Message{Role: "user", Content: prompt}
 
 		// 更新上下文消息
 		if s.Config.Chat.EnableContext {
-			context = append(context, message)
+			chatCtx = append(chatCtx, useMsg)  // 提问消息
+			chatCtx = append(chatCtx, message) // 回复消息
 			s.ChatContexts[ctxKey] = types.ChatContext{
-				Messages:       context,
+				Messages:       chatCtx,
 				LastAccessTime: time.Now().Unix(),
 			}
 		}
 
 		// 追加历史消息
 		if user.EnableHistory {
-			err = AppendChatHistory(user.Name, role.Key, useMsg)
+			err = AppendChatHistory(user.Name, role.Key, useMsg) // 提问消息
 			if err != nil {
 				return err
 			}
-			err = AppendChatHistory(user.Name, role.Key, message)
+			err = AppendChatHistory(user.Name, role.Key, message) // 回复消息
 		}
 	}
 
@@ -431,3 +453,12 @@ func (s *Server) ClearHistoryHandle(c *gin.Context) {
 
 	c.JSON(http.StatusOK, types.BizVo{Code: types.Success})
 }
+
+// StopGenerateHandle 停止生成
+func (s *Server) StopGenerateHandle(c *gin.Context) {
+	sessionId := c.GetHeader(types.TokenName)
+	cancel := s.ReqCancelFunc[sessionId]
+	cancel()
+	delete(s.ReqCancelFunc, sessionId)
+	c.JSON(http.StatusOK, types.BizVo{Code: types.Success})
+}