进一步优化大模型服务的响应内容

Author: MartinDai

robot-web/src/main/java/com/doodl6/wechatrobot/service/assistant/impl/DashscopeService.java

@@ -34,7 +34,7 @@ public BaseMessage processText(String content, String fromUserName) {
 
         try {
             Generation gen = new Generation();
-            Message systemMsg = Message.builder().role(Role.SYSTEM.getValue()).content("你是一个AI助手，保持回复内容尽量简短").build();
+            Message systemMsg = Message.builder().role(Role.SYSTEM.getValue()).content("你是一个AI助手，尽量保证回复内容在200个字符以内").build();
             Message userMsg = Message.builder().role(Role.USER.getValue()).content(content).build();
             GenerationParam param =
                     GenerationParam.builder().model(Generation.Models.QWEN_TURBO).messages(Lists.newArrayList(systemMsg, userMsg))

robot-web/src/main/java/com/doodl6/wechatrobot/service/assistant/impl/OpenAIService.java

@@ -89,7 +89,7 @@ public BaseMessage processText(String content, String fromUserName) {
 
         ChatMessage systemMsg = new ChatMessage();
         systemMsg.setRole("system");
-        systemMsg.setContent("你是一个AI助手，保持回复内容尽量简短");
+        systemMsg.setContent("你是一个AI助手，尽量保证回复内容在200个字符以内");
 
         ChatMessage userMsg = new ChatMessage();
         userMsg.setRole("user");