feat(chat): 重构 LLM 流式输出并扩展 ChatSaveReq 字段

- 将原始整段 chunk 拆分为 3 字批次推送，降低前端卡顿
- ChatSaveReq 新增 userId、lang、liked 等 8 个字段并补充 Swagger 注解
- QdrantVectorService 改用 Map<String,JsonWithInt.Value> 载荷，新增 QdrantPayloadMapper 统一转换

src/main/java/com/yolo/keyborad/controller/ChatController.java

@@ -4,28 +4,32 @@ import cn.dev33.satoken.stp.StpUtil;
 import cn.hutool.core.util.IdUtil;
 import com.yolo.keyborad.common.BaseResponse;
 import com.yolo.keyborad.common.ResultUtils;
+import com.yolo.keyborad.mapper.QdrantPayloadMapper;
 import com.yolo.keyborad.model.dto.chat.ChatReq;
 import com.yolo.keyborad.model.dto.chat.ChatSaveReq;
 import com.yolo.keyborad.model.dto.chat.ChatStreamMessage;
 import com.yolo.keyborad.model.entity.KeyboardCharacter;
 import com.yolo.keyborad.service.KeyboardCharacterService;
 import com.yolo.keyborad.service.impl.QdrantVectorService;
+import io.qdrant.client.grpc.JsonWithInt;
 import io.swagger.v3.oas.annotations.Operation;
 import io.swagger.v3.oas.annotations.Parameter;
 import io.swagger.v3.oas.annotations.tags.Tag;
 import jakarta.annotation.Resource;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.ai.chat.client.ChatClient;
-import org.springframework.ai.embedding.EmbeddingResponse;
 import org.springframework.ai.openai.OpenAiChatOptions;
 import org.springframework.ai.openai.OpenAiEmbeddingModel;
-import org.springframework.boot.context.properties.bind.DefaultValue;
 import org.springframework.http.codec.ServerSentEvent;
 import org.springframework.web.bind.annotation.*;
 import reactor.core.publisher.Flux;
 import reactor.core.publisher.Mono;
 import reactor.core.scheduler.Schedulers;
 
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
 /*
  * @author: ziin
  * @date: 2025/12/8 15:05
@@ -52,7 +56,7 @@ public class ChatController {
 
     @PostMapping("/talk")
     @Operation(summary = "聊天润色接口", description = "聊天润色接口")
-    public Flux<ServerSentEvent<ChatStreamMessage>> testTalk(@RequestBody ChatReq chatReq){
+    public Flux<ServerSentEvent<ChatStreamMessage>> talk(@RequestBody ChatReq chatReq){
         KeyboardCharacter character = keyboardCharacterService.getById(chatReq.getCharacterId());
         // 1. LLM 流式输出
         Flux<ChatStreamMessage> llmFlux = client
@@ -69,7 +73,30 @@ public class ChatController {
                         .build())
                 .stream()
                 .content()
-                .map(chunk -> new ChatStreamMessage("llm_chunk", chunk));
+                .concatMap(chunk -> {
+                    // 拆成单字符
+                    List<String> chars = chunk.codePoints()
+                            .mapToObj(cp -> new String(Character.toChars(cp)))
+                            .toList();
+
+                    // 你可以在这里按 3～5 个字符再拼一拼
+                    List<String> batched = new ArrayList<>();
+                    StringBuilder sb = new StringBuilder();
+                    for (String ch : chars) {
+                        sb.append(ch);
+                        if (sb.length() >= 3) { // 这里的 3 可以自己调
+                            batched.add(sb.toString());
+                            sb.setLength(0);
+                        }
+                    }
+                    if (!sb.isEmpty()) {
+                        batched.add(sb.toString());
+                    }
+
+                    return Flux.fromIterable(batched)
+                            .map(s -> new ChatStreamMessage("llm_chunk", s));
+                });
+//                .map(chunk -> new ChatStreamMessage("llm_chunk", chunk));
 
         // 2. 向量搜索Flux（一次性发送搜索结果）
         Flux<ChatStreamMessage> searchFlux = Mono
@@ -99,8 +126,10 @@ public class ChatController {
     @Operation(summary = "保存润色后的句子", description = "保存润色后的句子")
     @Parameter(name = "userInput",required = true,description = "测试聊天接口",example = "talk to something")
     public BaseResponse<Boolean> testTalkWithVector(@RequestBody ChatSaveReq chatSaveReq) {
-        float[] embed = embeddingModel.embed(chatSaveReq.getUserInputMessage());
-        qdrantVectorService.upsertPoint(IdUtil.getSnowflakeNextId(), embed, chatSaveReq.getUserSelectMessage());
+        float[] embed = embeddingModel.embed(chatSaveReq.getUserText());
+        chatSaveReq.setUserId(StpUtil.getLoginIdAsLong());
+        Map<String, JsonWithInt.Value> map = QdrantPayloadMapper.toQdrantPayload(chatSaveReq);
+        qdrantVectorService.upsertPoint(IdUtil.getSnowflakeNextId(), embed, map);
         return ResultUtils.success(true);
     }
 }