feat(ai-nutritionist): Coze TTS and streaming robustness

- Add Coze TTS endpoint and service; expose binary MP3 from controller.
- Bypass ResponseFilter for /audio/speech so MP3 bodies are not UTF-8 wrapped.
- UniApp: cozeTextToSpeech, TTS UI and play flow; SSE HTTP errors and diagnostics.
- Document TTS in docs/features.md; extend test-0325-1 with curl verification.

Made-with: Cursor

msh_crmeb_22/crmeb-front/src/main/java/com/zbkj/front/controller/CozeController.java

@@ -22,6 +22,7 @@ import org.springframework.web.servlet.mvc.method.annotation.SseEmitter;
 import javax.servlet.http.HttpServletResponse;
 import java.io.File;
 import java.io.IOException;
+import java.util.Map;
 
 /**
  * Coze API 控制器
@@ -132,4 +133,23 @@ public class CozeController {
     public CozeBaseResponse<Object> workflowResume(@RequestBody CozeWorkflowResumeRequest request) {
         return toolCozeService.workflowResume(request);
     }
+
+    /**
+     * 文本转语音 (TTS)
+     */
+    @ApiOperation(value = "文本转语音", notes = "调用 Coze TTS 将文本合成为 MP3 音频并直接返回二进制流")
+    @PostMapping("/audio/speech")
+    public void textToSpeech(@RequestBody Map<String, Object> params, HttpServletResponse response) throws IOException {
+        String input = (String) params.get("input");
+        String voiceId = (String) params.get("voiceId");
+        String format = params.get("format") != null ? (String) params.get("format") : "mp3";
+        Float speed = params.get("speed") != null ? ((Number) params.get("speed")).floatValue() : null;
+
+        byte[] audioData = toolCozeService.textToSpeech(input, voiceId, format, speed);
+        response.setContentType("audio/mpeg");
+        response.setHeader("Content-Disposition", "inline; filename=speech.mp3");
+        response.setContentLength(audioData.length);
+        response.getOutputStream().write(audioData);
+        response.getOutputStream().flush();
+    }
 }

msh_crmeb_22/crmeb-front/src/main/java/com/zbkj/front/filter/ResponseFilter.java

@@ -27,6 +27,17 @@ public class ResponseFilter implements Filter {
     @Override
     public void doFilter(ServletRequest request, ServletResponse response, FilterChain filterChain)
             throws IOException, ServletException {
+        HttpServletRequest httpRequest = (HttpServletRequest) request;
+        String uri = httpRequest.getRequestURI();
+        String accept = httpRequest.getHeader("Accept");
+        // SSE 流式响应和二进制音频响应不能被缓冲，直接透传
+        boolean isSseStream = uri != null && uri.contains("/stream");
+        boolean acceptsSse = accept != null && accept.contains("text/event-stream");
+        boolean isAudioResponse = uri != null && uri.contains("/audio/speech");
+        if (isSseStream || acceptsSse || isAudioResponse) {
+            filterChain.doFilter(request, response);
+            return;
+        }
         ResponseWrapper wrapperResponse = new ResponseWrapper((HttpServletResponse) response);//转换成代理类
         // 这里只拦截返回，直接让请求过去，如果在请求前有处理，可以在这里处理
         filterChain.doFilter(request, wrapperResponse);

msh_crmeb_22/crmeb-service/src/main/java/com/zbkj/service/service/impl/tool/ToolCozeServiceImpl.java

@@ -1,5 +1,8 @@
 package com.zbkj.service.service.impl.tool;
 
+import com.coze.openapi.client.audio.common.AudioFormat;
+import com.coze.openapi.client.audio.speech.CreateSpeechReq;
+import com.coze.openapi.client.audio.speech.CreateSpeechResp;
 import com.coze.openapi.client.chat.CreateChatReq;
 import com.coze.openapi.client.chat.CreateChatResp;
 import com.coze.openapi.client.chat.RetrieveChatReq;
@@ -437,6 +440,27 @@ public class ToolCozeServiceImpl implements ToolCozeService {
         }
     }
 
+    private static final String DEFAULT_VOICE_ID = "7468518753626652709";
+
+    @Override
+    public byte[] textToSpeech(String input, String voiceId, String format, Float speed) {
+        try {
+            CozeAPI client = getClient();
+            AudioFormat audioFormat = (format != null) ? AudioFormat.fromString(format) : AudioFormat.MP3;
+            CreateSpeechReq req = CreateSpeechReq.builder()
+                    .input(input)
+                    .voiceID(voiceId != null ? voiceId : DEFAULT_VOICE_ID)
+                    .responseFormat(audioFormat)
+                    .speed(speed != null ? speed : 1.0f)
+                    .build();
+            CreateSpeechResp resp = client.audio().speech().create(req);
+            return resp.getResponse().bytes();
+        } catch (Exception e) {
+            logger.error("Coze TTS error", e);
+            throw new RuntimeException("语音合成失败: " + e.getMessage(), e);
+        }
+    }
+
     /**
      * 获取访问令牌
      */

msh_crmeb_22/crmeb-service/src/main/java/com/zbkj/service/service/tool/ToolCozeService.java

@@ -79,4 +79,15 @@ public interface ToolCozeService {
      * @return 恢复结果
      */
     CozeBaseResponse<Object> workflowResume(CozeWorkflowResumeRequest request);
+
+    /**
+     * 文本转语音 (TTS)
+     *
+     * @param input   要合成的文本
+     * @param voiceId 音色ID，为 null 时使用默认中文音色
+     * @param format  音频格式，如 "mp3"，为 null 时默认 mp3
+     * @param speed   语速，1.0 为正常速度，为 null 时使用默认值
+     * @return 音频二进制数据
+     */
+    byte[] textToSpeech(String input, String voiceId, String format, Float speed);
 }