音频模型调整

src/main/java/com/rj/common/PasswordUtil.java

@@ -180,6 +180,8 @@ public class PasswordUtil {
 
 
 
+
+
 
 
 

src/main/java/com/rj/common/ServiceManager.java

@@ -164,6 +164,8 @@ public class ServiceManager {
 
 
 
+
+
 
 
 

src/main/java/com/rj/config/AliyunConfig.java

@@ -141,6 +141,8 @@ public class AliyunConfig {
 
 
 
+
+
 
 
 

src/main/java/com/rj/controller/SiliconFlowTtsController.java

@@ -419,7 +419,7 @@ public class SiliconFlowTtsController {
             String fileExtension = originalFilename != null && originalFilename.contains(".") 
                 ? originalFilename.substring(originalFilename.lastIndexOf(".")) 
                 : ".mp3";
-            String uniqueFileName = "audio/" + UUID.randomUUID().toString() + fileExtension;
+            String uniqueFileName = "/" + UUID.randomUUID().toString() + fileExtension;
             
             // 上传文件到MinIO
             String minioUrl = minIOService.uploadFileWithName(file, uniqueFileName);

src/main/java/com/rj/controller/TtsRequestLogController.java

@@ -5,6 +5,8 @@ import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
 import com.rj.entity.TtsRequestLog;
 import com.rj.mapper.TtsRequestLogMapper;
 import com.rj.service.ITtsRequestLogService;
+import com.rj.dto.AsrRequest;
+import com.rj.dto.AsrResponse;
 import io.swagger.v3.oas.annotations.Operation;
 import io.swagger.v3.oas.annotations.Parameter;
 import io.swagger.v3.oas.annotations.tags.Tag;
@@ -13,7 +15,6 @@ import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.http.ResponseEntity;
 import org.springframework.web.bind.annotation.*;
 
-import java.time.LocalDateTime;
 import java.util.HashMap;
 import java.util.Map;
 
@@ -40,6 +41,7 @@ public class TtsRequestLogController {
     public ResponseEntity<Map<String, Object>> getTtsRequestLogs(
             @Parameter(description = "页码") @RequestParam(defaultValue = "1") Integer pageNum,
             @Parameter(description = "每页大小") @RequestParam(defaultValue = "10") Integer pageSize,
+            @Parameter(description = "音乐名称") @RequestParam(required = false) String audioName,
             @Parameter(description = "状态") @RequestParam(required = false) String status,
             @Parameter(description = "模型") @RequestParam(required = false) String model,
             @Parameter(description = "创建人姓名") @RequestParam(required = false) String creatorName,
@@ -48,12 +50,15 @@ public class TtsRequestLogController {
             @Parameter(description = "结束时间") @RequestParam(required = false) String endTime) {
         
         Map<String, Object> result = new HashMap<>();
-        
+
         try {
             Page<TtsRequestLog> page = new Page<>(pageNum, pageSize);
             QueryWrapper<TtsRequestLog> queryWrapper = new QueryWrapper<>();
             
             // 添加查询条件
+            if (audioName != null && !audioName.trim().isEmpty()) {
+                queryWrapper.like("audio_name", audioName);
+            }
             if (status != null && !status.trim().isEmpty()) {
                 queryWrapper.eq("status", status);
             }
@@ -149,4 +154,28 @@ public class TtsRequestLogController {
             return ResponseEntity.status(500).body(result);
         }
     }
+    
+    @PostMapping("/speech-to-text")
+    @Operation(summary = "语音转文本", description = "将语音文件转换为文本")
+    public ResponseEntity<AsrResponse> speechToText(
+            @Parameter(description = "语音转文本请求参数") @RequestBody AsrRequest request) {
+        
+        try {
+            if (!ttsRequestLogService.isAsrServiceAvailable()) {
+                return ResponseEntity.status(500).body(AsrResponse.error("ASR服务不可用，请检查配置"));
+            }
+            
+            AsrResponse response = ttsRequestLogService.speechToText(request);
+            
+            if (response.isSuccess()) {
+                return ResponseEntity.ok(response);
+            } else {
+                return ResponseEntity.status(500).body(response);
+            }
+            
+        } catch (Exception e) {
+            log.error("语音转文本失败: {}", e.getMessage(), e);
+            return ResponseEntity.status(500).body(AsrResponse.error("语音转文本失败: " + e.getMessage()));
+        }
+    }
 }

src/main/java/com/rj/controller/sys/MenuController.java

@@ -402,6 +402,8 @@ public class MenuController {
 
 
 
+
+
 
 
 

src/main/java/com/rj/controller/sys/RoleController.java

@@ -372,6 +372,8 @@ public class RoleController {
 
 
 
+
+
 
 
 

src/main/java/com/rj/controller/sys/UserRoleController.java

@@ -378,6 +378,8 @@ public class UserRoleController {
 
 
 
+
+
 
 
 

src/main/java/com/rj/dto/DifyWorkflowResponseDto.java

@@ -144,6 +144,8 @@ public class DifyWorkflowResponseDto {
 
 
 
+
+
 
 
 

src/main/java/com/rj/mapper/CustomerProfileAnalysisMapper.java

@@ -87,6 +87,8 @@ public interface CustomerProfileAnalysisMapper extends BaseMapper<CustomerProfil
 
 
 
+
+
 
 
 

src/main/java/com/rj/service/ITtsRequestLogService.java

@@ -2,6 +2,8 @@ package com.rj.service;
 
 import com.rj.entity.TtsRequestLog;
 import com.rj.utils.MinIOUrlGenerator;
+import com.rj.dto.AsrRequest;
+import com.rj.dto.AsrResponse;
 
 /**
  * TTS请求日志服务接口
@@ -41,4 +43,19 @@ public interface ITtsRequestLogService {
      * @return 是否设置成功
      */
     boolean setShortUrlByAudioName(String audioName);
+    
+    /**
+     * 语音转文本
+     * 
+     * @param request ASR请求
+     * @return ASR响应
+     */
+    AsrResponse speechToText(AsrRequest request);
+    
+    /**
+     * 检查ASR服务是否可用
+     * 
+     * @return 是否可用
+     */
+    boolean isAsrServiceAvailable();
 }

src/main/java/com/rj/service/biz/ICustomerProfileAnalysisService.java

@@ -108,6 +108,8 @@ public interface ICustomerProfileAnalysisService extends IService<CustomerProfil
 
 
 
+
+
 
 
 

src/main/java/com/rj/service/biz/impl/CustomerProfileAnalysisServiceImpl.java

@@ -143,6 +143,8 @@ public class CustomerProfileAnalysisServiceImpl extends ServiceImpl<CustomerProf
 
 
 
+
+
 
 
 

src/main/java/com/rj/service/impl/TtsRequestLogServiceImpl.java

@@ -1,15 +1,22 @@
 package com.rj.service.impl;
 
+import com.alibaba.dashscope.audio.asr.transcription.*;
 import com.rj.entity.TtsRequestLog;
 import com.rj.mapper.TtsRequestLogMapper;
 import com.rj.service.ITtsRequestLogService;
 import com.rj.utils.MinIOUrlGenerator;
+import com.rj.dto.AsrRequest;
+import com.rj.dto.AsrResponse;
 import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.beans.factory.annotation.Value;
 import org.springframework.stereotype.Service;
 
 import java.time.LocalDateTime;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
 
 /**
  * TTS请求日志服务实现类
@@ -27,6 +34,15 @@ public class TtsRequestLogServiceImpl implements ITtsRequestLogService {
     @Autowired
     private MinIOUrlGenerator urlGenerator;
     
+    @Value("${dashscope.api.key}")
+    private String apiKey;
+    
+    @Value("${asr.enabled:true}")
+    private boolean asrEnabled;
+    
+    @Value("${asr.default-model:paraformer-v2}")
+    private String defaultAsrModel;
+    
     @Override
     public boolean saveTtsRequestLog(TtsRequestLog requestLog) {
         try {
@@ -52,7 +68,7 @@ public class TtsRequestLogServiceImpl implements ITtsRequestLogService {
     @Override
     public boolean setShortUrlByAudioName(MinIOUrlGenerator.UrlInfo urlInfo ) {
         try {
-            if (urlInfo == null || urlInfo.getFileName().trim().isEmpty()) {
+            if (urlInfo == null || urlInfo.getFileName() == null || urlInfo.getFileName().trim().isEmpty()) {
                 log.error("音频名称不能为空");
                 return false;
             }
@@ -92,4 +108,131 @@ public class TtsRequestLogServiceImpl implements ITtsRequestLogService {
     public boolean setShortUrlByAudioName(String audioName) {
         return false;
     }
+    
+    @Override
+    public AsrResponse speechToText(AsrRequest request) {
+        long startTime = System.currentTimeMillis();
+        
+        try {
+            if (!isAsrServiceAvailable()) {
+                return AsrResponse.error("ASR服务不可用，请检查配置");
+            }
+            
+            if (request.getAudioUrl() == null || request.getAudioUrl().trim().isEmpty()) {
+                return AsrResponse.error("音频文件URL不能为空");
+            }
+            
+            // 设置默认值
+            if (request.getModel() == null || request.getModel().trim().isEmpty()) {
+                request.setModel(defaultAsrModel);
+            }
+            if (request.getFormat() == null || request.getFormat().trim().isEmpty()) {
+                request.setFormat("wav");
+            }
+            if (request.getSampleRate() == null) {
+                request.setSampleRate(16000);
+            }
+            if (request.getEnablePunctuation() == null) {
+                request.setEnablePunctuation(true);
+            }
+            if (request.getEnableNumberConversion() == null) {
+                request.setEnableNumberConversion(true);
+            }
+            if (request.getEnableSpeakerDiarization() == null) {
+                request.setEnableSpeakerDiarization(false);
+            }
+            if (request.getSpeakerCount() == null) {
+                request.setSpeakerCount(1);
+            }
+            
+            log.info("开始语音识别 - 音频URL: {}, 模型: {}, 格式: {}, 采样率: {}", 
+                    request.getAudioUrl(), request.getModel(), request.getFormat(), request.getSampleRate());
+            
+            // 创建转写请求参数
+            TranscriptionParam param = TranscriptionParam.builder()
+                    .apiKey(apiKey)
+                    .model(request.getModel())
+                    .fileUrls(Arrays.asList(request.getAudioUrl()))
+                    .build();
+            
+            // 创建转写对象
+            Transcription transcription = new Transcription();
+            
+            // 提交转写请求
+            TranscriptionResult result = transcription.asyncCall(param);
+            String taskId = result.getTaskId();
+            String requestId = result.getRequestId();
+            
+            log.info("ASR任务已提交 - TaskId: {}, RequestId: {}", taskId, requestId);
+            
+            // 等待任务完成
+            TranscriptionQueryParam queryParam = TranscriptionQueryParam.FromTranscriptionParam(param, taskId);
+            result = transcription.wait(queryParam);
+            log.info("语音识别完成 - result: {} ", result);
+            long processingTime = System.currentTimeMillis() - startTime;
+            
+            if (result.getOutput() != null) {
+                // 解析识别结果 - 从JSON中提取文本
+                String fullText = "";
+                Double duration = 0.0;
+                
+                try {
+                    // 从输出中提取文本和时长信息
+                    if (result.getOutput().toString().contains("\"text\"")) {
+                        // 简单解析JSON获取文本
+                        String outputStr = result.getOutput().toString();
+                        int textStart = outputStr.indexOf("\"text\":\"") + 8;
+                        int textEnd = outputStr.indexOf("\"", textStart);
+                        if (textStart > 7 && textEnd > textStart) {
+                            fullText = outputStr.substring(textStart, textEnd);
+                        }
+                    }
+                    
+                    // 提取时长信息
+                    if (result.getOutput().toString().contains("\"duration\"")) {
+                        String outputStr = result.getOutput().toString();
+                        int durationStart = outputStr.indexOf("\"duration\":") + 11;
+                        int durationEnd = outputStr.indexOf(",", durationStart);
+                        if (durationEnd == -1) {
+                            durationEnd = outputStr.indexOf("}", durationStart);
+                        }
+                        if (durationStart > 10 && durationEnd > durationStart) {
+                            try {
+                                duration = Double.parseDouble(outputStr.substring(durationStart, durationEnd).trim());
+                            } catch (NumberFormatException e) {
+                                duration = 0.0;
+                            }
+                        }
+                    }
+                } catch (Exception e) {
+                    log.warn("解析ASR结果时出错: {}", e.getMessage());
+                }
+                
+                List<AsrResponse.AsrResult> asrResults = new ArrayList<>();
+                
+                // 创建简单的结果对象
+                AsrResponse.AsrResult asrResult = new AsrResponse.AsrResult();
+                asrResult.setText(fullText);
+                asrResult.setStartTime(0.0);
+                asrResult.setEndTime(duration);
+                asrResults.add(asrResult);
+                
+                log.info("语音识别完成 - 识别文本长度: {}, 处理时间: {}ms", fullText.length(), processingTime);
+                
+                return AsrResponse.success(fullText, asrResults, duration, processingTime, taskId, requestId);
+            } else {
+                log.warn("ASR识别结果为空 - TaskId: {}", taskId);
+                return AsrResponse.error("语音识别结果为空");
+            }
+            
+        } catch (Exception e) {
+            log.error("语音识别失败: {}", e.getMessage(), e);
+            return AsrResponse.error("语音识别失败: " + e.getMessage());
+        }
+    }
+    
+    @Override
+    public boolean isAsrServiceAvailable() {
+        return asrEnabled && apiKey != null && !apiKey.trim().isEmpty();
+    }
 }

src/main/java/com/rj/service/sys/impl/MenuServiceImpl.java

@@ -115,6 +115,8 @@ public class MenuServiceImpl extends ServiceImpl<MenuMapper, Menu> implements IM
 
 
 
+
+
 
 
 

src/main/java/com/rj/service/sys/impl/RoleServiceImpl.java

@@ -115,6 +115,8 @@ public class RoleServiceImpl extends ServiceImpl<RoleMapper, Role> implements IR
 
 
 
+
+
 
 
 

src/main/java/com/rj/service/sys/impl/UserRoleServiceImpl.java

@@ -115,6 +115,8 @@ public class UserRoleServiceImpl extends ServiceImpl<UserRoleMapper, UserRole> i
 
 
 
+
+
 
 
 

src/main/resources/application-audio.yml

@@ -124,6 +124,8 @@ spring:
 
 
 
+
+
 
 
 

src/main/resources/application.yml

@@ -125,6 +125,27 @@ aliyun:
   # 是否启用阿里云服务
   enabled: true
 
+# ASR语音识别配置
+asr:
+  # 是否启用ASR服务
+  enabled: true
+  # 默认ASR模型
+  default-model: paraformer-v2
+  # 请求超时时间（毫秒）
+  timeout: 60000
+  # 是否启用标点符号
+  enable-punctuation: true
+  # 是否启用数字转换
+  enable-number-conversion: true
+  # 是否启用说话人分离
+  enable-speaker-diarization: false
+  # 默认说话人数量
+  default-speaker-count: 1
+  # 默认音频格式
+  default-format: wav
+  # 默认采样率
+  default-sample-rate: 16000
+
 
 springdoc:
   swagger-ui:

src/main/resources/mapper/AudioManagementStatisticsMapper.xml

@@ -101,6 +101,8 @@
 
 
 
+
+
 
 
 

src/main/resources/static/tts-demo.html

@@ -300,6 +300,8 @@
 
 
 
+
+
 
 
 

src/test/java/com/rj/service/FaceDetectImageCountTest.java

@@ -238,6 +238,8 @@ public class FaceDetectImageCountTest {
 
 
 
+
+
 
 
 

src/test/java/com/rj/service/TtsRequestLogShortUrlTest.java

@@ -179,6 +179,8 @@ public class TtsRequestLogShortUrlTest {
 
 
 
+
+
 
 
 

src/test/java/com/rj/service/VideoSynthesisTempUrlTest.java

@@ -157,6 +157,8 @@ public class VideoSynthesisTempUrlTest {
 
 
 
+
+
 
 
 

src/test/java/com/rj/service/VideoSynthesisVideoNameTest.java

@@ -132,6 +132,8 @@ public class VideoSynthesisVideoNameTest {
 
 
 
+
+