speech-demo-service/src/main/java/com/supervision/service/impl/ChatServiceImpl.java

package com.supervision.service.impl;

import cn.hutool.core.codec.Base64;
import com.supervision.dto.dify.ChatResDTO;
import com.supervision.dto.paddlespeech.res.TtsResultDTO;
import com.supervision.dto.robot.AnswerInfo;
import com.supervision.dto.robot.AskInfo;
import com.supervision.dto.robot.RobotTalkDTO;
import com.supervision.model.RobotTalkReq;
import com.supervision.model.dify.DIFYChatReqInputVO;
import com.supervision.model.dify.DifyChatReqVO;
import com.supervision.service.IChatService;
import com.supervision.util.AsrUtil;
import com.supervision.util.DifyApiUtil;
import com.supervision.util.TtsUtil;
import jakarta.annotation.Resource;
import jakarta.servlet.http.HttpServletResponse;
import lombok.extern.slf4j.Slf4j;
import org.springframework.stereotype.Service;
import org.springframework.util.StopWatch;
import org.springframework.web.multipart.MultipartFile;

import java.io.IOException;
import java.util.HashMap;
import java.util.Map;
import java.util.UUID;

@Slf4j
@Service
public class ChatServiceImpl implements IChatService {

    @Resource
    private DifyApiUtil difyApiUtil;

    Map<String, String> voiceCache = new HashMap<>();


    @Override
    public RobotTalkDTO talk(MultipartFile file, RobotTalkReq robotTalkReq) {
        log.info("robotTalkReq:{}", robotTalkReq);
        RobotTalkDTO.RobotTalkDTOBuilder builder = RobotTalkDTO.builder();

        try {
            byte[] bytes = file.getBytes();
            StopWatch stopWatch = new StopWatch();
            DifyChatReqVO difyChatReqVO = new DifyChatReqVO();
            difyChatReqVO.setUser("admin");
            DIFYChatReqInputVO inputs = new DIFYChatReqInputVO();
            stopWatch.start("stt");
            stopWatch.stop();
            difyChatReqVO.setQuery(AsrUtil.asrTransformByBytes(bytes));
            difyChatReqVO.setConversation_id(robotTalkReq.getSessionId());
            stopWatch.start("dify");
            ChatResDTO chatResDTO = difyApiUtil.chat(difyChatReqVO);
            stopWatch.stop();
            log.info("response:{}", chatResDTO.getAnswer());
            builder.askInfo(AskInfo.builder().contentType(2).message(inputs.getQuery()).audioLength(100L).askId(chatResDTO.getMessage_id()).build());
            voiceCache.put(chatResDTO.getMessage_id(), Base64.encode(bytes));
            stopWatch.start("tts");
            TtsResultDTO ttsResultDTO = TtsUtil.ttsTransform(chatResDTO.getAnswer());
            stopWatch.stop();
            String voiceBaseId = UUID.randomUUID().toString();
            builder.answerInfo(AnswerInfo.builder().contentType(2).message(chatResDTO.getAnswer()).voiceBaseId(voiceBaseId).voiceBase64(ttsResultDTO.getAudio()).build());
            builder.sessionId(chatResDTO.getConversation_id());
            voiceCache.put(voiceBaseId, ttsResultDTO.getAudio());
            log.info("耗时：{}", stopWatch.prettyPrint());
        } catch (IOException e) {
            throw new RuntimeException(e);
        }

        return builder.build();
    }

    @Override
    public void getAudio(HttpServletResponse response, String audioId) throws IOException {

        log.info("audioId:{}", audioId);
        Base64.decodeToStream(voiceCache.get(audioId), response.getOutputStream(), false);
    }
}