feat: 음성 답변 STT 흐름 안정화

backend/src/main/java/com/stackup/stackup/common/exception/ApiErrorCode.java

@@ -34,6 +34,8 @@ public enum ApiErrorCode {
     SESSION_NOT_FOUND(HttpStatus.NOT_FOUND, "세션을 찾을 수 없습니다."),
     SESSION_FORBIDDEN(HttpStatus.FORBIDDEN, "세션에 접근할 수 없습니다."),
     FEEDBACK_NOT_READY(HttpStatus.NOT_FOUND, "피드백이 아직 생성되지 않았습니다."),
+    VOICE_EMPTY_FILE(HttpStatus.BAD_REQUEST, "음성 파일을 업로드할 수 없습니다."),
+    VOICE_FILE_TOO_LARGE(HttpStatus.BAD_REQUEST, "음성 파일 크기가 너무 큽니다."),
     VOICE_INVALID_CONTENT_TYPE(HttpStatus.BAD_REQUEST, "지원하지 않는 음성 형식입니다."),
     VOICE_MESSAGE_NOT_FOUND(HttpStatus.NOT_FOUND, "음성 메시지를 찾을 수 없습니다."),
 

backend/src/main/java/com/stackup/stackup/session/application/InterviewMessageService.java

@@ -8,6 +8,8 @@
 import com.stackup.stackup.session.domain.InterviewMessageRepository;
 import com.stackup.stackup.session.domain.InterviewSession;
 import com.stackup.stackup.session.domain.InterviewSessionRepository;
+import com.stackup.stackup.session.domain.MessageRole;
+import com.stackup.stackup.session.domain.MessageStatus;
 import com.stackup.stackup.session.domain.SessionStatus;
 import java.util.List;
 import lombok.RequiredArgsConstructor;
@@ -51,22 +53,33 @@ public MessageResult submitAnswer(Long userId, Long sessionId, String content, S
         InterviewMessage latest = messageRepository
             .findFirstBySession_IdOrderBySequenceNumberDesc(sessionId)
             .orElseThrow(() -> new DomainException(ApiErrorCode.SESSION_INVALID_STATE));
-        if (latest.getRole() != com.stackup.stackup.session.domain.MessageRole.INTERVIEWER) {
-            // 직전 메시지가 질문이 아니면 답변 불가
-            throw new DomainException(ApiErrorCode.SESSION_INVALID_STATE);
-        }
+        InterviewMessage parentQuestion = resolveAnswerParent(latest);
         int nextSeq = latest.getSequenceNumber() + 1;
         InterviewMessage answer = messageRepository.save(
-            InterviewMessage.interviewee(session, nextSeq, content, latest,
+            InterviewMessage.interviewee(session, nextSeq, content, parentQuestion,
                 idempotencyKey != null && !idempotencyKey.isBlank() ? idempotencyKey : null)
         );
 
         events.publishEvent(new AnswerSubmittedEvent(
-            userId, sessionId, latest.getId(), answer.getId()
+            userId, sessionId, parentQuestion.getId(), answer.getId()
         ));
         return MessageResult.of(answer);
     }
 
+    private InterviewMessage resolveAnswerParent(InterviewMessage latest) {
+        if (latest.getRole() == MessageRole.INTERVIEWER) {
+            return latest;
+        }
+        if (latest.getRole() == MessageRole.INTERVIEWEE
+            && latest.getStatus() == MessageStatus.FAILED
+            && latest.getAudioFilePath() != null
+            && latest.getParentMessage() != null
+            && latest.getParentMessage().getRole() == MessageRole.INTERVIEWER) {
+            return latest.getParentMessage();
+        }
+        throw new DomainException(ApiErrorCode.SESSION_INVALID_STATE);
+    }
+
     private InterviewSession ownedSession(Long userId, Long sessionId) {
         return sessionRepository.findByIdAndUser_IdAndDeletedFalse(sessionId, userId)
             .orElseThrow(() -> new DomainException(ApiErrorCode.SESSION_NOT_FOUND));

backend/src/main/java/com/stackup/stackup/session/application/VoiceAnswerUploadService.java

@@ -96,16 +96,19 @@ public MessageResult submit(Long userId, Long sessionId, VoiceAnswerUploadComman
     }
 
     private void validate(VoiceAnswerUploadCommand cmd) {
-        if (cmd.size() <= 0 || cmd.size() > MAX_BYTES) {
-            throw new DomainException(ApiErrorCode.RESUME_FILE_TOO_LARGE);
+        if (cmd == null || cmd.content() == null || cmd.size() <= 0) {
+            throw new DomainException(ApiErrorCode.VOICE_EMPTY_FILE);
         }
-        if (cmd.contentType() == null || !ALLOWED_CONTENT_TYPES.contains(cmd.contentType().toLowerCase())) {
+        if (cmd.size() > MAX_BYTES) {
+            throw new DomainException(ApiErrorCode.VOICE_FILE_TOO_LARGE);
+        }
+        if (cmd.contentType() == null || !ALLOWED_CONTENT_TYPES.contains(cmd.contentType().trim().toLowerCase())) {
             throw new DomainException(ApiErrorCode.VOICE_INVALID_CONTENT_TYPE);
         }
     }
 
     private static String buildKey(Long sessionId, Long messageId, String contentType) {
-        String ext = switch (contentType.toLowerCase()) {
+        String ext = switch (contentType.trim().toLowerCase()) {
             case "audio/webm" -> "webm";
             case "audio/ogg" -> "ogg";
             case "audio/mpeg" -> "mp3";

backend/src/main/java/com/stackup/stackup/session/application/VoiceCallbackService.java

@@ -11,7 +11,6 @@
 import com.stackup.stackup.session.application.event.AnswerSubmittedEvent;
 import com.stackup.stackup.session.domain.InterviewMessage;
 import com.stackup.stackup.session.domain.InterviewMessageRepository;
-import com.stackup.stackup.session.domain.MessageStatus;
 import com.stackup.stackup.session.domain.MessageVoiceAnalysis;
 import com.stackup.stackup.session.domain.MessageVoiceAnalysisRepository;
 import lombok.RequiredArgsConstructor;
@@ -64,14 +63,19 @@ public void apply(VoiceCallbackEnvelope envelope) {
         }
 
         if (p.errorCode() != null && !p.errorCode().isBlank()) {
-            message.markStatus(MessageStatus.FAILED);
-            sseEventPublisher.publishToSession(p.sessionId(), SseEventType.SESSION_MESSAGE,
-                new VoiceFailedNotice(p.sessionId(), message.getId(), p.errorCode()));
+            failVoiceMessage(p.sessionId(), message, p.errorCode());
             markProcessed(envelope.messageId());
             log.warn("callback.voice STT failed. sessionId={}, msg={}, code={}",
                 p.sessionId(), message.getId(), p.errorCode());
             return;
         }
+        if (p.transcript() == null || p.transcript().isBlank()) {
+            failVoiceMessage(p.sessionId(), message, "STT_EMPTY_TRANSCRIPT");
+            markProcessed(envelope.messageId());
+            log.warn("callback.voice STT empty transcript. sessionId={}, msg={}",
+                p.sessionId(), message.getId());
+            return;
+        }
 
         message.completeWithTranscript(p.transcript());
 
@@ -113,7 +117,20 @@ public void apply(VoiceCallbackEnvelope envelope) {
     public record VoiceTranscribedNotice(Long sessionId, Long messageId, String transcript) {
     }
 
-    public record VoiceFailedNotice(Long sessionId, Long messageId, String errorCode) {
+    public record VoiceFailedNotice(Long sessionId, Long messageId, String errorCode, String fallbackText) {
+    }
+
+    private void failVoiceMessage(Long sessionId, InterviewMessage message, String errorCode) {
+        message.failVoiceTranscription();
+        VoiceFailedNotice notice = new VoiceFailedNotice(
+            sessionId,
+            message.getId(),
+            errorCode,
+            message.getContent()
+        );
+        sseEventPublisher.publishToSession(sessionId, SseEventType.SESSION_MESSAGE, notice);
+        sseEventPublisher.publishToUser(message.getSession().getUser().getId(),
+            SseEventType.SESSION_MESSAGE, notice);
     }
 
     private String fillerToJson(java.util.Map<String, Integer> map) {

backend/src/main/java/com/stackup/stackup/session/application/dto/MessageResult.java

@@ -11,6 +11,7 @@ public record MessageResult(
     Integer sequenceNumber,
     MessageRole role,
     String content,
+    String audioFilePath,
     Long parentMessageId,
     MessageStatus status,
     Instant createdAt
@@ -22,6 +23,7 @@ public static MessageResult of(InterviewMessage m) {
             m.getSequenceNumber(),
             m.getRole(),
             m.getContent(),
+            m.getAudioFilePath(),
             m.getParentMessage() == null ? null : m.getParentMessage().getId(),
             m.getStatus(),
             m.getCreatedAt()

backend/src/main/java/com/stackup/stackup/session/domain/InterviewMessage.java

@@ -33,6 +33,10 @@
 @NoArgsConstructor(access = AccessLevel.PROTECTED)
 public class InterviewMessage extends BaseTimeEntity {
 
+    public static final String VOICE_TRANSCRIPTION_PENDING_TEXT = "(transcribing)";
+    public static final String VOICE_TRANSCRIPTION_FAILED_TEXT =
+        "음성 인식에 실패했습니다. 텍스트로 다시 답변해 주세요.";
+
     @Id
     @GeneratedValue(strategy = GenerationType.IDENTITY)
     private Long id;
@@ -106,7 +110,8 @@ public static InterviewMessage interviewee(InterviewSession session, int seq, St
     // 음성 답변 placeholder. content 는 STT 완료 후 채움. 생성 시 CHECK 제약 만족을 위해 임시 공백 + audio_file_path 후속 갱신.
     public static InterviewMessage voiceInterviewee(InterviewSession session, int seq,
                                                     InterviewMessage parent, String idempotencyKey) {
-        InterviewMessage m = new InterviewMessage(session, seq, MessageRole.INTERVIEWEE, "(transcribing)",
+        InterviewMessage m = new InterviewMessage(session, seq, MessageRole.INTERVIEWEE,
+            VOICE_TRANSCRIPTION_PENDING_TEXT,
             parent, MessageStatus.CREATED, idempotencyKey);
         return m;
     }
@@ -127,4 +132,9 @@ public void completeWithTranscript(String transcript) {
         }
         this.status = MessageStatus.COMPLETED;
     }
+
+    public void failVoiceTranscription() {
+        this.content = VOICE_TRANSCRIPTION_FAILED_TEXT;
+        this.status = MessageStatus.FAILED;
+    }
 }

backend/src/main/java/com/stackup/stackup/session/presentation/dto/MessageResponse.java

@@ -11,6 +11,7 @@ public record MessageResponse(
     Integer sequenceNumber,
     MessageRole role,
     String content,
+    String audioFilePath,
     Long parentMessageId,
     MessageStatus status,
     Instant createdAt
@@ -22,6 +23,7 @@ public static MessageResponse from(MessageResult r) {
             r.sequenceNumber(),
             r.role(),
             r.content(),
+            r.audioFilePath(),
             r.parentMessageId(),
             r.status(),
             r.createdAt()

backend/src/test/java/com/stackup/stackup/session/application/InterviewMessageServiceTest.java

@@ -20,6 +20,7 @@
 import java.util.Optional;
 import org.junit.jupiter.api.Test;
 import org.junit.jupiter.api.extension.ExtendWith;
+import org.mockito.ArgumentCaptor;
 import org.mockito.InjectMocks;
 import org.mockito.Mock;
 import org.mockito.junit.jupiter.MockitoExtension;
@@ -84,6 +85,35 @@ void submitAnswer_rejectsWhenPreviousMessageNotInterviewer() {
             .isInstanceOf(DomainException.class);
     }
 
+    @Test
+    void submitAnswer_allowsTextRetryAfterFailedVoiceAnswer() {
+        InterviewSession session = sessionInProgress(10L);
+        InterviewMessage question = InterviewMessage.interviewer(session, 1, "Q1?");
+        ReflectionTestUtils.setField(question, "id", 100L);
+        InterviewMessage failedVoice = InterviewMessage.voiceInterviewee(session, 2, question, null);
+        ReflectionTestUtils.setField(failedVoice, "id", 200L);
+        failedVoice.attachAudio("interview/voice/raw/10/200.webm");
+        failedVoice.failVoiceTranscription();
+
+        when(sessionRepository.findByIdAndUser_IdAndDeletedFalse(10L, 1L)).thenReturn(Optional.of(session));
+        when(messageRepository.findFirstBySession_IdOrderBySequenceNumberDesc(10L)).thenReturn(Optional.of(failedVoice));
+        when(messageRepository.save(any(InterviewMessage.class))).thenAnswer(inv -> {
+            InterviewMessage m = inv.getArgument(0);
+            ReflectionTestUtils.setField(m, "id", 201L);
+            return m;
+        });
+
+        MessageResult result = service.submitAnswer(1L, 10L, "retry as text", null);
+
+        assertThat(result.id()).isEqualTo(201L);
+        assertThat(result.sequenceNumber()).isEqualTo(3);
+        assertThat(result.parentMessageId()).isEqualTo(100L);
+        ArgumentCaptor<AnswerSubmittedEvent> eventCaptor = ArgumentCaptor.forClass(AnswerSubmittedEvent.class);
+        verify(events).publishEvent(eventCaptor.capture());
+        assertThat(eventCaptor.getValue().parentQuestionMessageId()).isEqualTo(100L);
+        assertThat(eventCaptor.getValue().answerMessageId()).isEqualTo(201L);
+    }
+
     private InterviewSession sessionInProgress(Long id) {
         User user = User.createGithubUser(1L, "u", null, null, "t");
         ReflectionTestUtils.setField(user, "id", 1L);