feat: 添加会议转录文件服务

- 新增 `MeetingTranscriptFileServiceImpl` 实现会议转录文件的初始化和导出功能
- 定义 `MeetingTranscriptExportResult` 数据传输对象，用于封装导出结果
- 定义 `MeetingTranscriptFileService` 接口，提供初始化和导出会议转录文件的方法

backend/src/main/java/com/imeeting/common/RedisKeys.java

@@ -43,6 +43,10 @@ public final class RedisKeys {
     return "biz:meeting:polling:lock:" + meetingId;
   }
 
+  public static String meetingSummaryLockKey(Long meetingId) {
+    return "biz:meeting:summary:lock:" + meetingId;
+  }
+
   public static String realtimeMeetingSocketSessionKey(String sessionToken) {
     return "biz:meeting:realtime:socket:" + sessionToken;
   }

backend/src/main/java/com/imeeting/common/SysParamKeys.java

@@ -5,5 +5,7 @@ public final class SysParamKeys {
 
   public static final String CAPTCHA_ENABLED = "security.captcha.enabled";
   public static final String MEETING_SUMMARY_SYSTEM_PROMPT = "meeting.summary.system_prompt";
+  public static final String MEETING_TRANSCRIPT_CLEANUP_FILLER_WORDS = "meeting.transcript.cleanup.filler_words";
+  public static final String MEETING_TRANSCRIPT_CLEANUP_REPLACEMENTS = "meeting.transcript.cleanup.replacements";
   public static final String MEETING_OFFLINE_AUDIO_MAX_SIZE_MB = "meeting.offline_audio.max_size_mb";
 }

backend/src/main/java/com/imeeting/controller/android/AndroidMeetingController.java

@@ -251,6 +251,20 @@ public class AndroidMeetingController {
 
         Integer realtimeProgress = resolveRealtimeProgress(meetingId);
         if (realtimeProgress != null) {
+            if (realtimeProgress >= 100) {
+                MeetingVO completedDetail = detail != null ? detail : meetingQueryService.getDetail(meetingId);
+                boolean completedHasSummary = completedDetail != null
+                        && completedDetail.getSummaryContent() != null
+                        && !completedDetail.getSummaryContent().isBlank();
+                if (completedHasSummary) {
+                    return new LegacyMeetingPreviewResult("200", "success", buildCompletedPreview(meeting, completedDetail, summaryTask));
+                }
+                return new LegacyMeetingPreviewResult(
+                        "504",
+                        "处理已完成，但摘要尚未同步，请稍后重试",
+                        buildProcessingPreview(meeting, summaryTask, processingStatus("摘要已生成，可查看详情", 100, STAGE_COMPLETED))
+                );
+            }
             if (realtimeProgress < 90) {
                 return new LegacyMeetingPreviewResult(
                         "400",
@@ -258,7 +272,7 @@ public class AndroidMeetingController {
                         buildProcessingPreview(meeting, summaryTask, processingStatus("正在转写音频", 50, STAGE_AUDIO_TRANSCRIPTION))
                 );
             }
-            if (realtimeProgress == 90) {
+            if (realtimeProgress >= 90) {
                 return new LegacyMeetingPreviewResult(
                         "400",
                         "会议正在处理中",

backend/src/main/java/com/imeeting/controller/android/legacy/LegacyMeetingController.java

@@ -237,6 +237,9 @@ public class LegacyMeetingController {
                 : null;
         boolean hasSummary = detail != null && detail.getSummaryContent() != null && !detail.getSummaryContent().isBlank();
 
+        if (hasSummary) {
+            return new LegacyMeetingPreviewResult("200", "success", buildCompletedPreview(meeting, detail, summaryTask));
+        }
         if (summaryCompleted) {
             return new LegacyMeetingPreviewResult("200", "success", buildCompletedPreview(meeting, detail, summaryTask));
         }
@@ -264,17 +267,31 @@ public class LegacyMeetingController {
 
         Integer realtimeProgress = resolveRealtimeProgress(meetingId);
         if (realtimeProgress != null) {
+            if (realtimeProgress >= 100) {
+                MeetingVO completedDetail = detail != null ? detail : meetingQueryService.getDetail(meetingId);
+                boolean completedHasSummary = completedDetail != null
+                        && completedDetail.getSummaryContent() != null
+                        && !completedDetail.getSummaryContent().isBlank();
+                if (completedHasSummary) {
+                    return new LegacyMeetingPreviewResult("200", "success", buildCompletedPreview(meeting, completedDetail, summaryTask));
+                }
+                return new LegacyMeetingPreviewResult(
+                        "504",
+                        "处理已完成，但摘要尚未同步，请稍后重试",
+                        buildProcessingPreview(meeting, summaryTask, processingStatus("摘要已生成，可扫码查看", 100, STAGE_COMPLETED))
+                );
+            }
             if (realtimeProgress < 90) {
                 return new LegacyMeetingPreviewResult(
                         "400",
-                        "浼氳姝ｅ湪澶勭悊涓?",
+                        "会议正在处理中",
                         buildProcessingPreview(meeting, summaryTask, processingStatus("姝ｅ湪杞瘧闊抽", 50, STAGE_AUDIO_TRANSCRIPTION))
                 );
             }
-            if (realtimeProgress == 90) {
+            if (realtimeProgress >= 90) {
                 return new LegacyMeetingPreviewResult(
                         "400",
-                        "浼氳姝ｅ湪澶勭悊涓?",
+                        "会议正在处理中",
                         buildProcessingPreview(meeting, summaryTask, processingStatus("姝ｅ湪鐢熸垚鎬荤粨", 75, STAGE_SUMMARY_GENERATION))
                 );
             }

backend/src/main/java/com/imeeting/dto/biz/MeetingSummarySource.java

@@ -0,0 +1,32 @@
+package com.imeeting.dto.biz;
+
+import lombok.Builder;
+import lombok.Data;
+
+import java.util.LinkedHashMap;
+import java.util.Map;
+
+@Data
+@Builder
+public class MeetingSummarySource {
+    private String text;
+    private String sourceType;
+    private Long revisionId;
+    private boolean fallbackUsed;
+    private String sourceFingerprint;
+    private String triggerTaskType;
+    private String semanticCorrector;
+    private String ruleProfileVersion;
+
+    public Map<String, Object> toSnapshot() {
+        Map<String, Object> snapshot = new LinkedHashMap<>();
+        snapshot.put("sourceType", sourceType);
+        snapshot.put("revisionId", revisionId);
+        snapshot.put("fallbackUsed", fallbackUsed);
+        snapshot.put("sourceFingerprint", sourceFingerprint);
+        snapshot.put("triggerTaskType", triggerTaskType);
+        snapshot.put("semanticCorrector", semanticCorrector);
+        snapshot.put("ruleProfileVersion", ruleProfileVersion);
+        return snapshot;
+    }
+}

backend/src/main/java/com/imeeting/entity/biz/MeetingTranscriptRevision.java

@@ -0,0 +1,57 @@
+package com.imeeting.entity.biz;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+import java.time.LocalDateTime;
+
+@Data
+@Schema(description = "会议转录修正版")
+@TableName("biz_meeting_transcript_revisions")
+public class MeetingTranscriptRevision {
+    @TableId(value = "id", type = IdType.AUTO)
+    @Schema(description = "修正版ID")
+    private Long id;
+
+    @Schema(description = "会议ID")
+    private Long meetingId;
+
+    @Schema(description = "触发本次修正版生成的任务ID")
+    private Long sourceTaskId;
+
+    @Schema(description = "版本号")
+    private Integer revisionNo;
+
+    @Schema(description = "状态")
+    private Integer status;
+
+    @Schema(description = "修正版全文")
+    private String cleanedFullText;
+
+    @Schema(description = "结果文件路径")
+    private String resultFilePath;
+
+    @Schema(description = "规则配置快照")
+    private String ruleProfile;
+
+    @Schema(description = "片段数")
+    private Integer segmentCount;
+
+    @Schema(description = "删除片段数")
+    private Integer droppedSegmentCount;
+
+    @Schema(description = "合并组数")
+    private Integer mergedGroupCount;
+
+    @Schema(description = "是否当前生效")
+    private Integer isCurrent;
+
+    @Schema(description = "创建时间")
+    private LocalDateTime createdAt;
+
+    @Schema(description = "更新时间")
+    private LocalDateTime updatedAt;
+}

backend/src/main/java/com/imeeting/entity/biz/MeetingTranscriptRevisionItem.java

@@ -0,0 +1,60 @@
+package com.imeeting.entity.biz;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+import java.time.LocalDateTime;
+
+@Data
+@Schema(description = "会议转录修正版明细")
+@TableName("biz_meeting_transcript_revision_items")
+public class MeetingTranscriptRevisionItem {
+    @TableId(value = "id", type = IdType.AUTO)
+    @Schema(description = "修正版明细ID")
+    private Long id;
+
+    @Schema(description = "修正版ID")
+    private Long revisionId;
+
+    @Schema(description = "原始转录ID")
+    private Long sourceTranscriptId;
+
+    @Schema(description = "原始排序值")
+    private Integer sourceSortOrder;
+
+    @Schema(description = "原始说话人ID")
+    private String sourceSpeakerId;
+
+    @Schema(description = "原始说话人名称")
+    private String sourceSpeakerName;
+
+    @Schema(description = "原始内容")
+    private String sourceContent;
+
+    @Schema(description = "清洗后内容")
+    private String cleanedContent;
+
+    @Schema(description = "清洗后说话人名称")
+    private String cleanedSpeakerName;
+
+    @Schema(description = "动作类型")
+    private String actionType;
+
+    @Schema(description = "合并组ID")
+    private String mergeGroupId;
+
+    @Schema(description = "置信度")
+    private java.math.BigDecimal confidence;
+
+    @Schema(description = "命中规则")
+    private String ruleHits;
+
+    @Schema(description = "上下文快照")
+    private String contextSnapshot;
+
+    @Schema(description = "创建时间")
+    private LocalDateTime createdAt;
+}

backend/src/main/java/com/imeeting/listener/MeetingTaskRecoveryListener.java

@@ -59,6 +59,7 @@ public class MeetingTaskRecoveryListener implements ApplicationRunner {
 
                 // 2. 清理旧的 Redis 锁和进度缓存，确保恢复线程能拿到控制权
                 redisTemplate.delete(RedisKeys.meetingPollingLockKey(meeting.getId()));
+                redisTemplate.delete(RedisKeys.meetingSummaryLockKey(meeting.getId()));
                 
                 // 3. 根据状态重新派发任务 (平滑拉起)
                 if (meeting.getStatus() == 1) {

backend/src/main/java/com/imeeting/mapper/biz/MeetingTranscriptRevisionItemMapper.java

@@ -0,0 +1,9 @@
+package com.imeeting.mapper.biz;
+
+import com.baomidou.mybatisplus.core.mapper.BaseMapper;
+import com.imeeting.entity.biz.MeetingTranscriptRevisionItem;
+import org.apache.ibatis.annotations.Mapper;
+
+@Mapper
+public interface MeetingTranscriptRevisionItemMapper extends BaseMapper<MeetingTranscriptRevisionItem> {
+}

backend/src/main/java/com/imeeting/mapper/biz/MeetingTranscriptRevisionMapper.java

@@ -0,0 +1,9 @@
+package com.imeeting.mapper.biz;
+
+import com.baomidou.mybatisplus.core.mapper.BaseMapper;
+import com.imeeting.entity.biz.MeetingTranscriptRevision;
+import org.apache.ibatis.annotations.Mapper;
+
+@Mapper
+public interface MeetingTranscriptRevisionMapper extends BaseMapper<MeetingTranscriptRevision> {
+}

backend/src/main/java/com/imeeting/service/biz/MeetingTranscriptRevisionService.java

@@ -0,0 +1,21 @@
+package com.imeeting.service.biz;
+
+import com.imeeting.dto.biz.AiModelVO;
+import com.imeeting.dto.biz.MeetingSummarySource;
+import com.imeeting.dto.biz.MeetingTranscriptVO;
+import com.imeeting.entity.biz.AiTask;
+import com.imeeting.entity.biz.Meeting;
+
+import java.util.List;
+
+public interface MeetingTranscriptRevisionService {
+    String generateOfflineCurrentRevision(Meeting meeting, AiTask task, AiModelVO asrModel);
+
+    MeetingSummarySource resolveSummarySource(Meeting meeting, AiTask summaryTask, AiModelVO asrModel);
+
+    List<MeetingTranscriptVO> listEffectiveTranscripts(Long meetingId);
+
+    boolean updateCurrentRevisionContent(Long meetingId, Long operatorId, String content);
+
+    void invalidateCurrentRevision(Long meetingId);
+}

backend/src/main/java/com/imeeting/service/biz/impl/AiTaskServiceImpl.java

@@ -7,6 +7,7 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 
 import com.imeeting.common.RedisKeys;
 import com.imeeting.dto.biz.AiModelVO;
+import com.imeeting.dto.biz.MeetingSummarySource;
 import com.imeeting.entity.biz.AiTask;
 import com.imeeting.entity.biz.HotWord;
 import com.imeeting.entity.biz.Meeting;
@@ -20,6 +21,7 @@ import com.imeeting.service.biz.AiTaskService;
 import com.imeeting.service.biz.HotWordService;
 import com.imeeting.service.biz.MeetingSummaryFileService;
 import com.imeeting.service.biz.MeetingTranscriptFileService;
+import com.imeeting.service.biz.MeetingTranscriptRevisionService;
 
 import com.unisbase.entity.SysUser;
 import com.unisbase.mapper.SysUserMapper;
@@ -60,6 +62,7 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
     private final StringRedisTemplate redisTemplate;
     private final MeetingSummaryFileService meetingSummaryFileService;
     private final MeetingTranscriptFileService meetingTranscriptFileService;
+    private final MeetingTranscriptRevisionService meetingTranscriptRevisionService;
     private final MeetingSummaryPromptAssembler meetingSummaryPromptAssembler;
     private final TaskSecurityContextRunner taskSecurityContextRunner;
 
@@ -99,7 +102,7 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
                     .last("limit 1"));
 
             String asrText = "";
-            if (asrTask != null && asrTask.getStatus() == 0) {
+            if (asrTask != null && canExecuteTask(asrTask)) {
                 asrText = processAsrTask(meeting, asrTask);
             } else {
                 List<MeetingTranscript> transcripts = transcriptMapper.selectList(new LambdaQueryWrapper<MeetingTranscript>()
@@ -129,13 +132,11 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
                 updateProgress(meetingId, -1, "未识别到可用于总结的转录内容", 0);
                 return;
             }
-            if (sumTask != null && sumTask.getStatus() == 0) {
-                processSummaryTask(meeting, asrText, sumTask);
+            if (sumTask != null && canExecuteTask(sumTask)) {
+                executeSummaryFlow(meeting, sumTask, resolveAsrModelForRevision(asrTask));
             } else if (meeting.getStatus() != 3) {
                 updateMeetingStatus(meetingId, 3);
             }
-
-            redisTemplate.delete(RedisKeys.meetingProgressKey(meetingId));
         } catch (Exception e) {
             log.error("Meeting {} AI Task Flow failed", meetingId, e);
             failPendingSummaryTask(findLatestSummaryTask(meetingId), "转录失败，已跳过总结任务：" + e.getMessage());
@@ -154,34 +155,19 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
 
     private void doDispatchSummaryTask(Long meetingId) {
         Meeting meeting = meetingMapper.selectById(meetingId);
-        if (meeting == null) return;
-        AiTask sumTask = this.getOne(new LambdaQueryWrapper<AiTask>()
-                .eq(AiTask::getMeetingId, meetingId)
-                .eq(AiTask::getTaskType, "SUMMARY")
-                .orderByDesc(AiTask::getId)
-                .last("limit 1"));
+        if (meeting == null) {
+            return;
+        }
+        AiTask sumTask = findLatestTask(meetingId, "SUMMARY");
+        AiTask asrTask = findLatestTask(meetingId, "ASR");
         try {
-            List<MeetingTranscript> transcripts = transcriptMapper.selectList(new LambdaQueryWrapper<MeetingTranscript>()
-                    .eq(MeetingTranscript::getMeetingId, meetingId)
-                    .orderByAsc(MeetingTranscript::getStartTime));
-
-            if (transcripts.isEmpty()) {
-                failPendingSummaryTask(sumTask, "没有可用于总结的转录内容");
-                throw new RuntimeException("没有找到可用的转录文本，无法生成总结");
-            }
-
-            String asrText = buildTranscriptText(transcripts);
-            if (asrText == null || asrText.isBlank()) {
-                failPendingSummaryTask(sumTask, "没有可用于总结的转录内容");
-                throw new RuntimeException("没有可用于总结的转录内容");
-            }
-            if (sumTask != null && sumTask.getStatus() == 0) {
-                processSummaryTask(meeting, asrText, sumTask);
+            if (sumTask != null && canExecuteTask(sumTask)) {
+                executeSummaryFlow(meeting, sumTask, resolveAsrModelForRevision(asrTask));
             }
         } catch (Exception e) {
             log.error("Re-summary failed for meeting {}", meetingId, e);
             updateMeetingStatus(meetingId, 4);
-            updateProgress(meetingId, -1, "总结失败: " + e.getMessage(), 0);
+            updateProgress(meetingId, -1, "Summary flow failed: " + e.getMessage(), 0);
         }
     }
 
@@ -197,8 +183,11 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
         if (asrModel == null) throw new RuntimeException("ASR模型配置不存在");
 
         String submitUrl = appendPath(asrModel.getBaseUrl(), "api/v1/asr/transcriptions");
-        String taskId = null;
+        String taskId = taskRecord.getResponseData() != null
+                ? String.valueOf(taskRecord.getResponseData().getOrDefault("task_id", ""))
+                : "";
 
+        if (taskId == null || taskId.isBlank()) {
             updateProgress(meeting.getId(), 5, "正在提交识别请求...", 0);
             Map<String, Object> req = buildAsrRequest(meeting, taskRecord, asrModel);
             taskRecord.setRequestData(req);
@@ -207,12 +196,16 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
             String respBody = postJson(submitUrl, req, asrModel.getApiKey());
             JsonNode submitNode = objectMapper.readTree(respBody);
             if (submitNode.path("code").asInt() != 0) {
-            updateAiTaskFail(taskRecord, "提交失败：" + respBody);
+                updateAiTaskFail(taskRecord, "提交失败: " + respBody);
                 throw new RuntimeException("ASR引擎拒绝请求: " + submitNode.path("msg").asText());
             }
             taskId = submitNode.path("data").path("task_id").asText();
             taskRecord.setResponseData(Map.of("task_id", taskId));
             this.updateById(taskRecord);
+        } else {
+            updateProgress(meeting.getId(), 5, "Resuming ASR polling...", 0);
+        }
+        this.updateById(taskRecord);
 
         String queryUrl = appendPath(asrModel.getBaseUrl(), "api/v1/asr/transcriptions/" + taskId);
 
@@ -473,12 +466,16 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
                 .last("limit 1"));
     }
 
-    private void processSummaryTask(Meeting meeting, String asrText, AiTask taskRecord) throws Exception {
+    private void processSummaryTask(Meeting meeting, MeetingSummarySource summarySource, AiTask taskRecord) throws Exception {
+        String asrText = summarySource.getText();
         updateMeetingStatus(meeting.getId(), 2);
         updateProgress(meeting.getId(), 90, "正在生成智能总结纪要...", 0);
 
         taskRecord.setStatus(1);
         taskRecord.setStartedAt(LocalDateTime.now());
+        Map<String, Object> initialResponseData = new HashMap<>();
+        initialResponseData.put("summarySource", summarySource.toSnapshot());
+        taskRecord.setResponseData(initialResponseData);
         this.updateById(taskRecord);
 
         Long summaryModelId = Long.valueOf(taskRecord.getTaskConfig().get("summaryModelId").toString());
@@ -555,6 +552,7 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
 
             taskRecord.setResultFilePath("meetings/" + meeting.getId() + "/summaries/" + fileName);
             Map<String, Object> responseData = objectMapper.convertValue(respNode, Map.class);
+            responseData.put("summarySource", summarySource.toSnapshot());
             if (summaryBundle != null) {
                 responseData.put("summaryBundle", summaryBundle);
             }
@@ -572,11 +570,62 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
 
             updateProgress(meeting.getId(), 100, "全流程分析完成", 0);
         } else {
-            updateAiTaskFail(taskRecord, "LLM 总结失败：" + response.body());
+            updateAiTaskFail(taskRecord, "LLM 总结失败: " + response.body());
             throw new RuntimeException("AI总结生成异常");
         }
     }
 
+    private void executeSummaryFlow(Meeting meeting, AiTask sumTask, AiModelVO asrModel) throws Exception {
+        String summaryLockKey = RedisKeys.meetingSummaryLockKey(meeting.getId());
+        Boolean acquired = redisTemplate.opsForValue().setIfAbsent(summaryLockKey, "locked", 30, TimeUnit.MINUTES);
+        if (Boolean.FALSE.equals(acquired)) {
+            log.warn("Meeting {} summary is already being processed", meeting.getId());
+            return;
+        }
+        try {
+            MeetingSummarySource summarySource = meetingTranscriptRevisionService.resolveSummarySource(meeting, sumTask, asrModel);
+            if (summarySource.getText() == null || summarySource.getText().isBlank()) {
+                failPendingSummaryTask(sumTask, "没有转录内容");
+                updateMeetingStatus(meeting.getId(), 4);
+                updateProgress(meeting.getId(), -1, "没有转录内容", 0);
+                return;
+            }
+            processSummaryTask(meeting, summarySource, sumTask);
+        } finally {
+            redisTemplate.delete(summaryLockKey);
+        }
+    }
+
+    private AiTask findLatestTask(Long meetingId, String taskType) {
+        return this.getOne(new LambdaQueryWrapper<AiTask>()
+                .eq(AiTask::getMeetingId, meetingId)
+                .eq(AiTask::getTaskType, taskType)
+                .orderByDesc(AiTask::getId)
+                .last("limit 1"));
+    }
+
+    private boolean canExecuteTask(AiTask task) {
+        return task != null
+                && !Integer.valueOf(2).equals(task.getStatus())
+                && !Integer.valueOf(3).equals(task.getStatus());
+    }
+
+    private AiModelVO resolveAsrModelForRevision(AiTask asrTask) {
+        if (asrTask == null || asrTask.getTaskConfig() == null) {
+            return null;
+        }
+        Object asrModelId = asrTask.getTaskConfig().get("asrModelId");
+        if (asrModelId == null) {
+            return null;
+        }
+        try {
+            return aiModelService.getModelById(Long.parseLong(String.valueOf(asrModelId)), "ASR");
+        } catch (Exception ex) {
+            log.warn("Failed to resolve ASR model for transcript revision, taskId={}", asrTask.getId(), ex);
+            return null;
+        }
+    }
+
     private void updateProgress(Long meetingId, int percent, String msg, int eta) {
         try {
             Map<String, Object> progress = new HashMap<>();

backend/src/main/java/com/imeeting/service/biz/impl/MeetingCommandServiceImpl.java

@@ -25,6 +25,7 @@ import com.imeeting.service.biz.MeetingRuntimeProfileResolver;
 import com.imeeting.service.biz.MeetingService;
 import com.imeeting.service.biz.MeetingSummaryFileService;
 import com.imeeting.service.biz.MeetingTranscriptFileService;
+import com.imeeting.service.biz.MeetingTranscriptRevisionService;
 import com.imeeting.service.biz.RealtimeMeetingSessionStateService;
 import com.imeeting.service.realtime.RealtimeMeetingAudioStorageService;
 import lombok.extern.slf4j.Slf4j;
@@ -55,6 +56,7 @@ public class MeetingCommandServiceImpl implements MeetingCommandService {
     private final com.imeeting.mapper.biz.MeetingTranscriptMapper transcriptMapper;
     private final MeetingSummaryFileService meetingSummaryFileService;
     private final MeetingTranscriptFileService meetingTranscriptFileService;
+    private final MeetingTranscriptRevisionService meetingTranscriptRevisionService;
     private final MeetingDomainSupport meetingDomainSupport;
     private final MeetingRuntimeProfileResolver meetingRuntimeProfileResolver;
     private final RealtimeMeetingSessionStateService realtimeMeetingSessionStateService;
@@ -436,6 +438,17 @@ public class MeetingCommandServiceImpl implements MeetingCommandService {
             throw new RuntimeException("转录内容不能为空");
         }
 
+        MeetingTranscript existing = transcriptMapper.selectOne(new LambdaQueryWrapper<MeetingTranscript>()
+                .eq(MeetingTranscript::getMeetingId, command.getMeetingId())
+                .eq(MeetingTranscript::getId, command.getTranscriptId())
+                .last("LIMIT 1"));
+        if (existing == null) {
+            throw new RuntimeException("转录记录不存在");
+        }
+        if (Objects.equals(normalizeTranscriptContent(existing.getContent()), content)) {
+            return;
+        }
+
         int updated = transcriptMapper.update(null, new LambdaUpdateWrapper<MeetingTranscript>()
                 .eq(MeetingTranscript::getMeetingId, command.getMeetingId())
                 .eq(MeetingTranscript::getId, command.getTranscriptId())
@@ -443,6 +456,7 @@ public class MeetingCommandServiceImpl implements MeetingCommandService {
         if (updated <= 0) {
             throw new RuntimeException("转录记录不存在");
         }
+        meetingTranscriptRevisionService.invalidateCurrentRevision(command.getMeetingId());
     }
 
     @Override
@@ -464,6 +478,10 @@ public class MeetingCommandServiceImpl implements MeetingCommandService {
         return normalized.isEmpty() ? null : normalized;
     }
 
+    private String normalizeTranscriptContent(String content) {
+        return content == null ? "" : content.trim();
+    }
+
     @Override
     @Transactional(rollbackFor = Exception.class)
     public void updateMeetingParticipants(Long meetingId, String participants) {

backend/src/main/java/com/imeeting/service/biz/impl/MeetingTranscriptFileServiceImpl.java

@@ -87,8 +87,8 @@ public class MeetingTranscriptFileServiceImpl implements MeetingTranscriptFileSe
                 .orderByAsc(MeetingTranscript::getStartTime)
                 .orderByAsc(MeetingTranscript::getId));
 
-        String title = firstNonBlank(meetingDetail != null ? meetingDetail.getTitle() : null, meeting.getTitle(), "未命名会议");
-        String hostName = firstNonBlank(meetingDetail != null ? meetingDetail.getHostName() : null, meeting.getHostName(), "未知");
+        String title = firstNonBlank(meetingDetail != null ? meetingDetail.getTitle() : null, meeting.getTitle(), "Untitled Meeting");
+        String hostName = firstNonBlank(meetingDetail != null ? meetingDetail.getHostName() : null, meeting.getHostName(), "Unknown");
         String meetingTime = formatDateTime(meetingDetail != null ? meetingDetail.getMeetingTime() : meeting.getMeetingTime());
 
         StringBuilder builder = new StringBuilder();
@@ -104,13 +104,13 @@ public class MeetingTranscriptFileServiceImpl implements MeetingTranscriptFileSe
         }
 
         for (MeetingTranscript transcript : transcripts) {
-            String speaker = firstNonBlank(transcript.getSpeakerName(), transcript.getSpeakerId(), "未知发言人");
+            String speaker = firstNonBlank(transcript.getSpeakerName(), transcript.getSpeakerId(), "Unknown Speaker");
             builder.append("- ");
             String timeRange = buildTimeRange(transcript.getStartTime(), transcript.getEndTime());
             if (!timeRange.isBlank()) {
                 builder.append(timeRange).append(' ');
             }
-            builder.append(speaker).append("：").append(normalizeTranscriptContent(transcript.getContent())).append("\n");
+            builder.append(speaker).append(": ").append(normalizeTranscriptContent(transcript.getContent())).append("\n");
         }
         return builder.toString();
     }

backend/src/main/java/com/imeeting/service/biz/impl/MeetingTranscriptRevisionServiceImpl.java

@@ -0,0 +1,833 @@
+package com.imeeting.service.biz.impl;
+
+import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
+import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
+import com.fasterxml.jackson.core.JsonProcessingException;
+import com.fasterxml.jackson.core.type.TypeReference;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import com.imeeting.common.SysParamKeys;
+import com.imeeting.dto.biz.AiModelVO;
+import com.imeeting.dto.biz.MeetingSummarySource;
+import com.imeeting.dto.biz.MeetingTranscriptVO;
+import com.imeeting.entity.biz.AiTask;
+import com.imeeting.entity.biz.Meeting;
+import com.imeeting.entity.biz.MeetingTranscript;
+import com.imeeting.entity.biz.MeetingTranscriptRevision;
+import com.imeeting.entity.biz.MeetingTranscriptRevisionItem;
+import com.imeeting.mapper.biz.MeetingTranscriptMapper;
+import com.imeeting.mapper.biz.MeetingTranscriptRevisionItemMapper;
+import com.imeeting.mapper.biz.MeetingTranscriptRevisionMapper;
+import com.imeeting.service.biz.MeetingTranscriptRevisionService;
+import com.unisbase.service.SysParamService;
+import lombok.RequiredArgsConstructor;
+import org.springframework.stereotype.Service;
+import org.springframework.transaction.annotation.Transactional;
+
+import java.nio.charset.StandardCharsets;
+import java.security.MessageDigest;
+import java.math.BigDecimal;
+import java.time.LocalDateTime;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Objects;
+import java.util.TreeMap;
+import java.util.UUID;
+import java.util.regex.Matcher;
+import java.util.regex.Pattern;
+import java.util.stream.Collectors;
+
+@Service
+@RequiredArgsConstructor
+public class MeetingTranscriptRevisionServiceImpl implements MeetingTranscriptRevisionService {
+
+    private static final String RULE_PROFILE_VERSION = "v1";
+    private static final String TRIGGER_TASK_TYPE = "SUMMARY";
+    private static final String SEMANTIC_CORRECTOR = "NONE_V1";
+    private static final String SOURCE_TYPE_REVISION = "REVISION";
+    private static final String SOURCE_TYPE_RAW_FALLBACK = "RAW_FALLBACK";
+    private static final int MERGE_GAP_THRESHOLD_MS = 3000;
+
+    private final MeetingTranscriptMapper transcriptMapper;
+    private final MeetingTranscriptRevisionMapper revisionMapper;
+    private final MeetingTranscriptRevisionItemMapper itemMapper;
+    private final ObjectMapper objectMapper;
+    private final SysParamService sysParamService;
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public String generateOfflineCurrentRevision(Meeting meeting, AiTask task, AiModelVO asrModel) {
+        MeetingTranscriptRevision revision = createCurrentRevision(meeting, task, asrModel);
+        return revision.getCleanedFullText();
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public MeetingSummarySource resolveSummarySource(Meeting meeting, AiTask summaryTask, AiModelVO asrModel) {
+        List<MeetingTranscript> transcripts = loadRawTranscripts(meeting.getId());
+        String rawText = buildRawTranscriptText(transcripts);
+        String fingerprint = buildSourceFingerprint(transcripts);
+        if (transcripts.isEmpty() || rawText.isBlank()) {
+            return buildFallbackSource(rawText, fingerprint);
+        }
+
+        Map<String, Object> ruleProfile = buildRuleProfile(fingerprint, asrModel);
+        String ruleProfileJson = toJson(ruleProfile);
+        MeetingTranscriptRevision current = findCurrentRevision(meeting.getId());
+        if (isReusableCurrentRevision(current, ruleProfileJson)) {
+            return MeetingSummarySource.builder()
+                    .text(current.getCleanedFullText())
+                    .sourceType(SOURCE_TYPE_REVISION)
+                    .revisionId(current.getId())
+                    .fallbackUsed(false)
+                    .sourceFingerprint(fingerprint)
+                    .triggerTaskType(TRIGGER_TASK_TYPE)
+                    .semanticCorrector(SEMANTIC_CORRECTOR)
+                    .ruleProfileVersion(RULE_PROFILE_VERSION)
+                    .build();
+        }
+
+        MeetingTranscriptRevision revision = createCurrentRevision(meeting, summaryTask, asrModel);
+        return MeetingSummarySource.builder()
+                .text(revision.getCleanedFullText())
+                .sourceType(SOURCE_TYPE_REVISION)
+                .revisionId(revision.getId())
+                .fallbackUsed(false)
+                .sourceFingerprint(fingerprint)
+                .triggerTaskType(TRIGGER_TASK_TYPE)
+                .semanticCorrector(SEMANTIC_CORRECTOR)
+                .ruleProfileVersion(RULE_PROFILE_VERSION)
+                .build();
+    }
+
+    @Override
+    public List<MeetingTranscriptVO> listEffectiveTranscripts(Long meetingId) {
+        List<MeetingTranscript> transcripts = loadRawTranscripts(meetingId);
+        MeetingTranscriptRevision current = findCurrentRevision(meetingId);
+        Map<Long, MeetingTranscriptRevisionItem> itemByTranscriptId = new LinkedHashMap<>();
+        if (current != null) {
+            itemByTranscriptId = itemMapper.selectList(new LambdaQueryWrapper<MeetingTranscriptRevisionItem>()
+                            .eq(MeetingTranscriptRevisionItem::getRevisionId, current.getId())
+                            .orderByAsc(MeetingTranscriptRevisionItem::getSourceSortOrder)
+                            .orderByAsc(MeetingTranscriptRevisionItem::getId))
+                    .stream()
+                    .collect(Collectors.toMap(
+                            MeetingTranscriptRevisionItem::getSourceTranscriptId,
+                            item -> item,
+                            (left, right) -> right.getCleanedContent() != null && !right.getCleanedContent().isBlank() ? right : left,
+                            LinkedHashMap::new
+                    ));
+        }
+        List<MeetingTranscriptVO> result = new ArrayList<>();
+        for (MeetingTranscript transcript : transcripts) {
+            MeetingTranscriptRevisionItem item = itemByTranscriptId.get(transcript.getId());
+            if (item != null && isSuppressedAction(item.getActionType())) {
+                continue;
+            }
+            MeetingTranscriptVO vo = new MeetingTranscriptVO();
+            vo.setId(transcript.getId());
+            vo.setSpeakerId(transcript.getSpeakerId());
+            vo.setSpeakerName(transcript.getSpeakerName());
+            vo.setSpeakerLabel(transcript.getSpeakerLabel());
+            vo.setStartTime(transcript.getStartTime());
+            vo.setEndTime(transcript.getEndTime());
+            vo.setContent(resolveEffectiveContent(transcript, item));
+            result.add(vo);
+        }
+        return result;
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public boolean updateCurrentRevisionContent(Long meetingId, Long operatorId, String content) {
+        MeetingTranscriptRevision current = findCurrentRevision(meetingId);
+        if (current == null) {
+            return false;
+        }
+        current.setCleanedFullText(content);
+        revisionMapper.updateById(current);
+        return true;
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void invalidateCurrentRevision(Long meetingId) {
+        revisionMapper.update(null, new LambdaUpdateWrapper<MeetingTranscriptRevision>()
+                .eq(MeetingTranscriptRevision::getMeetingId, meetingId)
+                .eq(MeetingTranscriptRevision::getIsCurrent, 1)
+                .set(MeetingTranscriptRevision::getIsCurrent, 0));
+    }
+
+    @Transactional(rollbackFor = Exception.class)
+    protected MeetingTranscriptRevision createCurrentRevision(Meeting meeting, AiTask task, AiModelVO asrModel) {
+        List<MeetingTranscript> transcripts = loadRawTranscripts(meeting.getId());
+        String fingerprint = buildSourceFingerprint(transcripts);
+        Map<String, Object> ruleProfile = buildRuleProfile(fingerprint, asrModel);
+        CleaningResult cleaningResult = cleanTranscripts(transcripts, ruleProfile);
+        MeetingTranscriptRevision current = findCurrentRevision(meeting.getId());
+        int nextRevisionNo = resolveNextRevisionNo(meeting.getId(), current);
+
+        MeetingTranscriptRevision draft = new MeetingTranscriptRevision();
+        draft.setMeetingId(meeting.getId());
+        draft.setSourceTaskId(task != null ? task.getId() : null);
+        draft.setRevisionNo(nextRevisionNo);
+        draft.setStatus(1);
+        draft.setCleanedFullText(cleaningResult.fullText());
+        draft.setRuleProfile(toJson(ruleProfile));
+        draft.setSegmentCount(transcripts.size());
+        draft.setDroppedSegmentCount(cleaningResult.droppedCount());
+        draft.setMergedGroupCount(cleaningResult.mergedGroupCount());
+        draft.setIsCurrent(0);
+        revisionMapper.insert(draft);
+
+        if (current != null) {
+            revisionMapper.update(null, new LambdaUpdateWrapper<MeetingTranscriptRevision>()
+                    .eq(MeetingTranscriptRevision::getMeetingId, meeting.getId())
+                    .eq(MeetingTranscriptRevision::getIsCurrent, 1)
+                    .set(MeetingTranscriptRevision::getIsCurrent, 0));
+        }
+
+        MeetingTranscriptRevision finalRevision = new MeetingTranscriptRevision();
+        finalRevision.setMeetingId(meeting.getId());
+        finalRevision.setSourceTaskId(task != null ? task.getId() : null);
+        finalRevision.setRevisionNo(nextRevisionNo + 1);
+        finalRevision.setStatus(2);
+        finalRevision.setCleanedFullText(cleaningResult.fullText());
+        finalRevision.setRuleProfile(toJson(ruleProfile));
+        finalRevision.setSegmentCount(transcripts.size());
+        finalRevision.setDroppedSegmentCount(cleaningResult.droppedCount());
+        finalRevision.setMergedGroupCount(cleaningResult.mergedGroupCount());
+        finalRevision.setIsCurrent(1);
+        revisionMapper.insert(finalRevision);
+
+        for (MeetingTranscriptRevisionItem item : cleaningResult.items()) {
+            item.setRevisionId(finalRevision.getId());
+            itemMapper.insert(item);
+        }
+        return finalRevision;
+    }
+
+    private List<MeetingTranscript> loadRawTranscripts(Long meetingId) {
+        return transcriptMapper.selectList(new LambdaQueryWrapper<MeetingTranscript>()
+                .eq(MeetingTranscript::getMeetingId, meetingId)
+                .orderByAsc(MeetingTranscript::getSortOrder)
+                .orderByAsc(MeetingTranscript::getStartTime)
+                .orderByAsc(MeetingTranscript::getId));
+    }
+
+    private MeetingTranscriptRevision findCurrentRevision(Long meetingId) {
+        return revisionMapper.selectOne(new LambdaQueryWrapper<MeetingTranscriptRevision>()
+                .eq(MeetingTranscriptRevision::getMeetingId, meetingId)
+                .eq(MeetingTranscriptRevision::getIsCurrent, 1)
+                .orderByDesc(MeetingTranscriptRevision::getRevisionNo)
+                .last("limit 1"));
+    }
+
+    private int resolveNextRevisionNo(Long meetingId, MeetingTranscriptRevision current) {
+        if (current != null && current.getRevisionNo() != null) {
+            return current.getRevisionNo() + 1;
+        }
+        MeetingTranscriptRevision latest = revisionMapper.selectOne(new LambdaQueryWrapper<MeetingTranscriptRevision>()
+                .eq(MeetingTranscriptRevision::getMeetingId, meetingId)
+                .orderByDesc(MeetingTranscriptRevision::getRevisionNo)
+                .last("limit 1"));
+        return latest == null || latest.getRevisionNo() == null ? 1 : latest.getRevisionNo() + 1;
+    }
+
+    private boolean isReusableCurrentRevision(MeetingTranscriptRevision current, String expectedRuleProfile) {
+        return current != null
+                && Integer.valueOf(1).equals(current.getIsCurrent())
+                && Integer.valueOf(2).equals(current.getStatus())
+                && current.getCleanedFullText() != null
+                && !current.getCleanedFullText().isBlank()
+                && Objects.equals(expectedRuleProfile, current.getRuleProfile());
+    }
+
+    private Map<String, Object> buildRuleProfile(String fingerprint, AiModelVO asrModel) {
+        Map<String, Object> profile = new LinkedHashMap<>();
+        profile.put("ruleProfileVersion", RULE_PROFILE_VERSION);
+        profile.put("sourceFingerprint", fingerprint);
+        profile.put("triggerTaskType", TRIGGER_TASK_TYPE);
+        profile.put("semanticCorrector", SEMANTIC_CORRECTOR);
+        profile.put("transcriptRuleFillerWords", resolveFillerWords(asrModel));
+        profile.put("transcriptRuleReplacements", resolveReplacementRules(asrModel));
+        return profile;
+    }
+
+    private List<String> resolveFillerWords(AiModelVO asrModel) {
+        List<String> configured = parseCleanupWords(sysParamService.getCachedParamValue(
+                SysParamKeys.MEETING_TRANSCRIPT_CLEANUP_FILLER_WORDS,
+                ""
+        ));
+        if (!configured.isEmpty()) {
+            return configured;
+        }
+        if (asrModel == null || asrModel.getMediaConfig() == null) {
+            return List.of();
+        }
+        Object raw = asrModel.getMediaConfig().get("transcriptRuleFillerWords");
+        if (!(raw instanceof List<?> list)) {
+            return List.of();
+        }
+        return list.stream()
+                .filter(Objects::nonNull)
+                .map(String::valueOf)
+                .map(String::trim)
+                .filter(value -> !value.isBlank())
+                .distinct()
+                .sorted()
+                .toList();
+    }
+
+    private Map<String, String> resolveReplacementRules(AiModelVO asrModel) {
+        Map<String, String> configured = parseCleanupReplacements(sysParamService.getCachedParamValue(
+                SysParamKeys.MEETING_TRANSCRIPT_CLEANUP_REPLACEMENTS,
+                ""
+        ));
+        if (!configured.isEmpty()) {
+            return configured;
+        }
+        if (asrModel == null || asrModel.getMediaConfig() == null) {
+            return Map.of();
+        }
+        Object raw = asrModel.getMediaConfig().get("transcriptRuleReplacements");
+        if (!(raw instanceof Map<?, ?> map)) {
+            return Map.of();
+        }
+        Map<String, String> result = new TreeMap<>();
+        for (Map.Entry<?, ?> entry : map.entrySet()) {
+            if (entry.getKey() == null || entry.getValue() == null) {
+                continue;
+            }
+            String key = String.valueOf(entry.getKey()).trim();
+            String value = String.valueOf(entry.getValue()).trim();
+            if (!key.isBlank() && !value.isBlank()) {
+                result.put(key, value);
+            }
+        }
+        return result;
+    }
+
+    private List<String> parseCleanupWords(String raw) {
+        if (raw == null || raw.isBlank()) {
+            return List.of();
+        }
+        String normalized = raw.trim();
+        try {
+            if (normalized.startsWith("[")) {
+                List<?> parsed = objectMapper.readValue(normalized, new TypeReference<List<?>>() {});
+                return parsed.stream()
+                        .filter(Objects::nonNull)
+                        .map(String::valueOf)
+                        .map(String::trim)
+                        .filter(value -> !value.isBlank())
+                        .distinct()
+                        .sorted()
+                        .toList();
+            }
+        } catch (Exception ignored) {
+            // Fall back to plain-text parsing.
+        }
+        return java.util.Arrays.stream(normalized.split("[,，;；\\r\\n]+"))
+                .map(String::trim)
+                .filter(value -> !value.isBlank())
+                .distinct()
+                .sorted()
+                .toList();
+    }
+
+    private Map<String, String> parseCleanupReplacements(String raw) {
+        if (raw == null || raw.isBlank()) {
+            return Map.of();
+        }
+        String normalized = raw.trim();
+        try {
+            if (normalized.startsWith("{")) {
+                Map<String, Object> parsed = objectMapper.readValue(normalized, new TypeReference<Map<String, Object>>() {});
+                Map<String, String> result = new TreeMap<>();
+                for (Map.Entry<String, Object> entry : parsed.entrySet()) {
+                    if (entry.getKey() == null || entry.getValue() == null) {
+                        continue;
+                    }
+                    String key = entry.getKey().trim();
+                    String value = String.valueOf(entry.getValue()).trim();
+                    if (!key.isBlank() && !value.isBlank()) {
+                        result.put(key, value);
+                    }
+                }
+                return result;
+            }
+        } catch (Exception ignored) {
+            // Fall back to line-based parsing.
+        }
+
+        Map<String, String> result = new TreeMap<>();
+        for (String line : normalized.split("[\\r\\n]+")) {
+            String candidate = line == null ? "" : line.trim();
+            if (candidate.isBlank()) {
+                continue;
+            }
+            String[] separatorCandidates = {"=>", "=", ":"};
+            for (String separator : separatorCandidates) {
+                int index = candidate.indexOf(separator);
+                if (index <= 0 || index >= candidate.length() - separator.length()) {
+                    continue;
+                }
+                String key = candidate.substring(0, index).trim();
+                String value = candidate.substring(index + separator.length()).trim();
+                if (!key.isBlank() && !value.isBlank()) {
+                    result.put(key, value);
+                }
+                break;
+            }
+        }
+        return result;
+    }
+
+    private CleaningResult cleanTranscripts(List<MeetingTranscript> transcripts, Map<String, Object> ruleProfile) {
+        @SuppressWarnings("unchecked")
+        List<String> fillerWords = (List<String>) ruleProfile.getOrDefault("transcriptRuleFillerWords", List.of());
+        @SuppressWarnings("unchecked")
+        Map<String, String> replacements = (Map<String, String>) ruleProfile.getOrDefault("transcriptRuleReplacements", Map.of());
+
+        List<String> orderedFillerWords = fillerWords.stream()
+                .filter(Objects::nonNull)
+                .map(String::trim)
+                .filter(value -> !value.isBlank())
+                .distinct()
+                .sorted(Comparator.comparingInt(String::length).reversed().thenComparing(String::compareTo))
+                .toList();
+        List<Map.Entry<String, String>> orderedReplacementRules = replacements.entrySet().stream()
+                .filter(entry -> entry.getKey() != null && entry.getValue() != null)
+                .sorted(Comparator.<Map.Entry<String, String>>comparingInt(entry -> entry.getKey().length()).reversed()
+                        .thenComparing(Map.Entry::getKey))
+                .toList();
+
+        List<MeetingTranscriptRevisionItem> items = new ArrayList<>();
+        List<SegmentGroupState> groups = new ArrayList<>();
+        MeetingTranscript previousTranscript = null;
+        SegmentGroupState currentGroup = null;
+        int droppedCount = 0;
+
+        for (int index = 0; index < transcripts.size(); index++) {
+            MeetingTranscript transcript = transcripts.get(index);
+            String normalizedSource = normalizeRawContent(transcript.getContent());
+            MeetingTranscriptRevisionItem item = createRevisionItem(transcript, transcripts, index);
+
+            if (previousTranscript != null
+                    && currentGroup != null
+                    && isAdjacentDuplicate(previousTranscript, transcript)) {
+                item.setActionType("DROP_DUPLICATE");
+                item.setCleanedContent("");
+                item.setCleanedSpeakerName(transcript.getSpeakerName());
+                item.setMergeGroupId(currentGroup.getGroupId());
+                item.setConfidence(confidenceForAction("DROP_DUPLICATE"));
+                item.setRuleHits(toJson(buildRuleHits(List.of(), List.of())));
+                items.add(item);
+                droppedCount++;
+                previousTranscript = transcript;
+                continue;
+            }
+
+            if (currentGroup != null && shouldMergeIntoCurrentGroup(currentGroup, transcript)) {
+                currentGroup.appendSourceContent(normalizedSource, transcript);
+                currentGroup.incrementSourceSegmentCount();
+                currentGroup.getRepresentativeItem().setMergeGroupId(currentGroup.getGroupId());
+                item.setActionType("MERGE_INTO_PREV");
+                item.setCleanedContent("");
+                item.setCleanedSpeakerName(transcript.getSpeakerName());
+                item.setMergeGroupId(currentGroup.getGroupId());
+                item.setConfidence(confidenceForAction("MERGE_INTO_PREV"));
+                item.setRuleHits(toJson(buildRuleHits(List.of(), List.of())));
+                items.add(item);
+                previousTranscript = transcript;
+                continue;
+            }
+
+            item.setMergeGroupId("");
+            items.add(item);
+            currentGroup = SegmentGroupState.start("merge-" + transcript.getId() + "-" + UUID.randomUUID(), transcript, item, normalizedSource);
+            groups.add(currentGroup);
+            previousTranscript = transcript;
+        }
+
+        List<String> finalLines = new ArrayList<>();
+        int mergedGroupCount = 0;
+        for (SegmentGroupState group : groups) {
+            if (group.getSourceSegmentCount() > 1) {
+                mergedGroupCount++;
+                group.getRepresentativeItem().setMergeGroupId(group.getGroupId());
+            }
+
+            TextCleanupResult cleanupResult = cleanTranscriptContent(group.getMergedSourceContent(), orderedFillerWords, orderedReplacementRules);
+            String cleaned = cleanupResult.cleanedContent();
+            MeetingTranscriptRevisionItem representativeItem = group.getRepresentativeItem();
+            representativeItem.setCleanedContent(cleaned);
+            representativeItem.setCleanedSpeakerName(group.getRepresentativeTranscript().getSpeakerName());
+            representativeItem.setActionType(resolveActionType(group.getMergedSourceContent(), cleaned,
+                    cleanupResult.matchedFillerWords(), cleanupResult.matchedReplacementRules()));
+            representativeItem.setRuleHits(toJson(buildRuleHits(
+                    cleanupResult.matchedFillerWords(),
+                    cleanupResult.matchedReplacementRules()
+            )));
+            representativeItem.setConfidence(confidenceForAction(representativeItem.getActionType()));
+
+            if (cleaned.isBlank()) {
+                droppedCount += group.getSourceSegmentCount();
+                continue;
+            }
+            finalLines.add(formatTranscriptLine(group.getRepresentativeTranscript(), cleaned));
+        }
+
+        String fullText = finalLines.stream()
+                .filter(line -> line != null && !line.isBlank())
+                .collect(Collectors.joining("\n"));
+        return new CleaningResult(fullText, items, droppedCount, mergedGroupCount);
+    }
+
+    private String buildSourceFingerprint(List<MeetingTranscript> transcripts) {
+        String raw = transcripts.stream()
+                .sorted(Comparator.comparing(MeetingTranscript::getSortOrder, Comparator.nullsLast(Integer::compareTo))
+                        .thenComparing(MeetingTranscript::getStartTime, Comparator.nullsLast(Integer::compareTo))
+                        .thenComparing(MeetingTranscript::getId, Comparator.nullsLast(Long::compareTo)))
+                .map(transcript -> String.join("|",
+                        String.valueOf(transcript.getId()),
+                        nullSafe(transcript.getSpeakerId()),
+                        nullSafe(transcript.getSpeakerName()),
+                        nullSafe(transcript.getContent()),
+                        String.valueOf(transcript.getStartTime()),
+                        String.valueOf(transcript.getEndTime()),
+                        String.valueOf(transcript.getSortOrder())))
+                .collect(Collectors.joining("\n"));
+        try {
+            MessageDigest digest = MessageDigest.getInstance("SHA-256");
+            byte[] hashed = digest.digest(raw.getBytes(StandardCharsets.UTF_8));
+            StringBuilder builder = new StringBuilder();
+            for (byte b : hashed) {
+                builder.append(String.format("%02x", b));
+            }
+            return builder.toString();
+        } catch (Exception ex) {
+            throw new RuntimeException("计算转录内容失败", ex);
+        }
+    }
+
+    private String buildRawTranscriptText(List<MeetingTranscript> transcripts) {
+        return transcripts.stream()
+                .filter(Objects::nonNull)
+                .map(transcript -> formatTranscriptLine(transcript, normalizeRawContent(transcript.getContent())))
+                .filter(line -> line != null && !line.isBlank())
+                .collect(Collectors.joining("\n"));
+    }
+
+    private MeetingSummarySource buildFallbackSource(String rawText, String fingerprint) {
+        return MeetingSummarySource.builder()
+                .text(rawText)
+                .sourceType(SOURCE_TYPE_RAW_FALLBACK)
+                .revisionId(null)
+                .fallbackUsed(true)
+                .sourceFingerprint(fingerprint)
+                .triggerTaskType(TRIGGER_TASK_TYPE)
+                .semanticCorrector(SEMANTIC_CORRECTOR)
+                .ruleProfileVersion(RULE_PROFILE_VERSION)
+                .build();
+    }
+
+    private String formatTranscriptLine(MeetingTranscript transcript, String content) {
+        if (content == null || content.isBlank()) {
+            return null;
+        }
+        String speaker = transcript.getSpeakerName();
+        if (speaker == null || speaker.isBlank()) {
+            speaker = transcript.getSpeakerId();
+        }
+        if (speaker == null || speaker.isBlank()) {
+            return content.trim();
+        }
+        return speaker.trim() + ": " + content.trim();
+    }
+
+    private String resolveEffectiveContent(MeetingTranscript transcript, MeetingTranscriptRevisionItem item) {
+        if (item == null) {
+            return transcript.getContent();
+        }
+        if (item.getCleanedContent() != null) {
+            return item.getCleanedContent();
+        }
+        return transcript.getContent();
+    }
+
+    private boolean isSuppressedAction(String actionType) {
+        return "MERGE_INTO_PREV".equals(actionType)
+                || "DROP_FILLER".equals(actionType)
+                || "DROP_DUPLICATE".equals(actionType);
+    }
+
+    private String resolveActionType(String normalizedSource,
+                                     String cleaned,
+                                     List<String> matchedFillerWords,
+                                     List<String> matchedReplacementRules) {
+        if (cleaned == null || cleaned.isBlank()) {
+            return !matchedFillerWords.isEmpty() ? "DROP_FILLER" : "EDIT";
+        }
+        if (Objects.equals(cleaned, normalizedSource)) {
+            return "KEEP";
+        }
+        return "RULE_REPLACED";
+    }
+
+    private Map<String, Object> buildRuleHits(List<String> matchedFillerWords, List<String> matchedReplacementRules) {
+        Map<String, Object> ruleHits = new LinkedHashMap<>();
+        ruleHits.put("fillerWords", matchedFillerWords == null ? List.of() : List.copyOf(matchedFillerWords));
+        ruleHits.put("replacements", matchedReplacementRules == null ? List.of() : List.copyOf(matchedReplacementRules));
+        return ruleHits;
+    }
+
+    private MeetingTranscriptRevisionItem createRevisionItem(MeetingTranscript transcript,
+                                                             List<MeetingTranscript> transcripts,
+                                                             int index) {
+        MeetingTranscriptRevisionItem item = new MeetingTranscriptRevisionItem();
+        item.setSourceTranscriptId(transcript.getId());
+        item.setSourceSortOrder(transcript.getSortOrder());
+        item.setSourceSpeakerId(transcript.getSpeakerId());
+        item.setSourceSpeakerName(transcript.getSpeakerName());
+        item.setSourceContent(transcript.getContent());
+        item.setCleanedSpeakerName(transcript.getSpeakerName());
+        item.setContextSnapshot(buildContextSnapshot(transcripts, index));
+        return item;
+    }
+
+    private String buildContextSnapshot(List<MeetingTranscript> transcripts, int index) {
+        Map<String, Object> snapshot = new LinkedHashMap<>();
+        snapshot.put("prevContent", index > 0 ? normalizeRawContent(transcripts.get(index - 1).getContent()) : "");
+        snapshot.put("currentContent", normalizeRawContent(transcripts.get(index).getContent()));
+        snapshot.put("nextContent", index + 1 < transcripts.size()
+                ? normalizeRawContent(transcripts.get(index + 1).getContent())
+                : "");
+        return toJson(snapshot);
+    }
+
+    private boolean isAdjacentDuplicate(MeetingTranscript previousTranscript, MeetingTranscript currentTranscript) {
+        return sameSpeaker(previousTranscript, currentTranscript)
+                && !normalizeComparisonText(previousTranscript.getContent()).isBlank()
+                && Objects.equals(
+                normalizeComparisonText(previousTranscript.getContent()),
+                normalizeComparisonText(currentTranscript.getContent())
+        );
+    }
+
+    private boolean shouldMergeIntoCurrentGroup(SegmentGroupState currentGroup, MeetingTranscript currentTranscript) {
+        if (currentGroup == null || currentTranscript == null) {
+            return false;
+        }
+        if (!sameSpeaker(currentGroup.getLastTranscript(), currentTranscript)) {
+            return false;
+        }
+        Integer previousEndTime = currentGroup.getLastTranscript().getEndTime();
+        Integer currentStartTime = currentTranscript.getStartTime();
+        if (previousEndTime == null || currentStartTime == null) {
+            return false;
+        }
+        int gap = currentStartTime - previousEndTime;
+        return gap >= 0 && gap <= MERGE_GAP_THRESHOLD_MS;
+    }
+
+    private boolean sameSpeaker(MeetingTranscript left, MeetingTranscript right) {
+        if (left == null || right == null) {
+            return false;
+        }
+        return Objects.equals(nullSafe(left.getSpeakerId()), nullSafe(right.getSpeakerId()))
+                && Objects.equals(nullSafe(left.getSpeakerName()), nullSafe(right.getSpeakerName()));
+    }
+
+    private String normalizeComparisonText(String content) {
+        return normalizeRawContent(content)
+                .replaceAll("\\s+", "")
+                .replaceAll("[，。？！；：、,.!?;:]", "");
+    }
+
+    private TextCleanupResult cleanTranscriptContent(String content,
+                                                     List<String> fillerWords,
+                                                     List<Map.Entry<String, String>> replacementRules) {
+        String cleaned = normalizeRawContent(content);
+        List<String> matchedFillerWords = new ArrayList<>();
+        List<String> matchedReplacementRules = new ArrayList<>();
+        for (String fillerWord : fillerWords) {
+            String updated = removeFillerWord(cleaned, fillerWord);
+            if (!Objects.equals(updated, cleaned)) {
+                matchedFillerWords.add(fillerWord);
+                cleaned = updated;
+            }
+        }
+        for (Map.Entry<String, String> entry : replacementRules) {
+            if (cleaned.contains(entry.getKey())) {
+                matchedReplacementRules.add(entry.getKey() + "->" + entry.getValue());
+                cleaned = cleaned.replace(entry.getKey(), entry.getValue());
+            }
+        }
+        cleaned = normalizeCleanedContent(cleaned);
+        return new TextCleanupResult(cleaned, matchedFillerWords, matchedReplacementRules);
+    }
+
+    private String removeFillerWord(String content, String fillerWord) {
+        if (content == null || content.isBlank() || fillerWord == null || fillerWord.isBlank()) {
+            return content == null ? "" : content;
+        }
+        Pattern pattern = Pattern.compile("(^|[\\s，。？！；：、,.!?;:])(" + Pattern.quote(fillerWord) + ")(?=($|[\\s，。？！；：、,.!?;:]))");
+        Matcher matcher = pattern.matcher(content);
+        StringBuffer buffer = new StringBuffer();
+        boolean changed = false;
+        while (matcher.find()) {
+            String prefix = matcher.group(1);
+            matcher.appendReplacement(buffer, Matcher.quoteReplacement(prefix == null ? "" : prefix));
+            changed = true;
+        }
+        matcher.appendTail(buffer);
+        return changed ? buffer.toString() : content;
+    }
+
+    private String normalizeRawContent(String content) {
+        if (content == null || content.isBlank()) {
+            return "";
+        }
+        return content.trim();
+    }
+
+    private String normalizeCleanedContent(String content) {
+        if (content == null || content.isBlank()) {
+            return "";
+        }
+        String normalized = content.replace("\r\n", " ")
+                .replace("\n", " ")
+                .replaceAll("\\s+", " ")
+                .trim();
+        normalized = normalized.replaceAll("\\s*([，。？！；：、,.!?;:])\\s*", "$1");
+        normalized = normalized.replaceAll("([，。？！；：、,.!?;:])[，。？！；：、,.!?;:]+", "$1");
+        normalized = normalized.replaceAll("^[，。？！；：、,.!?;:]+", "");
+        normalized = normalized.replaceAll("\\(\\s+\\)", "()");
+        return normalized.trim();
+    }
+
+    private BigDecimal confidenceForAction(String actionType) {
+        return switch (actionType) {
+            case "DROP_DUPLICATE" -> BigDecimal.valueOf(0.98D);
+            case "DROP_FILLER" -> BigDecimal.valueOf(0.95D);
+            case "MERGE_INTO_PREV" -> BigDecimal.valueOf(0.90D);
+            case "RULE_REPLACED" -> BigDecimal.valueOf(0.88D);
+            default -> BigDecimal.ONE;
+        };
+    }
+
+    private String nullSafe(String value) {
+        return value == null ? "" : value;
+    }
+
+    private String toJson(Object value) {
+        try {
+            return objectMapper.writeValueAsString(value);
+        } catch (JsonProcessingException ex) {
+            throw new RuntimeException("序列化修正版元数据失败", ex);
+        }
+    }
+
+    @SuppressWarnings("unused")
+    private Map<String, Object> readRuleProfile(String raw) {
+        if (raw == null || raw.isBlank()) {
+            return Map.of();
+        }
+        try {
+            return objectMapper.readValue(raw, new TypeReference<>() {});
+        } catch (JsonProcessingException ex) {
+            return Map.of();
+        }
+    }
+
+    private record CleaningResult(String fullText,
+                                  List<MeetingTranscriptRevisionItem> items,
+                                  int droppedCount,
+                                  int mergedGroupCount) {
+    }
+
+    private record TextCleanupResult(String cleanedContent,
+                                     List<String> matchedFillerWords,
+                                     List<String> matchedReplacementRules) {
+    }
+
+    private static final class SegmentGroupState {
+        private final String groupId;
+        private final MeetingTranscript representativeTranscript;
+        private final MeetingTranscriptRevisionItem representativeItem;
+        private MeetingTranscript lastTranscript;
+        private String mergedSourceContent;
+        private int sourceSegmentCount;
+
+        private SegmentGroupState(String groupId,
+                                  MeetingTranscript representativeTranscript,
+                                  MeetingTranscriptRevisionItem representativeItem,
+                                  String mergedSourceContent) {
+            this.groupId = groupId;
+            this.representativeTranscript = representativeTranscript;
+            this.representativeItem = representativeItem;
+            this.lastTranscript = representativeTranscript;
+            this.mergedSourceContent = mergedSourceContent;
+            this.sourceSegmentCount = 1;
+        }
+
+        private static SegmentGroupState start(String groupId,
+                                               MeetingTranscript representativeTranscript,
+                                               MeetingTranscriptRevisionItem representativeItem,
+                                               String mergedSourceContent) {
+            return new SegmentGroupState(groupId, representativeTranscript, representativeItem, mergedSourceContent);
+        }
+
+        private void appendSourceContent(String nextContent, MeetingTranscript transcript) {
+            this.mergedSourceContent = joinTranscriptContent(this.mergedSourceContent, nextContent);
+            this.lastTranscript = transcript;
+        }
+
+        private void incrementSourceSegmentCount() {
+            this.sourceSegmentCount++;
+        }
+
+        private static String joinTranscriptContent(String left, String right) {
+            if (left == null || left.isBlank()) {
+                return right == null ? "" : right;
+            }
+            if (right == null || right.isBlank()) {
+                return left;
+            }
+            boolean leftAsciiTail = Character.isLetterOrDigit(left.charAt(left.length() - 1));
+            boolean rightAsciiHead = Character.isLetterOrDigit(right.charAt(0));
+            if (leftAsciiTail && rightAsciiHead) {
+                return left + " " + right;
+            }
+            return left + right;
+        }
+
+        private String getGroupId() {
+            return groupId;
+        }
+
+        private MeetingTranscript getRepresentativeTranscript() {
+            return representativeTranscript;
+        }
+
+        private MeetingTranscriptRevisionItem getRepresentativeItem() {
+            return representativeItem;
+        }
+
+        private MeetingTranscript getLastTranscript() {
+            return lastTranscript;
+        }
+
+        private String getMergedSourceContent() {
+            return mergedSourceContent;
+        }
+
+        private int getSourceSegmentCount() {
+            return sourceSegmentCount;
+        }
+    }
+}

backend/src/test/java/com/imeeting/controller/android/legacy/LegacyMeetingControllerTest.java

@@ -614,4 +614,47 @@ class LegacyMeetingControllerTest {
         assertEquals(75, data.getProcessingStatus().getOverallProgress());
         assertEquals("summary_generation", data.getProcessingStatus().getCurrentStage());
     }
+
+    @Test
+    void previewDataShouldTreatHundredPercentProgressAsCompletedWhenSummaryAlreadyReadable() {
+        MeetingService meetingService = mock(MeetingService.class);
+        AiTaskService aiTaskService = mock(AiTaskService.class);
+        MeetingQueryService meetingQueryService = mock(MeetingQueryService.class);
+        MeetingTranscriptMapper transcriptMapper = mock(MeetingTranscriptMapper.class);
+        StringRedisTemplate redisTemplate = mock(StringRedisTemplate.class);
+        @SuppressWarnings("unchecked")
+        ValueOperations<String, String> valueOperations = mock(ValueOperations.class);
+
+        Meeting meeting = new Meeting();
+        meeting.setId(28L);
+        meeting.setTitle("completed by progress");
+        meeting.setStatus(0);
+        when(meetingService.getById(28L)).thenReturn(meeting);
+        when(aiTaskService.getOne(any())).thenReturn((AiTask) null, (AiTask) null);
+        when(redisTemplate.opsForValue()).thenReturn(valueOperations);
+        when(valueOperations.get("biz:meeting:progress:28")).thenReturn("{\"percent\":100}");
+
+        MeetingVO detail = new MeetingVO();
+        detail.setSummaryContent("done");
+        when(meetingQueryService.getDetail(28L)).thenReturn(detail);
+
+        LegacyMeetingController controller = new LegacyMeetingController(
+                mock(LegacyMeetingAdapterService.class),
+                meetingQueryService,
+                mock(MeetingAccessService.class),
+                mock(MeetingCommandService.class),
+                meetingService,
+                aiTaskService,
+                mock(PromptTemplateService.class),
+                transcriptMapper,
+                mock(SysUserMapper.class),
+                redisTemplate,
+                new ObjectMapper()
+        );
+
+        LegacyApiResponse<?> response = controller.previewData(28L);
+
+        assertEquals("200", response.getCode());
+        assertNotNull(response.getData());
+    }
 }

backend/src/test/java/com/imeeting/service/biz/impl/AiTaskServiceImplTest.java

@@ -1,6 +1,7 @@
 package com.imeeting.service.biz.impl;
 
 import com.fasterxml.jackson.databind.ObjectMapper;
+import com.imeeting.dto.biz.MeetingSummarySource;
 import com.imeeting.entity.biz.AiTask;
 import com.imeeting.entity.biz.Meeting;
 import com.imeeting.entity.biz.MeetingTranscript;
@@ -10,6 +11,7 @@ import com.imeeting.service.biz.AiModelService;
 import com.imeeting.service.biz.HotWordService;
 import com.imeeting.service.biz.MeetingSummaryFileService;
 import com.imeeting.service.biz.MeetingTranscriptFileService;
+import com.imeeting.service.biz.MeetingTranscriptRevisionService;
 import com.imeeting.support.TaskSecurityContextRunner;
 import com.unisbase.mapper.SysUserMapper;
 import org.junit.jupiter.api.Test;
@@ -22,15 +24,15 @@ import java.util.List;
 
 import static org.junit.jupiter.api.Assertions.assertEquals;
 import static org.junit.jupiter.api.Assertions.assertTrue;
-import static org.mockito.ArgumentMatchers.eq;
 import static org.mockito.ArgumentMatchers.any;
 import static org.mockito.ArgumentMatchers.anyLong;
 import static org.mockito.ArgumentMatchers.anyString;
-import static org.mockito.Mockito.times;
+import static org.mockito.ArgumentMatchers.eq;
 import static org.mockito.Mockito.doReturn;
 import static org.mockito.Mockito.mock;
 import static org.mockito.Mockito.never;
 import static org.mockito.Mockito.spy;
+import static org.mockito.Mockito.times;
 import static org.mockito.Mockito.verify;
 import static org.mockito.Mockito.when;
 
@@ -94,7 +96,8 @@ class AiTaskServiceImplTest {
                 aiModelService,
                 redisTemplate,
                 new TaskSecurityContextRunner(),
-                mock(MeetingTranscriptFileService.class)
+                mock(MeetingTranscriptFileService.class),
+                mock(MeetingTranscriptRevisionService.class)
         ));
         doReturn(true).when(service).updateById(any());
 
@@ -118,7 +121,7 @@ class AiTaskServiceImplTest {
         service.dispatchTasks(66L, 1L, 2L);
 
         assertEquals(3, summaryTask.getStatus());
-        assertTrue(summaryTask.getErrorMsg().contains("没有可用于总结的转录内容"));
+        assertTrue(summaryTask.getErrorMsg() != null && !summaryTask.getErrorMsg().isBlank());
         verify(aiModelService, never()).getModelById(anyLong(), anyString());
     }
 
@@ -128,9 +131,19 @@ class AiTaskServiceImplTest {
         MeetingTranscriptMapper transcriptMapper = mock(MeetingTranscriptMapper.class);
         AiModelService aiModelService = mock(AiModelService.class);
         StringRedisTemplate redisTemplate = mock(StringRedisTemplate.class);
+        MeetingTranscriptRevisionService revisionService = mock(MeetingTranscriptRevisionService.class);
         @SuppressWarnings("unchecked")
         ValueOperations<String, String> valueOperations = mock(ValueOperations.class);
         when(redisTemplate.opsForValue()).thenReturn(valueOperations);
+        when(valueOperations.setIfAbsent(anyString(), anyString(), anyLong(), any())).thenReturn(true);
+        when(revisionService.resolveSummarySource(any(), any(), any())).thenReturn(MeetingSummarySource.builder()
+                .text(" ")
+                .sourceType("RAW_FALLBACK")
+                .fallbackUsed(true)
+                .triggerTaskType("SUMMARY")
+                .semanticCorrector("NONE_V1")
+                .ruleProfileVersion("v1")
+                .build());
 
         AiTaskServiceImpl service = spy(createService(
                 meetingMapper,
@@ -138,7 +151,8 @@ class AiTaskServiceImplTest {
                 aiModelService,
                 redisTemplate,
                 new TaskSecurityContextRunner(),
-                mock(MeetingTranscriptFileService.class)
+                mock(MeetingTranscriptFileService.class),
+                revisionService
         ));
         doReturn(true).when(service).updateById(any());
 
@@ -146,22 +160,17 @@ class AiTaskServiceImplTest {
         meeting.setId(77L);
         when(meetingMapper.selectById(77L)).thenReturn(meeting);
 
-        MeetingTranscript transcript = new MeetingTranscript();
-        transcript.setSpeakerId("spk-1");
-        transcript.setContent("  ");
-        when(transcriptMapper.selectList(any())).thenReturn(List.of(transcript));
-
         AiTask summaryTask = new AiTask();
         summaryTask.setId(100L);
         summaryTask.setMeetingId(77L);
         summaryTask.setTaskType("SUMMARY");
         summaryTask.setStatus(0);
-        doReturn(summaryTask).when(service).getOne(any());
+        doReturn(summaryTask, null).when(service).getOne(any());
 
         service.dispatchSummaryTask(77L, 1L, 2L);
 
         assertEquals(3, summaryTask.getStatus());
-        assertTrue(summaryTask.getErrorMsg().contains("没有可用于总结的转录内容"));
+        assertTrue(summaryTask.getErrorMsg() != null && !summaryTask.getErrorMsg().isBlank());
         verify(aiModelService, never()).getModelById(anyLong(), anyString());
     }
 
@@ -176,7 +185,8 @@ class AiTaskServiceImplTest {
                 mock(AiModelService.class),
                 mock(StringRedisTemplate.class),
                 mock(TaskSecurityContextRunner.class),
-                transcriptFileService
+                transcriptFileService,
+                mock(MeetingTranscriptRevisionService.class)
         );
 
         Meeting meeting = new Meeting();
@@ -190,9 +200,9 @@ class AiTaskServiceImplTest {
                         {
                           "segments": [
                             {
-                              "speaker_id": "spk-1",
+                              "speaker_id": "123",
                               "speaker_name": "Alice",
-                              "text": "第一段转录",
+                              "text": "hello world",
                               "timestamp": [0, 1200]
                             }
                           ]
@@ -211,7 +221,8 @@ class AiTaskServiceImplTest {
                 mock(AiModelService.class),
                 mock(StringRedisTemplate.class),
                 mock(TaskSecurityContextRunner.class),
-                mock(MeetingTranscriptFileService.class)
+                mock(MeetingTranscriptFileService.class),
+                mock(MeetingTranscriptRevisionService.class)
         );
     }
 
@@ -220,7 +231,8 @@ class AiTaskServiceImplTest {
                                             AiModelService aiModelService,
                                             StringRedisTemplate redisTemplate,
                                             TaskSecurityContextRunner taskSecurityContextRunner,
-                                            MeetingTranscriptFileService meetingTranscriptFileService) {
+                                            MeetingTranscriptFileService meetingTranscriptFileService,
+                                            MeetingTranscriptRevisionService revisionService) {
         return new AiTaskServiceImpl(
                 meetingMapper,
                 transcriptMapper,
@@ -231,6 +243,7 @@ class AiTaskServiceImplTest {
                 redisTemplate,
                 mock(MeetingSummaryFileService.class),
                 meetingTranscriptFileService,
+                revisionService,
                 mock(MeetingSummaryPromptAssembler.class),
                 taskSecurityContextRunner
         );

backend/src/test/java/com/imeeting/service/biz/impl/HotWordGroupServiceImplTest.java

@@ -0,0 +1,66 @@
+package com.imeeting.service.biz.impl;
+
+import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
+import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
+import com.imeeting.entity.biz.HotWordGroup;
+import com.imeeting.mapper.biz.HotWordMapper;
+import com.imeeting.mapper.biz.PromptTemplateMapper;
+import org.junit.jupiter.api.Test;
+import org.mockito.ArgumentCaptor;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.Mockito.doReturn;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.spy;
+import static org.mockito.Mockito.verify;
+import static org.mockito.Mockito.when;
+
+class HotWordGroupServiceImplTest {
+
+    @Test
+    void pageGroupsShouldApplyNameAndStatusFilters() {
+        HotWordMapper hotWordMapper = mock(HotWordMapper.class);
+        PromptTemplateMapper promptTemplateMapper = mock(PromptTemplateMapper.class);
+        when(hotWordMapper.selectList(any())).thenReturn(java.util.List.of());
+
+        HotWordGroupServiceImpl service = spy(new HotWordGroupServiceImpl(hotWordMapper, promptTemplateMapper));
+        Page<HotWordGroup> page = new Page<>(2, 8);
+        page.setRecords(java.util.List.of());
+        page.setTotal(0);
+        doReturn(page).when(service).page(any(Page.class), any(LambdaQueryWrapper.class));
+
+        service.pageGroups(2, 8, "项目", 1, 9L);
+
+        ArgumentCaptor<LambdaQueryWrapper<HotWordGroup>> wrapperCaptor = ArgumentCaptor.forClass(LambdaQueryWrapper.class);
+        verify(service).page(any(Page.class), wrapperCaptor.capture());
+        String sqlSegment = wrapperCaptor.getValue().getSqlSegment().toLowerCase();
+        assertTrue(sqlSegment.contains("group_name"));
+        assertTrue(sqlSegment.contains("status"));
+        assertTrue(sqlSegment.contains("tenant"));
+    }
+
+    @Test
+    void removeGroupByIdShouldRejectWhenPromptTemplateStillReferencesGroup() {
+        HotWordMapper hotWordMapper = mock(HotWordMapper.class);
+        PromptTemplateMapper promptTemplateMapper = mock(PromptTemplateMapper.class);
+        when(promptTemplateMapper.selectCount(any())).thenReturn(1L);
+
+        HotWordGroup group = new HotWordGroup();
+        group.setId(9L);
+        group.setTenantId(9L);
+        HotWordGroupServiceImpl service = new HotWordGroupServiceImpl(hotWordMapper, promptTemplateMapper) {
+            @Override
+            public HotWordGroup getById(java.io.Serializable id) {
+                return Long.valueOf(9L).equals(id) ? group : null;
+            }
+        };
+
+        IllegalArgumentException exception = assertThrows(IllegalArgumentException.class,
+                () -> service.removeGroupById(9L, 9L));
+
+        assertEquals("该热词组已被会议总结模板引用，无法删除", exception.getMessage());
+    }
+}

backend/src/test/java/com/imeeting/service/biz/impl/HotWordServiceImplTest.java

@@ -0,0 +1,71 @@
+package com.imeeting.service.biz.impl;
+
+import com.imeeting.dto.biz.HotWordDTO;
+import com.imeeting.dto.biz.HotWordVO;
+import com.imeeting.entity.biz.HotWord;
+import com.imeeting.entity.biz.HotWordGroup;
+import com.imeeting.mapper.biz.HotWordGroupMapper;
+import org.junit.jupiter.api.Test;
+
+import java.util.List;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.Mockito.doAnswer;
+import static org.mockito.Mockito.doReturn;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.spy;
+import static org.mockito.Mockito.when;
+
+class HotWordServiceImplTest {
+
+    @Test
+    void saveHotWordShouldRejectWhenGroupLimitReached() {
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordServiceImpl service = spy(new HotWordServiceImpl(hotWordGroupMapper));
+        doReturn(200L).when(service).count(any());
+
+        HotWordGroup group = new HotWordGroup();
+        group.setId(5L);
+        group.setTenantId(9L);
+        group.setGroupName("客户名单");
+        group.setStatus(1);
+        when(hotWordGroupMapper.selectById(5L)).thenReturn(group);
+
+        HotWordDTO dto = new HotWordDTO();
+        dto.setWord("阿里");
+        dto.setMatchStrategy(1);
+        dto.setWeight(2);
+        dto.setStatus(1);
+        dto.setHotWordGroupId(5L);
+
+        IllegalArgumentException exception = assertThrows(IllegalArgumentException.class,
+                () -> service.saveHotWord(dto, 7L, 9L));
+
+        assertEquals("热词组最多只能包含 200 个热词", exception.getMessage());
+    }
+    @Test
+    void saveHotWordShouldGeneratePinyinWhenRequestDoesNotProvideIt() {
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordServiceImpl service = spy(new HotWordServiceImpl(hotWordGroupMapper));
+        doAnswer(invocation -> {
+            HotWord entity = invocation.getArgument(0);
+            entity.setId(11L);
+            return true;
+        }).when(service).save(any(HotWord.class));
+
+        HotWordDTO dto = new HotWordDTO();
+        dto.setWord("会议");
+        dto.setMatchStrategy(1);
+        dto.setWeight(2);
+        dto.setStatus(1);
+        dto.setPinyinList(List.of());
+
+        HotWordVO result = service.saveHotWord(dto, 7L, 9L);
+
+        assertFalse(result.getPinyinList().isEmpty());
+        assertEquals("hui yi", result.getPinyinList().get(0));
+    }
+}

backend/src/test/java/com/imeeting/service/biz/impl/MeetingAudioUploadSupportTest.java

@@ -0,0 +1,157 @@
+package com.imeeting.service.biz.impl;
+
+import com.imeeting.common.SysParamKeys;
+import com.unisbase.service.SysParamService;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
+import org.springframework.mock.web.MockMultipartFile;
+import org.springframework.test.util.ReflectionTestUtils;
+
+import java.nio.charset.StandardCharsets;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.Arrays;
+
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+import static org.mockito.ArgumentMatchers.eq;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.when;
+
+class MeetingAudioUploadSupportTest {
+
+    @TempDir
+    Path tempDir;
+
+    @Test
+    void shouldStoreValidatedWavInPrivateStagingDirectory() throws Exception {
+        MeetingAudioUploadSupport support = createSupport(null);
+
+        MockMultipartFile file = new MockMultipartFile("file", "demo.wav", "audio/wav", buildWavHeader());
+
+        String token = support.storeUploadedAudio(file);
+
+        assertTrue(MeetingAudioUploadSupport.isStagingAudioToken(token));
+        Path storedPath = MeetingAudioUploadSupport.resolveStagingAudioPath(tempDir.resolve("uploads").toString(), token);
+        assertTrue(Files.exists(storedPath));
+        assertFalse(storedPath.startsWith(tempDir.resolve("uploads")));
+    }
+
+    @Test
+    void shouldStoreAacM4aInPrivateStagingDirectory() throws Exception {
+        MeetingAudioUploadSupport support = createSupport(null);
+
+        MockMultipartFile file = new MockMultipartFile("file", "demo.m4a", "audio/mp4", buildM4a("mp4a"));
+
+        String token = support.storeUploadedAudio(file);
+
+        assertTrue(MeetingAudioUploadSupport.isStagingAudioToken(token));
+        Path storedPath = MeetingAudioUploadSupport.resolveStagingAudioPath(tempDir.resolve("uploads").toString(), token);
+        assertTrue(Files.exists(storedPath));
+    }
+
+    @Test
+    void shouldRejectM4aWithBrowserIncompatibleCodec() {
+        MeetingAudioUploadSupport support = createSupport(null);
+
+        MockMultipartFile file = new MockMultipartFile("file", "demo.m4a", "audio/mp4", buildM4a("samr"));
+
+        RuntimeException ex = assertThrows(RuntimeException.class, () -> support.storeUploadedAudio(file));
+        assertTrue(ex.getMessage().contains("AAC"));
+    }
+
+    @Test
+    void shouldRejectUnsupportedExtension() {
+        MeetingAudioUploadSupport support = createSupport(null);
+
+        MockMultipartFile file = new MockMultipartFile(
+                "file",
+                "demo.html",
+                "text/html",
+                "<html>alert(1)</html>".getBytes(StandardCharsets.UTF_8)
+        );
+
+        assertThrows(RuntimeException.class, () -> support.storeUploadedAudio(file));
+    }
+
+    @Test
+    void shouldRejectFakeMp3Payload() {
+        MeetingAudioUploadSupport support = createSupport(null);
+
+        MockMultipartFile file = new MockMultipartFile(
+                "file",
+                "fake.mp3",
+                "audio/mpeg",
+                "<script>alert(1)</script>".getBytes(StandardCharsets.UTF_8)
+        );
+
+        assertThrows(RuntimeException.class, () -> support.storeUploadedAudio(file));
+    }
+
+    @Test
+    void shouldRejectAudioLargerThanConfiguredSystemParamLimit() {
+        MeetingAudioUploadSupport support = createSupport("1");
+        byte[] payload = new byte[2 * 1024 * 1024];
+        byte[] header = buildWavHeader();
+        System.arraycopy(header, 0, payload, 0, header.length);
+        MockMultipartFile file = new MockMultipartFile("file", "large.wav", "audio/wav", payload);
+
+        assertThrows(RuntimeException.class, () -> support.storeUploadedAudio(file));
+    }
+
+    private MeetingAudioUploadSupport createSupport(String maxSizeMb) {
+        SysParamService sysParamService = mock(SysParamService.class);
+        when(sysParamService.getCachedParamValue(
+                eq(SysParamKeys.MEETING_OFFLINE_AUDIO_MAX_SIZE_MB),
+                eq("1024")
+        )).thenReturn(maxSizeMb);
+        MeetingAudioUploadSupport support = new MeetingAudioUploadSupport(sysParamService);
+        ReflectionTestUtils.setField(support, "uploadPath", tempDir.resolve("uploads").toString());
+        return support;
+    }
+
+    private byte[] buildWavHeader() {
+        byte[] header = new byte[16];
+        System.arraycopy("RIFF".getBytes(StandardCharsets.US_ASCII), 0, header, 0, 4);
+        System.arraycopy("WAVE".getBytes(StandardCharsets.US_ASCII), 0, header, 8, 4);
+        return header;
+    }
+
+    private byte[] buildM4a(String sampleEntryType) {
+        byte[] sampleEntry = atom(sampleEntryType, new byte[8]);
+        byte[] stsdPayload = concat(new byte[4], intBytes(1), sampleEntry);
+        byte[] stsd = atom("stsd", stsdPayload);
+        byte[] stbl = atom("stbl", stsd);
+        byte[] minf = atom("minf", stbl);
+        byte[] mdia = atom("mdia", minf);
+        byte[] trak = atom("trak", mdia);
+        byte[] moov = atom("moov", trak);
+        byte[] ftyp = atom("ftyp", concat("M4A ".getBytes(StandardCharsets.US_ASCII), new byte[8]));
+        return concat(ftyp, moov);
+    }
+
+    private byte[] atom(String type, byte[] payload) {
+        return concat(intBytes(payload.length + 8), type.getBytes(StandardCharsets.US_ASCII), payload);
+    }
+
+    private byte[] intBytes(int value) {
+        return new byte[]{
+                (byte) ((value >> 24) & 0xFF),
+                (byte) ((value >> 16) & 0xFF),
+                (byte) ((value >> 8) & 0xFF),
+                (byte) (value & 0xFF)
+        };
+    }
+
+    private byte[] concat(byte[]... parts) {
+        int totalLength = Arrays.stream(parts).mapToInt(part -> part.length).sum();
+        byte[] result = new byte[totalLength];
+        int offset = 0;
+        for (byte[] part : parts) {
+            System.arraycopy(part, 0, result, offset, part.length);
+            offset += part.length;
+        }
+        return result;
+    }
+}

backend/src/test/java/com/imeeting/service/biz/impl/MeetingPlaybackAudioResolverTest.java

@@ -0,0 +1,229 @@
+package com.imeeting.service.biz.impl;
+
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
+import org.springframework.test.util.ReflectionTestUtils;
+
+import java.io.ByteArrayOutputStream;
+import java.io.IOException;
+import java.nio.charset.StandardCharsets;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.attribute.FileTime;
+import java.util.Arrays;
+
+import static org.junit.jupiter.api.Assertions.assertArrayEquals;
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+class MeetingPlaybackAudioResolverTest {
+
+    @TempDir
+    Path tempDir;
+
+    @Test
+    void shouldGenerateBrowserPlaybackWavFor16000Source() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path sourcePath = tempDir.resolve("uploads/meetings/101/source_audio.wav");
+        byte[] sourceFrames = new byte[]{1, 2, 3, 4};
+        Files.createDirectories(sourcePath.getParent());
+        Files.write(sourcePath, buildWav(16_000, sourceFrames));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/101/source_audio.wav");
+
+        assertEquals("/api/static/meetings/101/source_audio_browser_48000.wav", playbackUrl);
+        Path convertedPath = tempDir.resolve("uploads/meetings/101/source_audio_browser_48000.wav");
+        assertTrue(Files.exists(convertedPath));
+
+        byte[] converted = Files.readAllBytes(convertedPath);
+        assertEquals(48_000, readIntLe(converted, 24));
+        assertEquals(12, readIntLe(converted, 40));
+        assertArrayEquals(
+                new byte[]{1, 2, 1, 2, 1, 2, 3, 4, 3, 4, 3, 4},
+                Arrays.copyOfRange(converted, 44, 56)
+        );
+    }
+
+    @Test
+    void shouldReuseExistingBrowserPlaybackWavFile() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path meetingDir = tempDir.resolve("uploads/meetings/102");
+        Path sourcePath = meetingDir.resolve("source_audio.wav");
+        Path convertedPath = meetingDir.resolve("source_audio_browser_48000.wav");
+        Files.createDirectories(meetingDir);
+        Files.write(sourcePath, buildWav(16_000, new byte[]{10, 20, 30, 40}));
+        Files.write(convertedPath, buildWav(48_000, new byte[]{9, 9, 8, 8, 7, 7}));
+        Files.setLastModifiedTime(sourcePath, FileTime.fromMillis(1_000));
+        Files.setLastModifiedTime(convertedPath, FileTime.fromMillis(2_000));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/102/source_audio.wav");
+
+        assertEquals("/api/static/meetings/102/source_audio_browser_48000.wav", playbackUrl);
+        byte[] converted = Files.readAllBytes(convertedPath);
+        assertEquals(48_000, readIntLe(converted, 24));
+        assertArrayEquals(new byte[]{9, 9, 8, 8, 7, 7}, Arrays.copyOfRange(converted, 44, 50));
+    }
+
+    @Test
+    void shouldKeep44100WavSourceForBrowserPlayback() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path sourcePath = tempDir.resolve("uploads/meetings/103/source_audio.wav");
+        Files.createDirectories(sourcePath.getParent());
+        Files.write(sourcePath, buildWav(44_100, new byte[]{1, 1, 2, 2}));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/103/source_audio.wav");
+
+        assertEquals("/api/static/meetings/103/source_audio.wav", playbackUrl);
+        assertFalse(Files.exists(tempDir.resolve("uploads/meetings/103/source_audio_browser_48000.wav")));
+    }
+
+    @Test
+    void shouldKeep44100M4aSourceForBrowserPlayback() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path sourcePath = tempDir.resolve("uploads/meetings/104/source_audio.m4a");
+        Files.createDirectories(sourcePath.getParent());
+        Files.write(sourcePath, buildM4a("mp4a", 44_100));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/104/source_audio.m4a");
+
+        assertEquals("/api/static/meetings/104/source_audio.m4a", playbackUrl);
+        assertFalse(Files.exists(tempDir.resolve("uploads/meetings/104/source_audio_browser_48000.m4a")));
+    }
+
+    @Test
+    void shouldReuseExistingBrowserPlaybackM4aFile() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path meetingDir = tempDir.resolve("uploads/meetings/105");
+        Path sourcePath = meetingDir.resolve("source_audio.m4a");
+        Path convertedPath = meetingDir.resolve("source_audio_browser_48000.m4a");
+        Files.createDirectories(meetingDir);
+        Files.write(sourcePath, buildM4a("mp4a", 16_000));
+        Files.write(convertedPath, buildM4a("mp4a", 48_000));
+        Files.setLastModifiedTime(sourcePath, FileTime.fromMillis(1_000));
+        Files.setLastModifiedTime(convertedPath, FileTime.fromMillis(2_000));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/105/source_audio.m4a");
+
+        assertEquals("/api/static/meetings/105/source_audio_browser_48000.m4a", playbackUrl);
+        assertTrue(Files.exists(convertedPath));
+    }
+
+    @Test
+    void shouldFallbackWhenM4aNeedsConversionButFfmpegIsUnavailable() throws Exception {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path sourcePath = tempDir.resolve("uploads/meetings/106/source_audio.m4a");
+        Files.createDirectories(sourcePath.getParent());
+        Files.write(sourcePath, buildM4a("mp4a", 16_000));
+
+        String playbackUrl = resolver.resolveBrowserPlaybackAudioUrl("/api/static/meetings/106/source_audio.m4a");
+
+        assertEquals("/api/static/meetings/106/source_audio.m4a", playbackUrl);
+        assertFalse(Files.exists(tempDir.resolve("uploads/meetings/106/source_audio_browser_48000.m4a")));
+    }
+
+    @Test
+    void shouldKeepOriginalExtensionForTemporaryM4aOutput() {
+        MeetingPlaybackAudioResolver resolver = newResolver();
+        Path targetPath = tempDir.resolve("uploads/meetings/107/source_audio_browser_48000.m4a");
+
+        Path tempPath = ReflectionTestUtils.invokeMethod(resolver, "buildTemporaryOutputPath", targetPath);
+
+        assertEquals("source_audio_browser_48000.tmp.m4a", tempPath.getFileName().toString());
+    }
+
+    private MeetingPlaybackAudioResolver newResolver() {
+        MeetingPlaybackAudioResolver resolver = new MeetingPlaybackAudioResolver();
+        ReflectionTestUtils.setField(resolver, "uploadPath", tempDir.resolve("uploads").toString());
+        ReflectionTestUtils.setField(resolver, "resourcePrefix", "/api/static/");
+        ReflectionTestUtils.setField(resolver, "ffmpegPath", "ffmpeg");
+        return resolver;
+    }
+
+    private byte[] buildWav(int sampleRate, byte[] data) throws IOException {
+        ByteArrayOutputStream output = new ByteArrayOutputStream();
+        int channels = 1;
+        int bitsPerSample = 16;
+        int blockAlign = channels * bitsPerSample / 8;
+        long byteRate = (long) sampleRate * blockAlign;
+
+        output.write(new byte[]{'R', 'I', 'F', 'F'});
+        writeIntLe(output, 36 + data.length);
+        output.write(new byte[]{'W', 'A', 'V', 'E'});
+        output.write(new byte[]{'f', 'm', 't', ' '});
+        writeIntLe(output, 16);
+        writeShortLe(output, 1);
+        writeShortLe(output, channels);
+        writeIntLe(output, sampleRate);
+        writeIntLe(output, byteRate);
+        writeShortLe(output, blockAlign);
+        writeShortLe(output, bitsPerSample);
+        output.write(new byte[]{'d', 'a', 't', 'a'});
+        writeIntLe(output, data.length);
+        output.write(data);
+        return output.toByteArray();
+    }
+
+    private byte[] buildM4a(String sampleEntryType, int sampleRate) {
+        byte[] sampleEntryPayload = new byte[28];
+        int fixedPointSampleRate = sampleRate << 16;
+        sampleEntryPayload[24] = (byte) ((fixedPointSampleRate >> 24) & 0xFF);
+        sampleEntryPayload[25] = (byte) ((fixedPointSampleRate >> 16) & 0xFF);
+        sampleEntryPayload[26] = (byte) ((fixedPointSampleRate >> 8) & 0xFF);
+        sampleEntryPayload[27] = (byte) (fixedPointSampleRate & 0xFF);
+
+        byte[] sampleEntry = atom(sampleEntryType, sampleEntryPayload);
+        byte[] stsdPayload = concat(new byte[4], intBytes(1), sampleEntry);
+        byte[] stsd = atom("stsd", stsdPayload);
+        byte[] stbl = atom("stbl", stsd);
+        byte[] minf = atom("minf", stbl);
+        byte[] mdia = atom("mdia", minf);
+        byte[] trak = atom("trak", mdia);
+        byte[] moov = atom("moov", trak);
+        byte[] ftyp = atom("ftyp", concat("M4A ".getBytes(StandardCharsets.US_ASCII), new byte[8]));
+        return concat(ftyp, moov);
+    }
+
+    private byte[] atom(String type, byte[] payload) {
+        return concat(intBytes(payload.length + 8), type.getBytes(StandardCharsets.US_ASCII), payload);
+    }
+
+    private byte[] intBytes(int value) {
+        return new byte[]{
+                (byte) ((value >> 24) & 0xFF),
+                (byte) ((value >> 16) & 0xFF),
+                (byte) ((value >> 8) & 0xFF),
+                (byte) (value & 0xFF)
+        };
+    }
+
+    private byte[] concat(byte[]... parts) {
+        int totalLength = Arrays.stream(parts).mapToInt(part -> part.length).sum();
+        byte[] result = new byte[totalLength];
+        int offset = 0;
+        for (byte[] part : parts) {
+            System.arraycopy(part, 0, result, offset, part.length);
+            offset += part.length;
+        }
+        return result;
+    }
+
+    private void writeShortLe(ByteArrayOutputStream output, int value) {
+        output.write(value & 0xff);
+        output.write((value >> 8) & 0xff);
+    }
+
+    private void writeIntLe(ByteArrayOutputStream output, long value) {
+        output.write((int) (value & 0xff));
+        output.write((int) ((value >> 8) & 0xff));
+        output.write((int) ((value >> 16) & 0xff));
+        output.write((int) ((value >> 24) & 0xff));
+    }
+
+    private int readIntLe(byte[] bytes, int offset) {
+        return (bytes[offset] & 0xff)
+                | ((bytes[offset + 1] & 0xff) << 8)
+                | ((bytes[offset + 2] & 0xff) << 16)
+                | ((bytes[offset + 3] & 0xff) << 24);
+    }
+}

backend/src/test/java/com/imeeting/service/biz/impl/MeetingTranscriptFileServiceImplTest.java

@@ -0,0 +1,106 @@
+package com.imeeting.service.biz.impl;
+
+import com.imeeting.dto.biz.MeetingTranscriptExportResult;
+import com.imeeting.dto.biz.MeetingVO;
+import com.imeeting.entity.biz.Meeting;
+import com.imeeting.entity.biz.MeetingTranscript;
+import com.imeeting.mapper.biz.MeetingMapper;
+import com.imeeting.mapper.biz.MeetingTranscriptMapper;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
+import org.springframework.test.util.ReflectionTestUtils;
+
+import java.nio.charset.StandardCharsets;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.time.LocalDateTime;
+import java.util.List;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.times;
+import static org.mockito.Mockito.verify;
+import static org.mockito.Mockito.when;
+
+class MeetingTranscriptFileServiceImplTest {
+
+    @TempDir
+    Path tempDir;
+
+    @Test
+    void initializeTranscriptFileIfAbsentShouldCreateMarkdownFile() throws Exception {
+        MeetingMapper meetingMapper = mock(MeetingMapper.class);
+        MeetingTranscriptMapper transcriptMapper = mock(MeetingTranscriptMapper.class);
+        MeetingTranscriptFileServiceImpl service = newService(meetingMapper, transcriptMapper);
+
+        Meeting meeting = new Meeting();
+        meeting.setId(1001L);
+        meeting.setTitle("Weekly Sync");
+        meeting.setHostName("Alice");
+        meeting.setMeetingTime(LocalDateTime.of(2026, 4, 27, 10, 0, 0));
+
+        MeetingTranscript transcript = new MeetingTranscript();
+        transcript.setSpeakerName("Bob");
+        transcript.setContent("Confirm this week's release plan");
+        transcript.setStartTime(0);
+        transcript.setEndTime(5000);
+
+        when(meetingMapper.selectById(1001L)).thenReturn(meeting);
+        when(transcriptMapper.selectList(any())).thenReturn(List.of(transcript));
+
+        service.initializeTranscriptFileIfAbsent(1001L);
+
+        Path transcriptPath = tempDir.resolve("uploads/meetings/1001/transcripts/current.md");
+        assertTrue(Files.exists(transcriptPath));
+        String markdown = Files.readString(transcriptPath, StandardCharsets.UTF_8);
+        assertTrue(markdown.contains("# Weekly Sync Transcript"));
+        assertTrue(markdown.contains("Bob: Confirm this week's release plan"));
+        verify(meetingMapper, times(1)).selectById(1001L);
+    }
+
+    @Test
+    void exportTranscriptShouldRewriteFileWithLatestTranscriptContent() throws Exception {
+        MeetingMapper meetingMapper = mock(MeetingMapper.class);
+        MeetingTranscriptMapper transcriptMapper = mock(MeetingTranscriptMapper.class);
+        MeetingTranscriptFileServiceImpl service = newService(meetingMapper, transcriptMapper);
+
+        Meeting meeting = new Meeting();
+        meeting.setId(1002L);
+        meeting.setTitle("Architecture Review");
+        meeting.setHostName("Carol");
+        meeting.setMeetingTime(LocalDateTime.of(2026, 4, 27, 14, 30, 0));
+
+        MeetingVO meetingVO = new MeetingVO();
+        meetingVO.setTitle("Architecture Review");
+        meetingVO.setHostName("Carol");
+        meetingVO.setMeetingTime(LocalDateTime.of(2026, 4, 27, 14, 30, 0));
+
+        MeetingTranscript transcript = new MeetingTranscript();
+        transcript.setSpeakerName("Dave");
+        transcript.setContent("Ship download support first");
+        transcript.setStartTime(1000);
+        transcript.setEndTime(4000);
+
+        when(transcriptMapper.selectList(any())).thenReturn(List.of(transcript));
+
+        Path transcriptPath = tempDir.resolve("uploads/meetings/1002/transcripts/current.md");
+        Files.createDirectories(transcriptPath.getParent());
+        Files.writeString(transcriptPath, "old content", StandardCharsets.UTF_8);
+
+        MeetingTranscriptExportResult result = service.exportTranscript(meeting, meetingVO);
+
+        String markdown = Files.readString(transcriptPath, StandardCharsets.UTF_8);
+        assertTrue(markdown.contains("Dave: Ship download support first"));
+        assertEquals("text/markdown; charset=UTF-8", result.getContentType());
+        assertEquals("Architecture Review-Transcript.md", result.getFileName());
+        assertTrue(new String(result.getContent(), StandardCharsets.UTF_8).contains("Ship download support first"));
+    }
+
+    private MeetingTranscriptFileServiceImpl newService(MeetingMapper meetingMapper, MeetingTranscriptMapper transcriptMapper) {
+        MeetingTranscriptFileServiceImpl service = new MeetingTranscriptFileServiceImpl(meetingMapper, transcriptMapper);
+        ReflectionTestUtils.setField(service, "uploadPath", tempDir.resolve("uploads").toString());
+        return service;
+    }
+}

backend/src/test/java/com/imeeting/service/biz/impl/PromptTemplateServiceImplTest.java

@@ -0,0 +1,217 @@
+package com.imeeting.service.biz.impl;
+
+import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
+import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
+import com.imeeting.dto.biz.PromptTemplateDTO;
+import com.imeeting.dto.biz.PromptTemplateVO;
+import com.imeeting.entity.biz.HotWord;
+import com.imeeting.entity.biz.HotWordGroup;
+import com.imeeting.entity.biz.PromptTemplate;
+import com.imeeting.mapper.biz.HotWordGroupMapper;
+import com.imeeting.mapper.biz.PromptTemplateUserConfigMapper;
+import com.imeeting.service.biz.HotWordService;
+import org.junit.jupiter.api.Test;
+
+import java.util.List;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertThrows;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.Mockito.doReturn;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.spy;
+import static org.mockito.Mockito.when;
+
+class PromptTemplateServiceImplTest {
+
+    @Test
+    void saveTemplateShouldRejectPlatformTemplateBindingTenantGroup() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+
+        HotWordGroup tenantGroup = new HotWordGroup();
+        tenantGroup.setId(11L);
+        tenantGroup.setTenantId(9L);
+        tenantGroup.setGroupName("租户组");
+        tenantGroup.setStatus(1);
+        when(hotWordGroupMapper.selectById(11L)).thenReturn(tenantGroup);
+
+        PromptTemplateDTO dto = new PromptTemplateDTO();
+        dto.setTenantId(0L);
+        dto.setTemplateName("平台模板");
+        dto.setCategory("default");
+        dto.setIsSystem(1);
+        dto.setPromptContent("content");
+        dto.setStatus(1);
+        dto.setHotWordGroupId(11L);
+
+        IllegalArgumentException exception = assertThrows(IllegalArgumentException.class,
+                () -> service.saveTemplate(dto, 7L, 9L));
+
+        assertEquals("平台级模板只能绑定平台级热词组", exception.getMessage());
+    }
+
+    @Test
+    void saveTemplateShouldReturnBoundGroupInfo() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+        doReturn(true).when(service).save(any(PromptTemplate.class));
+
+        HotWordGroup group = new HotWordGroup();
+        group.setId(11L);
+        group.setTenantId(9L);
+        group.setGroupName("项目术语");
+        group.setStatus(1);
+        when(hotWordGroupMapper.selectById(11L)).thenReturn(group);
+        when(hotWordGroupMapper.selectByIdsIgnoreTenant(any())).thenReturn(java.util.List.of(group));
+
+        PromptTemplateDTO dto = new PromptTemplateDTO();
+        dto.setTemplateName("租户模板");
+        dto.setCategory("default");
+        dto.setIsSystem(0);
+        dto.setPromptContent("content");
+        dto.setStatus(1);
+        dto.setHotWordGroupId(11L);
+
+        PromptTemplateVO result = service.saveTemplate(dto, 7L, 9L);
+
+        assertEquals(11L, result.getHotWordGroupId());
+        assertEquals("项目术语", result.getHotWordGroupName());
+    }
+
+    @Test
+    void saveTemplateShouldAllowPlatformTemplateBindingPlatformGroup() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+        doReturn(true).when(service).save(any(PromptTemplate.class));
+
+        HotWordGroup group = new HotWordGroup();
+        group.setId(11L);
+        group.setTenantId(0L);
+        group.setGroupName("平台术语");
+        group.setStatus(1);
+        when(hotWordGroupMapper.selectById(11L)).thenReturn(group);
+        when(hotWordGroupMapper.selectByIdsIgnoreTenant(any())).thenReturn(java.util.List.of(group));
+
+        PromptTemplateDTO dto = new PromptTemplateDTO();
+        dto.setTenantId(0L);
+        dto.setTemplateName("平台模板");
+        dto.setCategory("default");
+        dto.setIsSystem(1);
+        dto.setPromptContent("content");
+        dto.setStatus(1);
+        dto.setHotWordGroupId(11L);
+
+        PromptTemplateVO result = service.saveTemplate(dto, 7L, 9L);
+
+        assertEquals(11L, result.getHotWordGroupId());
+        assertEquals("平台术语", result.getHotWordGroupName());
+    }
+
+    @Test
+    void pageTemplatesShouldHandleTemplateWithoutBoundGroup() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+
+        PromptTemplate template = new PromptTemplate();
+        template.setId(21L);
+        template.setTenantId(9L);
+        template.setCreatorId(7L);
+        template.setTemplateName("未绑定模板");
+        template.setCategory("default");
+        template.setIsSystem(0);
+        template.setPromptContent("content");
+        template.setStatus(1);
+        template.setHotWordGroupId(null);
+
+        Page<PromptTemplate> page = new Page<>(1, 10);
+        page.setRecords(List.of(template));
+        page.setTotal(1);
+
+        doReturn(page).when(service).page(any(Page.class), any(LambdaQueryWrapper.class));
+        when(userConfigMapper.selectList(any())).thenReturn(List.of());
+
+        PromptTemplateVO result = service.pageTemplates(1, 10, null, null, 9L, 7L, false, false)
+                .getRecords()
+                .get(0);
+
+        assertEquals(21L, result.getId());
+        assertNull(result.getHotWordGroupId());
+        assertNull(result.getHotWordGroupName());
+    }
+
+    @Test
+    void getTemplateDetailShouldReturnBoundHotWords() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+
+        PromptTemplate template = new PromptTemplate();
+        template.setId(31L);
+        template.setTenantId(0L);
+        template.setCreatorId(1L);
+        template.setTemplateName("平台模板");
+        template.setCategory("default");
+        template.setIsSystem(1);
+        template.setPromptContent("content");
+        template.setStatus(1);
+        template.setHotWordGroupId(11L);
+
+        HotWordGroup group = new HotWordGroup();
+        group.setId(11L);
+        group.setTenantId(0L);
+        group.setGroupName("平台热词组");
+
+        HotWord word1 = new HotWord();
+        word1.setWord("OpenAI");
+        HotWord word2 = new HotWord();
+        word2.setWord("Codex");
+
+        doReturn(template).when(service).getOne(any(LambdaQueryWrapper.class));
+        when(hotWordGroupMapper.selectByIdsIgnoreTenant(any())).thenReturn(List.of(group));
+        when(hotWordService.listEnabledByGroupIdIgnoreTenant(11L)).thenReturn(List.of(word1, word2));
+
+        PromptTemplateVO result = service.getTemplateDetail(31L, 9L, 7L, false, false);
+
+        assertEquals("平台热词组", result.getHotWordGroupName());
+        assertEquals(List.of("OpenAI", "Codex"), result.getHotWords());
+    }
+
+    @Test
+    void getTemplateDetailShouldHandleTemplateWithoutBoundGroup() {
+        PromptTemplateUserConfigMapper userConfigMapper = mock(PromptTemplateUserConfigMapper.class);
+        HotWordGroupMapper hotWordGroupMapper = mock(HotWordGroupMapper.class);
+        HotWordService hotWordService = mock(HotWordService.class);
+        PromptTemplateServiceImpl service = spy(new PromptTemplateServiceImpl(userConfigMapper, hotWordGroupMapper, hotWordService));
+
+        PromptTemplate template = new PromptTemplate();
+        template.setId(32L);
+        template.setTenantId(9L);
+        template.setCreatorId(7L);
+        template.setTemplateName("detail-without-group");
+        template.setCategory("default");
+        template.setIsSystem(0);
+        template.setPromptContent("content");
+        template.setStatus(1);
+        template.setHotWordGroupId(null);
+
+        doReturn(template).when(service).getOne(any(LambdaQueryWrapper.class));
+
+        PromptTemplateVO result = service.getTemplateDetail(32L, 9L, 7L, false, false);
+
+        assertEquals(32L, result.getId());
+        assertNull(result.getHotWordGroupId());
+        assertNull(result.getHotWordGroupName());
+        assertEquals(List.of(), result.getHotWords());
+    }
+}