feat(speech): add qwen tts voice preferences

backend/ruoyi-modules/ruoyi-aihr/src/main/java/org/dromara/aihr/service/AihrModelSeedService.java

@@ -58,6 +58,12 @@ public class AihrModelSeedService {
     @Value("${aihr.ai-runtime.speech-enabled:${AIHR_AI_SPEECH_ENABLED:true}}")
     private boolean speechEnabled;
 
+    @Value("${aihr.realtime-practice.endpoint:${AIHR_QWEN_REALTIME_ENDPOINT:}}")
+    private String qwenRealtimeEndpoint;
+
+    @Value("${aihr.realtime-practice.api-key:${AIHR_QWEN_REALTIME_API_KEY:}}")
+    private String qwenRealtimeApiKey;
+
     public List<ProviderResponse> providers() {
         List<ProviderResponse> rows = dbProviders();
         return rows.isEmpty() ? seedProviders() : mergeProviders(rows);
@@ -267,7 +273,7 @@ public class AihrModelSeedService {
                 rs.getString("resolved_api_key")
             ), tenantId(), category);
             return rows.stream()
-                .filter(model -> configured(model.baseUrl(), model.modelName(), model.apiKey()))
+                .filter(model -> configured(model.baseUrl(), model.modelName(), model.apiKey()) || qwenTtsConfigured(model))
                 .findFirst();
         } catch (DataAccessException e) {
             log.debug("aihr speech model db fallback（处理错误已隐藏）");
@@ -282,6 +288,16 @@ public class AihrModelSeedService {
     public record SpeechModel(String providerCode, String modelName, String baseUrl, String apiKey) {
     }
 
+    private boolean qwenTtsConfigured(SpeechModel model) {
+        return isDashScopeQwenTts(model) && !isBlank(qwenRealtimeEndpoint) && !isBlank(qwenRealtimeApiKey);
+    }
+
+    private static boolean isDashScopeQwenTts(SpeechModel model) {
+        return model != null
+            && ("dashscope".equalsIgnoreCase(model.providerCode()) || "qianwen".equalsIgnoreCase(model.providerCode()))
+            && "qwen-audio-3.0-tts-flash".equals(model.modelName());
+    }
+
     /**
      * 供其他模块(如三角色对练)复用的 chat 调用：模型未配置或调用失败返回 empty，由调用方决定兜底。
      */
@@ -354,19 +370,28 @@ public class AihrModelSeedService {
                     on p.tenant_id = c.tenant_id and p.provider_code = c.provider_code
                  where c.tenant_id = ?
                  order by case c.category when 'chat' then 0 when 'vector' then 1 when 'rerank' then 2 else 9 end, c.id asc
-                """, (rs, rowNum) -> new ConfigResponse(
-                rs.getLong("id"),
-                rs.getString("category"),
-                rs.getString("model_name"),
-                rs.getString("provider_code"),
-                normalizeModelDescribe(rs.getString("model_describe")),
-                rs.getObject("model_dimension", Integer.class),
-                rs.getString("model_show"),
-                rs.getString("api_host"),
-                rs.getInt("enabled") == 1,
-                !isBlank(rs.getString("resolved_api_key")),
-                configured(rs.getString("resolved_api_host"), rs.getString("model_name"), rs.getString("resolved_api_key"))
-            ), tenantId());
+            """, (rs, rowNum) -> {
+                SpeechModel runtime = new SpeechModel(
+                    rs.getString("provider_code"),
+                    rs.getString("model_name"),
+                    rs.getString("resolved_api_host"),
+                    rs.getString("resolved_api_key")
+                );
+                boolean qwenTts = qwenTtsConfigured(runtime);
+                return new ConfigResponse(
+                    rs.getLong("id"),
+                    rs.getString("category"),
+                    rs.getString("model_name"),
+                    rs.getString("provider_code"),
+                    normalizeModelDescribe(rs.getString("model_describe")),
+                    rs.getObject("model_dimension", Integer.class),
+                    rs.getString("model_show"),
+                    rs.getString("api_host"),
+                    rs.getInt("enabled") == 1,
+                    qwenTts || !isBlank(runtime.apiKey()),
+                    qwenTts || configured(runtime.baseUrl(), runtime.modelName(), runtime.apiKey())
+                );
+            }, tenantId());
         } catch (DataAccessException e) {
             log.debug("aihr model config db fallback（处理错误已隐藏）");
             return List.of();

backend/ruoyi-modules/ruoyi-aihr/src/main/java/org/dromara/aihr/service/AihrSpeechService.java

@@ -6,6 +6,7 @@ import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 import org.dromara.aihr.domain.AihrSpeechDto.VoiceProfile;
 import org.dromara.aihr.service.AihrModelSeedService.SpeechModel;
+import org.springframework.beans.factory.annotation.Value;
 import org.springframework.stereotype.Service;
 
 import java.io.ByteArrayOutputStream;
@@ -33,7 +34,10 @@ public class AihrSpeechService {
 
     private static final Duration CONNECT_TIMEOUT = Duration.ofSeconds(15);
     private static final Duration REQUEST_TIMEOUT = Duration.ofSeconds(5);
+    private static final Duration DASHSCOPE_TTS_TIMEOUT = Duration.ofSeconds(20);
     private static final String DEFAULT_TTS_VOICE = "anna";
+    private static final String QWEN_TTS_MODEL = "qwen-audio-3.0-tts-flash";
+    private static final String QWEN_TTS_DEFAULT_VOICE = "longanhuan_v3.6";
     private static final Pattern SAFE_VOICE = Pattern.compile("[A-Za-z0-9_./:-]{1,200}");
     private static final Map<String, String> ROLE_VOICES = Map.of(
         "mentor", "speech:shifu-warm-v1:cm3hz4wfz02jy106j6z6muix7:ysvbyzypjgmnceokxedr",
@@ -47,15 +51,48 @@ public class AihrSpeechService {
         "intense", "请用情绪强烈、急切、有压力且明显不满的语气说",
         "calm", "请用平静、自然的语气说"
     );
-    private static final Map<String, String> DIALECT_PROMPTS = Map.of(
-        "mandarin", "请使用自然普通话表达",
-        "cantonese", "请使用自然粤语口语表达",
-        "sichuanese", "请使用自然四川话口语表达"
+    private static final Map<String, String> DIALECT_PROMPTS = Map.ofEntries(
+        Map.entry("mandarin", "请使用自然普通话表达"),
+        Map.entry("cantonese", "请使用自然粤语口语表达"),
+        Map.entry("chongqing", "请使用自然重庆话口语表达"),
+        Map.entry("northeastern", "请使用自然东北话口语表达"),
+        Map.entry("gansu", "请使用自然甘肃话口语表达"),
+        Map.entry("guizhou", "请使用自然贵州话口语表达"),
+        Map.entry("zhejiang", "请使用自然浙江话口语表达"),
+        Map.entry("hebei", "请使用自然河北话口语表达"),
+        Map.entry("henan", "请使用自然河南话口语表达"),
+        Map.entry("hubei", "请使用自然湖北话口语表达"),
+        Map.entry("hunan", "请使用自然湖南话口语表达"),
+        Map.entry("jiangxi", "请使用自然江西话口语表达"),
+        Map.entry("ningbo", "请使用自然宁波话口语表达"),
+        Map.entry("ningxia", "请使用自然宁夏话口语表达"),
+        Map.entry("qingdao", "请使用自然青岛话口语表达"),
+        Map.entry("shaanxi", "请使用自然陕西话口语表达"),
+        Map.entry("shanxi", "请使用自然山西话口语表达"),
+        Map.entry("shandong", "请使用自然山东话口语表达"),
+        Map.entry("shanghai", "请使用自然上海话口语表达"),
+        Map.entry("sichuanese", "请使用自然四川话口语表达"),
+        Map.entry("yunnan", "请使用自然云南话口语表达")
+    );
+    private static final Map<String, String> QWEN_TTS_ROLE_VOICES = Map.of(
+        "mentor", QWEN_TTS_DEFAULT_VOICE,
+        "customer", QWEN_TTS_DEFAULT_VOICE,
+        "interviewer", QWEN_TTS_DEFAULT_VOICE,
+        "neutral", QWEN_TTS_DEFAULT_VOICE
+    );
+    private static final java.util.Set<String> QWEN_TTS_VOICES = java.util.Set.of(
+        "longanhuan_v3.6", "longjielidou_v3.6", "loongeva_v3.6", "loongjohn"
     );
 
     private final AihrModelSeedService modelService;
     private final ObjectMapper objectMapper;
 
+    @Value("${aihr.realtime-practice.endpoint:${AIHR_QWEN_REALTIME_ENDPOINT:}}")
+    private String qwenRealtimeEndpoint;
+
+    @Value("${aihr.realtime-practice.api-key:${AIHR_QWEN_REALTIME_API_KEY:}}")
+    private String qwenRealtimeApiKey;
+
     public boolean asrConfigured() {
         return modelService.speechModel("asr").isPresent();
     }
@@ -71,7 +108,7 @@ public class AihrSpeechService {
         if (rawDialect.isEmpty() || DIALECT_PROMPTS.containsKey(rawDialect.toLowerCase(Locale.ROOT))) {
             return Optional.empty();
         }
-        return Optional.of("dialect 仅支持 mandarin、cantonese、sichuanese");
+        return Optional.of("dialect 不受支持");
     }
 
     /**
@@ -117,6 +154,9 @@ public class AihrSpeechService {
         }
         SpeechModel runtime = model.get();
         try {
+            if (isDashScopeQwenTts(runtime)) {
+                return synthesizeDashScopeQwen(runtime, text, voice, voiceProfile);
+            }
             ObjectNode body = buildRequestBody(objectMapper, runtime, text, voice, voiceProfile);
             HttpRequest httpRequest = authorized(runtime, "/audio/speech")
                 .header("Content-Type", "application/json")
@@ -134,6 +174,38 @@ public class AihrSpeechService {
         }
     }
 
+    private Optional<byte[]> synthesizeDashScopeQwen(SpeechModel runtime, String text, String voice, VoiceProfile voiceProfile) {
+        if (isBlank(qwenRealtimeEndpoint) || isBlank(qwenRealtimeApiKey)) {
+            return Optional.empty();
+        }
+        try {
+            ObjectNode body = buildDashScopeRequestBody(objectMapper, runtime, text, voiceProfile == null
+                ? new VoiceProfile("neutral", voice, null, null, null)
+                : withLegacyVoice(voiceProfile, voice));
+            HttpRequest request = HttpRequest.newBuilder()
+                .uri(dashScopeSynthesisUri(qwenRealtimeEndpoint))
+                .timeout(DASHSCOPE_TTS_TIMEOUT)
+                .header("Content-Type", "application/json")
+                .header("Authorization", "Bearer " + qwenRealtimeApiKey.trim())
+                .POST(HttpRequest.BodyPublishers.ofString(objectMapper.writeValueAsString(body)))
+                .build();
+            HttpResponse<String> response = client().send(request, HttpResponse.BodyHandlers.ofString());
+            if (response.statusCode() < 200 || response.statusCode() >= 300) {
+                log.warn("dashscope tts http {}（外部响应体已隐藏）", response.statusCode());
+                return Optional.empty();
+            }
+            Optional<String> audioUrl = dashScopeAudioUrl(objectMapper, response.body());
+            if (audioUrl.isEmpty()) {
+                log.warn("dashscope tts response missing audio url");
+                return Optional.empty();
+            }
+            return downloadDashScopeAudio(audioUrl.get());
+        } catch (Exception e) {
+            log.warn("dashscope tts call failed（处理错误已隐藏）");
+            return Optional.empty();
+        }
+    }
+
     static ObjectNode buildRequestBody(ObjectMapper objectMapper, SpeechModel runtime, String text,
                                        String legacyVoice, VoiceProfile voiceProfile) {
         String role = normalizedRole(voiceProfile == null ? null : voiceProfile.role());
@@ -142,6 +214,9 @@ public class AihrSpeechService {
         if (requestedVoice == null) {
             requestedVoice = safeVoice(legacyVoice);
         }
+        if (requestedVoice != null && QWEN_TTS_VOICES.contains(requestedVoice)) {
+            requestedVoice = null;
+        }
         if (requestedVoice == null && expressiveCosyVoice) {
             requestedVoice = ROLE_VOICES.get(role);
         }
@@ -165,6 +240,80 @@ public class AihrSpeechService {
         return body;
     }
 
+    static ObjectNode buildDashScopeRequestBody(ObjectMapper objectMapper, SpeechModel runtime, String text,
+                                                VoiceProfile voiceProfile) {
+        String role = normalizedRole(voiceProfile == null ? null : voiceProfile.role());
+        String emotion = normalizedEmotion(voiceProfile == null ? null : voiceProfile.emotion(), role);
+        String dialect = normalizedDialect(voiceProfile == null ? null : voiceProfile.dialect());
+        String requestedVoice = safeVoice(voiceProfile == null ? null : voiceProfile.voice());
+        String voice = requestedVoice != null && QWEN_TTS_VOICES.contains(requestedVoice)
+            ? requestedVoice
+            : QWEN_TTS_ROLE_VOICES.getOrDefault(role, QWEN_TTS_DEFAULT_VOICE);
+        ObjectNode input = objectMapper.createObjectNode();
+        input.put("text", AihrSensitiveText.forModel(text.trim()));
+        input.put("voice", voice);
+        input.put("format", "mp3");
+        input.put("sample_rate", 24000);
+        input.put("rate", resolveSpeed(voiceProfile == null ? null : voiceProfile.speed(), emotion));
+        String instruction = expressivePrompt(emotion, dialect);
+        if (!instruction.isBlank()) {
+            input.put("instruction", instruction);
+        }
+        ObjectNode body = objectMapper.createObjectNode();
+        body.put("model", runtime == null || isBlank(runtime.modelName()) ? QWEN_TTS_MODEL : runtime.modelName());
+        body.set("input", input);
+        return body;
+    }
+
+    static URI dashScopeSynthesisUri(String endpoint) {
+        URI configured = URI.create(endpoint == null ? "" : endpoint.trim());
+        String scheme = configured.getScheme();
+        String authority = configured.getRawAuthority();
+        if (("https".equalsIgnoreCase(scheme) || "http".equalsIgnoreCase(scheme)) && authority != null && !authority.isBlank()) {
+            return URI.create(scheme + "://" + authority + "/api/v1/services/audio/tts/SpeechSynthesizer");
+        }
+        throw new IllegalArgumentException("invalid DashScope endpoint");
+    }
+
+    static Optional<String> dashScopeAudioUrl(ObjectMapper objectMapper, String responseBody) {
+        try {
+            String value = objectMapper.readTree(responseBody == null ? "" : responseBody)
+                .path("output").path("audio").path("url").asText("").trim();
+            URI uri = value.isBlank() ? null : URI.create(value);
+            if (uri == null || !("http".equalsIgnoreCase(uri.getScheme()) || "https".equalsIgnoreCase(uri.getScheme()))
+                || uri.getHost() == null || !uri.getHost().endsWith(".oss-cn-beijing.aliyuncs.com")) {
+                return Optional.empty();
+            }
+            return Optional.of(uri.toString());
+        } catch (Exception e) {
+            return Optional.empty();
+        }
+    }
+
+    private Optional<byte[]> downloadDashScopeAudio(String url) {
+        try {
+            HttpRequest request = HttpRequest.newBuilder()
+                .uri(URI.create(url))
+                .timeout(DASHSCOPE_TTS_TIMEOUT)
+                .GET()
+                .build();
+            HttpResponse<byte[]> response = client().send(request, HttpResponse.BodyHandlers.ofByteArray());
+            if (response.statusCode() < 200 || response.statusCode() >= 300 || response.body().length == 0) {
+                return Optional.empty();
+            }
+            return Optional.of(response.body());
+        } catch (Exception e) {
+            return Optional.empty();
+        }
+    }
+
+    private static VoiceProfile withLegacyVoice(VoiceProfile profile, String legacyVoice) {
+        if (safeVoice(profile.voice()) != null || safeVoice(legacyVoice) == null) {
+            return profile;
+        }
+        return new VoiceProfile(profile.role(), legacyVoice, profile.speed(), profile.emotion(), profile.dialect());
+    }
+
     /**
      * 硅基流动 voice 格式为 "{model}:{voice}"；调用方只传短名（如 anna/粤语音色名）时自动补模型前缀。
      */
@@ -196,7 +345,7 @@ public class AihrSpeechService {
         if (value.isBlank() || DIALECT_PROMPTS.containsKey(value)) {
             return value;
         }
-        throw new IllegalArgumentException("dialect 仅支持 mandarin、cantonese、sichuanese");
+        throw new IllegalArgumentException("dialect 不受支持");
     }
 
     private static String expressivePrompt(String emotion, String dialect) {
@@ -236,6 +385,12 @@ public class AihrSpeechService {
             && runtime.modelName().toLowerCase(Locale.ROOT).contains("cosyvoice");
     }
 
+    private static boolean isDashScopeQwenTts(SpeechModel runtime) {
+        return runtime != null
+            && ("dashscope".equalsIgnoreCase(runtime.providerCode()) || "qianwen".equalsIgnoreCase(runtime.providerCode()))
+            && QWEN_TTS_MODEL.equals(runtime.modelName());
+    }
+
     private HttpRequest.Builder authorized(SpeechModel runtime, String path) {
         HttpRequest.Builder builder = HttpRequest.newBuilder()
             .uri(URI.create(normalizeBaseUrl(runtime.baseUrl()) + path))
@@ -286,6 +441,10 @@ public class AihrSpeechService {
         return normalized;
     }
 
+    private static boolean isBlank(String value) {
+        return value == null || value.isBlank();
+    }
+
     private static String truncate(String value) {
         if (value == null || value.length() <= 240) {
             return value;

backend/ruoyi-modules/ruoyi-aihr/src/test/java/org/dromara/aihr/service/AihrSpeechServiceTest.java

@@ -32,6 +32,9 @@ class AihrSpeechServiceTest {
             new VoiceProfile("customer", null, 9.0, "intense", "sichuanese"));
         var legacy = AihrSpeechService.buildRequestBody(
             mapper, runtime, "普通播报", "alex", null);
+        var qwenVoiceBeforeSwitch = AihrSpeechService.buildRequestBody(
+            mapper, runtime, "普通播报", null,
+            new VoiceProfile("neutral", "longanhuan_v3.6", null, null, null));
 
         assertEquals("speech:shifu-warm-v1:cm3hz4wfz02jy106j6z6muix7:ysvbyzypjgmnceokxedr", mentor.path("voice").asText());
         assertEquals(0.88, mentor.path("speed").asDouble());
@@ -41,6 +44,7 @@ class AihrSpeechServiceTest {
         assertTrue(customer.path("input").asText().contains("情绪强烈"));
         assertTrue(customer.path("input").asText().contains("四川话"));
         assertEquals("FunAudioLLM/CosyVoice2-0.5B:alex", legacy.path("voice").asText());
+        assertEquals("FunAudioLLM/CosyVoice2-0.5B:anna", qwenVoiceBeforeSwitch.path("voice").asText());
         assertFalse(legacy.has("speed"));
         assertFalse(legacy.path("input").asText().contains("<|endofprompt|>"));
     }
@@ -60,4 +64,45 @@ class AihrSpeechServiceTest {
         assertThrows(IllegalArgumentException.class, () -> AihrSpeechService.buildRequestBody(
             mapper, runtime, "测试", null, invalid));
     }
+
+    @Test
+    void buildsDashScopeQwenTtsRequestAndNormalizesWorkspaceEndpoint() {
+        ObjectMapper mapper = new ObjectMapper();
+        SpeechModel runtime = new SpeechModel(
+            "dashscope",
+            "qwen-audio-3.0-tts-flash",
+            "",
+            ""
+        );
+
+        var body = AihrSpeechService.buildDashScopeRequestBody(
+            mapper, runtime, "请先确认业主的诉求。",
+            new VoiceProfile("mentor", null, 0.88, "warm", "shanghai"));
+
+        assertEquals("qwen-audio-3.0-tts-flash", body.path("model").asText());
+        assertEquals("请先确认业主的诉求。", body.path("input").path("text").asText());
+        assertEquals("longanhuan_v3.6", body.path("input").path("voice").asText());
+        assertEquals("mp3", body.path("input").path("format").asText());
+        assertEquals(24000, body.path("input").path("sample_rate").asInt());
+        assertEquals(0.88, body.path("input").path("rate").asDouble());
+        assertTrue(body.path("input").path("instruction").asText().contains("上海话"));
+        assertEquals(
+            "https://ws-example.cn-beijing.maas.aliyuncs.com/api/v1/services/audio/tts/SpeechSynthesizer",
+            AihrSpeechService.dashScopeSynthesisUri("https://ws-example.cn-beijing.maas.aliyuncs.com/compatible-mode/v1").toString()
+        );
+    }
+
+    @Test
+    void extractsDashScopeAudioUrlOnlyWhenPresent() throws Exception {
+        ObjectMapper mapper = new ObjectMapper();
+        String response = """
+            {"output":{"audio":{"url":"http://dashscope-result-bj.oss-cn-beijing.aliyuncs.com/audio.mp3?sig=ok"}}}
+            """;
+
+        assertEquals(
+            "http://dashscope-result-bj.oss-cn-beijing.aliyuncs.com/audio.mp3?sig=ok",
+            AihrSpeechService.dashScopeAudioUrl(mapper, response).orElseThrow()
+        );
+        assertTrue(AihrSpeechService.dashScopeAudioUrl(mapper, "{\"output\":{}}").isEmpty());
+    }
 }

mobile-uni/src/pages/user/profile/index.vue

@@ -271,6 +271,25 @@
         </view>
         <text class="display-note">切换后同步到今日、练、问、我全部页面，并自动保存。</text>
       </template>
+
+      <view v-if="loggedIn" class="voice-preference-row">
+        <view class="display-heading-copy">
+          <text class="menu-item-title">AI 播报音色</text>
+          <text class="menu-item-desc">{{ currentTtsVoiceLabel }} · 对练情绪和语速仍会保留</text>
+        </view>
+        <picker :range="qwenTtsVoiceOptions" range-key="label" :value="ttsVoiceOptionIndex" @change="changeTtsVoice">
+          <view class="voice-picker-value">{{ currentTtsVoiceLabel }} <uni-icons type="right" size="12" color="#94a3b8" /></view>
+        </picker>
+      </view>
+      <view v-if="loggedIn" class="voice-preference-row">
+        <view class="display-heading-copy">
+          <text class="menu-item-title">AI 播报方言</text>
+          <text class="menu-item-desc">{{ currentTtsDialectLabel }} · 仅影响服务端语音，失败时不退回普通话</text>
+        </view>
+        <picker :range="qwenTtsDialectOptions" range-key="label" :value="ttsDialectOptionIndex" @change="changeTtsDialect">
+          <view class="voice-picker-value">{{ currentTtsDialectLabel }} <uni-icons type="right" size="12" color="#94a3b8" /></view>
+        </picker>
+      </view>
     </view>
   </view>
 </template>
@@ -297,6 +316,16 @@ import {
 import { getCompetencyProfile, getPracticeHistory, getPromotionEvidence } from '@/services/practice';
 import { resetPageScroll } from '@/services/navigation';
 import { ensureEmployeePosition } from '@/services/position';
+import {
+  getTtsVoicePreference,
+  getTtsDialectPreference,
+  qwenTtsDialectOptions,
+  qwenTtsVoiceOptions,
+  setTtsDialectPreference,
+  setTtsVoicePreference,
+  type TtsDialectPreference,
+  type TtsVoicePreference
+} from '@/services/tts-voice-preference';
 
 const loggedIn = ref(false);
 const phone = ref('');
@@ -313,9 +342,19 @@ const evidence = ref<PromotionEvidence | null>(null);
 const positionProfile = ref(getSelectedPositionProfile());
 const supervisorLearning = ref(false);
 const fontSize = ref<AppFontSize>(getAppFontSize());
+const ttsVoice = ref<TtsVoicePreference>('');
+const ttsDialect = ref<TtsDialectPreference>('');
 const currentFontSizeLabel = computed(() =>
   appFontSizeOptions.find((item) => item.value === fontSize.value)?.label || '标准'
 );
+const ttsVoiceOptionIndex = computed(() => Math.max(0, qwenTtsVoiceOptions.findIndex((item) => item.value === ttsVoice.value)));
+const currentTtsVoiceLabel = computed(() =>
+  qwenTtsVoiceOptions[ttsVoiceOptionIndex.value]?.label || '跟随场景'
+);
+const ttsDialectOptionIndex = computed(() => Math.max(0, qwenTtsDialectOptions.findIndex((item) => item.value === ttsDialect.value)));
+const currentTtsDialectLabel = computed(() =>
+  qwenTtsDialectOptions[ttsDialectOptionIndex.value]?.label || '跟随场景'
+);
 const maskedPhone = computed(() => {
   const value = phone.value.trim();
   if (!value) return '未登录';
@@ -333,6 +372,8 @@ const refreshAuth = () => {
   positionProfile.value = getSelectedPositionProfile();
   supervisorLearning.value = isSupervisorLearnerMode();
   fontSize.value = getAppFontSize();
+  ttsVoice.value = getTtsVoicePreference(auth.tenantId, auth.phone);
+  ttsDialect.value = getTtsDialectPreference(auth.tenantId, auth.phone);
 };
 
 const changeFontSize = (value: AppFontSize) => {
@@ -340,6 +381,20 @@ const changeFontSize = (value: AppFontSize) => {
   uni.showToast({ title: `已切换为${currentFontSizeLabel.value}`, icon: 'none' });
 };
 
+const changeTtsVoice = (event: { detail?: { value?: string | number } }) => {
+  const option = qwenTtsVoiceOptions[Number(event.detail?.value)];
+  const auth = getAuth();
+  ttsVoice.value = setTtsVoicePreference(auth.tenantId, auth.phone, option?.value);
+  uni.showToast({ title: `已切换为${currentTtsVoiceLabel.value}`, icon: 'none' });
+};
+
+const changeTtsDialect = (event: { detail?: { value?: string | number } }) => {
+  const option = qwenTtsDialectOptions[Number(event.detail?.value)];
+  const auth = getAuth();
+  ttsDialect.value = setTtsDialectPreference(auth.tenantId, auth.phone, option?.value);
+  uni.showToast({ title: `已切换为${currentTtsDialectLabel.value}`, icon: 'none' });
+};
+
 const returnToSupervisor = () => {
   leaveSupervisorLearnerMode();
   supervisorLearning.value = false;
@@ -650,6 +705,28 @@ onShow(() => {
   box-shadow: 0 4px 12px rgba(24, 34, 48, 0.08);
 }
 
+.voice-preference-row {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 12px;
+  padding-top: 12px;
+  border-top: 1px solid var(--employee-line);
+}
+
+.voice-picker-value {
+  display: flex;
+  align-items: center;
+  gap: 3px;
+  max-width: 150px;
+  padding: 8px 10px;
+  border: 1px solid #e3e6ea;
+  border-radius: 10px;
+  color: var(--employee-text);
+  font-size: 12px;
+  white-space: nowrap;
+}
+
 .profile-summary {
   border-color: var(--employee-line);
 }

mobile-uni/src/services/speech.ts

@@ -1,6 +1,8 @@
 import type { AsrResponse, PracticeTtsContext, SpeechPlaybackStatus, SpeechSelectedFile, SpeechVoiceProfile, TtsResponse } from '@/types/api';
 import { apiRequest, apiUrl, authHeaders, readTextPayload } from './api';
+import { getAuth } from './auth';
 import type { SpeechCapture } from './speech-capture';
+import { applyTtsVoicePreference, getTtsDialectPreference, getTtsVoicePreference } from './tts-voice-preference';
 
 const audioExtensions = ['.mp3', '.wav', '.m4a', '.webm', '.aac', '.ogg'];
 
@@ -103,11 +105,20 @@ export const transcribeSpeechCapture = (
   ? transcribeSpeechBlob(capture.blob, filename, registerAbort)
   : transcribeSpeechFile(capture.file, registerAbort);
 
+export const resolveTtsVoiceProfile = (voiceProfile?: SpeechVoiceProfile) => {
+  const auth = getAuth();
+  return applyTtsVoicePreference(
+    voiceProfile,
+    getTtsVoicePreference(auth.tenantId, auth.phone),
+    getTtsDialectPreference(auth.tenantId, auth.phone)
+  );
+};
+
 export const synthesizeSpeech = (text: string, voiceProfile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext) =>
   apiRequest<TtsResponse>({
     url: '/api/ai/tts',
     method: 'POST',
-    data: { text: text.trim().slice(0, 300), voiceProfile, practiceContext },
+    data: { text: text.trim().slice(0, 300), voiceProfile: resolveTtsVoiceProfile(voiceProfile), practiceContext },
     timeout: 60000
   });
 
@@ -118,7 +129,7 @@ export interface SpeechPlaybackSnapshot {
 
 const browserDialectLanguage = (dialect: SpeechVoiceProfile['dialect']) => {
   if (dialect === 'cantonese') return 'zh-HK';
-  if (dialect === 'sichuanese') return '';
+  if (dialect && dialect !== 'mandarin') return '';
   return 'zh-CN';
 };
 
@@ -221,8 +232,8 @@ export const createSpeechPlaybackController = (
     stop();
     const value = text.trim();
     if (!value) return;
-    const profile = voiceProfile || defaultVoiceProfile;
-    const sourceKey = cacheKey(value, profile, practiceContext);
+    const effectiveProfile = resolveTtsVoiceProfile(voiceProfile || defaultVoiceProfile);
+    const sourceKey = cacheKey(value, effectiveProfile, practiceContext);
     const currentGeneration = generation;
     publish(key, 'loading');
     try {
@@ -233,7 +244,7 @@ export const createSpeechPlaybackController = (
         playAudioSource(key, cachedSource, sourceKey, currentGeneration);
         return;
       }
-      const result = await synthesizeSpeech(value, profile, practiceContext);
+      const result = await synthesizeSpeech(value, effectiveProfile, practiceContext);
       if (currentGeneration !== generation || activeKey !== key) return;
       const source = result.inlineAudioUrl || result.audioUrl;
       if (!source) throw new Error('语音合成未返回音频');
@@ -246,7 +257,7 @@ export const createSpeechPlaybackController = (
       playAudioSource(key, source, sourceKey, currentGeneration);
     } catch (error) {
       if (currentGeneration !== generation) return;
-      const browserFallback = playWithBrowserSpeech(key, value, currentGeneration, profile);
+      const browserFallback = playWithBrowserSpeech(key, value, currentGeneration, effectiveProfile);
       if (browserFallback.played) return;
       stop();
       onError(browserFallback.message || (error instanceof Error ? error.message : '语音生成失败'));
@@ -261,12 +272,12 @@ export const createSpeechPlaybackController = (
   const preload = async (text: string, voiceProfile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext): Promise<string | null> => {
     const value = text.trim();
     if (!value) return null;
-    const profile = voiceProfile || defaultVoiceProfile;
-    const sourceKey = cacheKey(value, profile, practiceContext);
+    const effectiveProfile = resolveTtsVoiceProfile(voiceProfile || defaultVoiceProfile);
+    const sourceKey = cacheKey(value, effectiveProfile, practiceContext);
     const cached = sourceCache.get(sourceKey);
     if (cached) return cached;
     try {
-      const result = await synthesizeSpeech(value, profile, practiceContext);
+      const result = await synthesizeSpeech(value, effectiveProfile, practiceContext);
       const source = result.inlineAudioUrl || result.audioUrl;
       if (!source) return null;
       sourceCache.set(sourceKey, source);

mobile-uni/src/services/tts-voice-preference.ts

@@ -0,0 +1,96 @@
+import type { SpeechVoiceProfile } from '@/types/api';
+
+export const qwenTtsVoiceOptions = [
+  { value: '', label: '跟随场景' },
+  { value: 'longanhuan_v3.6', label: '龙安欢 · 中文女声' },
+  { value: 'longjielidou_v3.6', label: '龙杰力豆 · 童声' },
+  { value: 'loongeva_v3.6', label: 'Loongeva · 英文女声' },
+  { value: 'loongjohn', label: 'loongJohn · 英文男声' }
+] as const;
+
+export const qwenTtsDialectOptions = [
+  { value: '', label: '跟随场景' },
+  { value: 'mandarin', label: '普通话' },
+  { value: 'cantonese', label: '粤语' },
+  { value: 'chongqing', label: '重庆话' },
+  { value: 'northeastern', label: '东北话' },
+  { value: 'gansu', label: '甘肃话' },
+  { value: 'guizhou', label: '贵州话' },
+  { value: 'zhejiang', label: '浙江话' },
+  { value: 'hebei', label: '河北话' },
+  { value: 'henan', label: '河南话' },
+  { value: 'hubei', label: '湖北话' },
+  { value: 'hunan', label: '湖南话' },
+  { value: 'jiangxi', label: '江西话' },
+  { value: 'ningbo', label: '宁波话' },
+  { value: 'ningxia', label: '宁夏话' },
+  { value: 'qingdao', label: '青岛话' },
+  { value: 'shaanxi', label: '陕西话' },
+  { value: 'shanxi', label: '山西话' },
+  { value: 'shandong', label: '山东话' },
+  { value: 'shanghai', label: '上海话' },
+  { value: 'sichuanese', label: '四川话' },
+  { value: 'yunnan', label: '云南话' }
+] as const;
+
+export type TtsVoicePreference = typeof qwenTtsVoiceOptions[number]['value'];
+export type TtsDialectPreference = typeof qwenTtsDialectOptions[number]['value'];
+
+const preferenceKeyPrefix = 'aihr_tts_voice';
+const dialectPreferenceKeyPrefix = 'aihr_tts_dialect';
+
+export const normalizeTtsVoicePreference = (value: unknown): TtsVoicePreference =>
+  qwenTtsVoiceOptions.some((option) => option.value === value) ? value as TtsVoicePreference : '';
+
+export const normalizeTtsDialectPreference = (value: unknown): TtsDialectPreference =>
+  qwenTtsDialectOptions.some((option) => option.value === value) ? value as TtsDialectPreference : '';
+
+export const ttsVoicePreferenceStorageKey = (tenantId: unknown, phone: unknown) => {
+  const tenant = String(tenantId || '').trim();
+  const account = String(phone || '').trim();
+  return tenant && account ? `${preferenceKeyPrefix}:${tenant}:${account}` : '';
+};
+
+export const getTtsVoicePreference = (tenantId: unknown, phone: unknown): TtsVoicePreference => {
+  const key = ttsVoicePreferenceStorageKey(tenantId, phone);
+  return key ? normalizeTtsVoicePreference(uni.getStorageSync(key)) : '';
+};
+
+const ttsDialectPreferenceStorageKey = (tenantId: unknown, phone: unknown) => {
+  const tenant = String(tenantId || '').trim();
+  const account = String(phone || '').trim();
+  return tenant && account ? `${dialectPreferenceKeyPrefix}:${tenant}:${account}` : '';
+};
+
+export const getTtsDialectPreference = (tenantId: unknown, phone: unknown): TtsDialectPreference => {
+  const key = ttsDialectPreferenceStorageKey(tenantId, phone);
+  return key ? normalizeTtsDialectPreference(uni.getStorageSync(key)) : '';
+};
+
+export const setTtsVoicePreference = (tenantId: unknown, phone: unknown, value: unknown): TtsVoicePreference => {
+  const key = ttsVoicePreferenceStorageKey(tenantId, phone);
+  const normalized = normalizeTtsVoicePreference(value);
+  if (key) uni.setStorageSync(key, normalized);
+  return normalized;
+};
+
+export const setTtsDialectPreference = (tenantId: unknown, phone: unknown, value: unknown): TtsDialectPreference => {
+  const key = ttsDialectPreferenceStorageKey(tenantId, phone);
+  const normalized = normalizeTtsDialectPreference(value);
+  if (key) uni.setStorageSync(key, normalized);
+  return normalized;
+};
+
+export const applyTtsVoicePreference = (
+  profile: SpeechVoiceProfile | undefined,
+  preference: unknown,
+  dialectPreference?: unknown
+): SpeechVoiceProfile | undefined => {
+  const voice = normalizeTtsVoicePreference(preference);
+  const dialect = normalizeTtsDialectPreference(dialectPreference);
+  if (!voice && !dialect) return profile;
+  const next: SpeechVoiceProfile = { ...(profile || { role: 'neutral' }) };
+  if (voice) next.voice = voice;
+  if (dialect) next.dialect = dialect;
+  return next;
+};

mobile-uni/src/types/api.ts

@@ -45,7 +45,7 @@ export interface SpeechVoiceProfile {
   voice?: string;
   speed?: number;
   emotion?: 'warm' | 'professional' | 'serious' | 'intense' | 'calm';
-  dialect?: 'mandarin' | 'cantonese' | 'sichuanese';
+  dialect?: 'mandarin' | 'cantonese' | 'chongqing' | 'northeastern' | 'gansu' | 'guizhou' | 'zhejiang' | 'hebei' | 'henan' | 'hubei' | 'hunan' | 'jiangxi' | 'ningbo' | 'ningxia' | 'qingdao' | 'shaanxi' | 'shanxi' | 'shandong' | 'shanghai' | 'sichuanese' | 'yunnan';
 }
 
 export interface ToolItem {
@@ -582,6 +582,13 @@ export interface PracticeScenarioOption {
   enabled?: boolean;
   position?: string;
   scenarioType?: string;
+  projectType?: string;
+  growthLevel?: string;
+  competencyCode?: string;
+  collaborationPositions?: string;
+  reviewStatus?: string;
+  curriculumVersion?: string;
+  difficulty?: number;
 }
 
 export interface ReviewDialogue {
@@ -601,6 +608,12 @@ export interface ReviewAnnotation {
 }
 
 export interface ReviewDetail extends PracticeRecord {
+  position?: string;
+  projectType?: string;
+  growthLevel?: string;
+  competencyCode?: string;
+  collaborationPositions?: string;
+  curriculumVersion?: string;
   mentorRewrite: string;
   aiComment: string;
   reviewAdvice?: string;
@@ -1087,6 +1100,16 @@ export interface CompetencyProfile {
   aiLevel?: string;
   dimensions: CompetencyDimension[];
   growthPath?: GrowthStage[];
+  capabilityProgress?: CapabilityProgress[];
+}
+
+export interface CapabilityProgress {
+  position: string;
+  growthLevel: string;
+  competencyCode: string;
+  completed: number;
+  averageScore: number;
+  status: string;
 }
 
 export interface PromotionEvidence {

mobile-uni/tests/tts-voice-preference.test.mjs

@@ -0,0 +1,57 @@
+import assert from 'node:assert/strict';
+import { readFile } from 'node:fs/promises';
+import test from 'node:test';
+import vm from 'node:vm';
+import ts from 'typescript';
+
+const source = (path) => readFile(new URL(path, import.meta.url), 'utf8');
+
+const preferenceRuntime = async (storage = new Map()) => {
+  const service = await source('../src/services/tts-voice-preference.ts');
+  const compiled = ts.transpileModule(service, {
+    compilerOptions: { module: ts.ModuleKind.CommonJS, target: ts.ScriptTarget.ES2020 }
+  });
+  const runtimeModule = { exports: {} };
+  vm.runInNewContext(compiled.outputText, {
+    exports: runtimeModule.exports,
+    module: runtimeModule,
+    uni: {
+      getStorageSync: (key) => storage.get(key),
+      setStorageSync: (key, value) => storage.set(key, value)
+    }
+  });
+  return runtimeModule.exports;
+};
+
+test('播报音色按租户和账号隔离保存，非法值回退跟随场景', async () => {
+  const storage = new Map();
+  const runtime = await preferenceRuntime(storage);
+
+  assert.equal(runtime.getTtsVoicePreference('000000', '13800000000'), '');
+  assert.equal(runtime.setTtsVoicePreference('000000', '13800000000', 'longanhuan_v3.6'), 'longanhuan_v3.6');
+  assert.equal(runtime.getTtsVoicePreference('000000', '13800000000'), 'longanhuan_v3.6');
+  assert.equal(runtime.getTtsVoicePreference('000000', '13900000000'), '');
+  assert.equal(runtime.setTtsVoicePreference('000000', '13800000000', 'untrusted-voice'), '');
+  assert.equal(runtime.setTtsDialectPreference('000000', '13800000000', 'shanghai'), 'shanghai');
+  assert.equal(runtime.getTtsDialectPreference('000000', '13800000000'), 'shanghai');
+});
+
+test('全局选择只覆盖 voice，保留场景角色、情绪、语速和方言', async () => {
+  const runtime = await preferenceRuntime();
+  const profile = { role: 'customer', emotion: 'intense', speed: 1.12, dialect: 'sichuanese' };
+
+  assert.deepEqual(
+    JSON.parse(JSON.stringify(runtime.applyTtsVoicePreference(profile, 'longanhuan_v3.6', 'shanghai'))),
+    { ...profile, voice: 'longanhuan_v3.6', dialect: 'shanghai' }
+  );
+  assert.deepEqual(JSON.parse(JSON.stringify(runtime.applyTtsVoicePreference(profile, ''))), profile);
+});
+
+test('语音播放链路在合成前统一注入全局音色，且缓存键包含处理后的 profile', async () => {
+  const sourceText = await source('../src/services/speech.ts');
+
+  assert.match(sourceText, /applyTtsVoicePreference/);
+  assert.match(sourceText, /getTtsVoicePreference/);
+  assert.match(sourceText, /cacheKey\(value, effectiveProfile, practiceContext\)/);
+  assert.match(sourceText, /synthesizeSpeech\(value, effectiveProfile, practiceContext\)/);
+});