test: add qwen tts adapter and voice preference reproducers

2026-07-23 22:21:52 +08:00
parent bf82ff9691
commit 39ff6f6ceb
2 changed files with 95 additions and 0 deletions
@@ -60,4 +60,44 @@ 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", "mandarin"));
+
+        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());
+        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());
+    }
 }
@@ -0,0 +1,55 @@
+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'), '');
+});
+
+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(
+    runtime.applyTtsVoicePreference(profile, 'longanhuan_v3.6'),
+    { ...profile, voice: 'longanhuan_v3.6' }
+  );
+  assert.deepEqual(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\(value, effectiveProfile, practiceContext\)/);
+});