fix: stabilize translation and audio generation

Validate translation responses and fail over between free providers. Fix logical audio URLs, bounded TTS generation, retries, and regression coverage for v1.2.1.

Author: Pi3-l22

app.py

@@ -166,20 +166,27 @@ async def generate(request: Request, text: str = Form(""), voice: str = Form("")
 
         # 翻译句子
         lang_code = lang if lang in LANGUAGE_CODES else "en"
-        translated_sentences = get_text_translated(sentences, from_lang=lang_code)
+        translated_sentences, translation_warning = get_text_translated(
+            sentences, from_lang=lang_code
+        )
 
         # 生成音频
         voice_name = voice if voice else "en-US-ChristopherNeural"
         audio_dir, audio_filenames, warning_msg = await generate_audio(sentences, voice_name, title)
 
-        # 构建结果列表
+        # 构建结果列表。结果中只保存可被 HTTP/导出页面使用的逻辑路径，
+        # 绝不把本机磁盘绝对路径暴露给浏览器。
         results = []
         for i, (sentence, translation, audio_filename) in enumerate(zip(sentences, translated_sentences, audio_filenames)):
-            audio_path = Path(audio_dir) / audio_filename
+            audio_path = None
+            if audio_filename:
+                audio_path = "/".join(
+                    (Path(AUDIO_DIR).name, title, audio_filename)
+                )
             results.append({
                 "sentence": sentence,
                 "translation": translation,
-                "audio_path": str(audio_path).replace("\\", "/")  # 确保路径分隔符统一
+                "audio_path": audio_path,
             })
 
         # 清理临时文件
@@ -196,13 +203,10 @@ async def generate(request: Request, text: str = Form(""), voice: str = Form("")
         CURRENT_UUID = title
 
         # 返回结果页面
-        if warning_msg:
-            return templates.TemplateResponse(request, "results.html", {
-                "results": results,
-                "warning": warning_msg
-            })
+        warnings = [warning for warning in (translation_warning, warning_msg) if warning]
         return templates.TemplateResponse(request, "results.html", {
-            "results": results
+            "results": results,
+            "warning": "；".join(warnings),
         })
 
     except Exception as e:

templates/export_template.html

@@ -51,10 +51,14 @@ <h1><a href="https://github.com/Pi3-l22/TingJu" target="_blank">听句 TingJu</a
             <!-- <input class="user-input" type="text" placeholder="播放音频，听写句子..."> -->
             <textarea class="user-input" placeholder="播放音频，听写句子..." spellcheck="true" autocomplete="off"></textarea>
             <div class=" audio-controls">
+                {% if item.audio_path %}
                 <audio controls class="audio-player">
                     <source src="./{{ item.audio_path }}" type="audio/mpeg">
                     <span class="warning">您的浏览器不支持音频播放，换一个浏览器试试</span>
                 </audio>
+                {% else %}
+                <span class="warning">该句音频生成失败，请稍后重试</span>
+                {% endif %}
             </div>
         </div>
         {% endfor %}
@@ -69,4 +73,4 @@ <h1><a href="https://github.com/Pi3-l22/TingJu" target="_blank">听句 TingJu</a
     <script src="./js/results.js"></script>
 </body>
 
-</html>
+</html>

templates/results.html

@@ -52,10 +52,14 @@ <h1><a href="https://github.com/Pi3-l22/TingJu" target="_blank">听句 TingJu</a
             <!-- <input class="user-input" type="text" placeholder="播放音频，听写句子..."> -->
             <textarea class="user-input" placeholder="播放音频，听写句子..." spellcheck="true" autocomplete="off"></textarea>
             <div class=" audio-controls">
+                {% if item.audio_path %}
                 <audio controls class="audio-player">
                     <source src="/{{ item.audio_path }}" type="audio/mpeg">
                     <span class="warning">您的浏览器不支持音频播放，换一个浏览器试试</span>
                 </audio>
+                {% else %}
+                <span class="warning">该句音频生成失败，请稍后重试</span>
+                {% endif %}
             </div>
         </div>
         {% endfor %}
@@ -71,4 +75,4 @@ <h1><a href="https://github.com/Pi3-l22/TingJu" target="_blank">听句 TingJu</a
     <script src="/static/js/results.js"></script>
 </body>
 
-</html>
+</html>

tests/test_app.py

@@ -2,10 +2,13 @@
 import tempfile
 import os
 import shutil
+import re
 from pathlib import Path
 from fastapi.testclient import TestClient
+from unittest.mock import AsyncMock, patch
 
 from app import app, TEMP_DIR, EXPORT_DIR, AUDIO_DIR, cleanup_temp_files
+from utils.audio_generator import _get_filename
 
 class TestApp(unittest.IsolatedAsyncioTestCase):
     def setUp(self):
@@ -48,7 +51,10 @@ async def test_manual_input(self):
 
     async def test_get_voices(self):
         """测试获取音色列表"""
-        response = self.client.get("/voices")
+        with patch("app.list_voices", new=AsyncMock(return_value=[
+            {"name": "en-US-TestNeural", "gender": "Female", "style": []}
+        ])):
+            response = self.client.get("/voices")
         self.assertEqual(response.status_code, 200)
         data = response.json()
         # 检查返回数据是否包含voices键
@@ -102,11 +108,25 @@ async def test_upload_file_unsupported(self):
     async def test_generate(self):
         """测试生成音频和翻译功能"""
         test_text = "This is a test sentence. This is another test sentence."
-        response = self.client.post("/generate", data={
-            "text": test_text,
-            "voice": "en-US-ChristopherNeural",
-            "lang": "en"
-        })
+
+        async def fake_generate_audio(sentences, voice_name, title):
+            audio_dir = Path(AUDIO_DIR) / title
+            audio_dir.mkdir(parents=True, exist_ok=True)
+            filenames = []
+            for index, sentence in enumerate(sentences, start=1):
+                filename = f"{index}_{_get_filename(sentence)}"
+                (audio_dir / filename).write_bytes(b"\xff\xf3\x64\xc4")
+                filenames.append(filename)
+            return audio_dir, filenames, ""
+
+        with patch("app.get_text_translated", return_value=(
+            ["这是测试句子。", "这是另一个测试句子。"], ""
+        )), patch("app.generate_audio", new=fake_generate_audio):
+            response = self.client.post("/generate", data={
+                "text": test_text,
+                "voice": "en-US-ChristopherNeural",
+                "lang": "en"
+            })
 
         # 检查响应
         self.assertEqual(response.status_code, 200)
@@ -118,6 +138,14 @@ async def test_generate(self):
         self.assertIn("This is a test sentence.", response.text)
         self.assertIn("This is another test sentence.", response.text)
 
+        audio_urls = re.findall(r'<source src="([^"]+)"', response.text)
+        self.assertEqual(len(audio_urls), 2)
+        for audio_url in audio_urls:
+            self.assertTrue(audio_url.startswith("/audios/"))
+            audio_response = self.client.get(audio_url)
+            self.assertEqual(audio_response.status_code, 200)
+            self.assertTrue(audio_response.content.startswith(b"\xff\xf3"))
+
     async def test_generate_empty_text(self):
         """测试生成空文本时的处理"""
         response = self.client.post("/generate", data={
@@ -136,11 +164,24 @@ async def test_export_content(self):
         """测试导出功能"""
         # 先生成一些内容，设置CURRENT_UUID
         test_text = "This is a test sentence. This is another test sentence."
-        response = self.client.post("/generate", data={
-            "text": test_text,
-            "voice": "en-US-ChristopherNeural",
-            "lang": "en"
-        })
+        async def fake_generate_audio(sentences, voice_name, title):
+            audio_dir = Path(AUDIO_DIR) / title
+            audio_dir.mkdir(parents=True, exist_ok=True)
+            filenames = []
+            for index, sentence in enumerate(sentences, start=1):
+                filename = f"{index}_{_get_filename(sentence)}"
+                (audio_dir / filename).write_bytes(b"\xff\xf3\x64\xc4")
+                filenames.append(filename)
+            return audio_dir, filenames, ""
+
+        with patch("app.get_text_translated", return_value=(
+            ["这是测试句子。", "这是另一个测试句子。"], ""
+        )), patch("app.generate_audio", new=fake_generate_audio):
+            response = self.client.post("/generate", data={
+                "text": test_text,
+                "voice": "en-US-ChristopherNeural",
+                "lang": "en"
+            })
 
         self.assertEqual(response.status_code, 200)
 
@@ -164,7 +205,11 @@ async def test_export_content(self):
         self.assertTrue((export_path / "css").exists())
         self.assertTrue((export_path / "js").exists())
         self.assertTrue((export_path / "img").exists())
-        self.assertTrue((export_path / AUDIO_DIR).exists())
+        audios_path = export_path / Path(AUDIO_DIR).name
+        self.assertTrue(audios_path.exists())
+        exported_html = (export_path / "index.html").read_text(encoding="utf-8")
+        self.assertIn("./audios/", exported_html)
+        self.assertNotIn("D:/", exported_html)
 
 if __name__ == "__main__":
-    unittest.main()
+    unittest.main()

tests/test_audio_generator.py

@@ -1,40 +1,90 @@
 import unittest
-import shutil
+import tempfile
 from pathlib import Path
-from utils.audio_generator import list_voices, generate_audio, AUDIO_DIR
+from unittest.mock import AsyncMock, patch
+
+from utils.audio_generator import list_voices, generate_audio
 
 class TestAudioGenerator(unittest.IsolatedAsyncioTestCase):
     async def test_list_voices(self):
         """测试音色获取功能"""
-        voices = await list_voices()
+        class FakeVoices:
+            def find(self, **kwargs):
+                return [{
+                    "ShortName": "en-US-TestNeural",
+                    "Gender": "Female",
+                    "VoiceTag": {
+                        "ContentCategories": [],
+                        "VoicePersonalities": [],
+                    },
+                }]
+
+        async def create_voices():
+            return FakeVoices()
+
+        with patch("utils.audio_generator.VoicesManager.create", new=create_voices):
+            voices = await list_voices()
+
         self.assertIsInstance(voices, list)
-        self.assertGreater(len(voices), 10)
+        self.assertEqual(voices[0]["name"], "en-US-TestNeural")
 
     async def test_generate_audio(self):
         """测试音频生成功能"""
-        text_list = ['It is often said that we are what we repeatedly do.', 
-                    'This simple statement highlights the incredible influence of our daily habits.',
-                    'Habits shape our thoughts, guide our actions, and ultimately determine the kind of life we live.',
-                    'Whether good or bad, habits are powerful forces that quietly direct our future.']
-        voice_name = "en-US-ChristopherNeural"
-        title = "test"
-        audio_dir, filenames, warning_msg = await generate_audio(text_list, voice_name, title)
-        print(audio_dir, filenames, warning_msg)
-        
-        self.assertEqual(len(text_list), len(filenames))
-        self.assertEqual(audio_dir, Path(AUDIO_DIR, title))
-        self.assertTrue(audio_dir.exists())
-        self.assertEqual(warning_msg, "")
-        for i, filename in enumerate(filenames):
-            self.assertTrue(audio_dir.joinpath(filename).exists())
-            self.assertEqual(filename.split("_")[0], str(i+1))
-            self.assertEqual(filename.split(".")[1], "mp3")
-            self.assertEqual(len(filename.split("_")[1].split(".")[0]), 32)
-            
-        # 清理测试目录
-        dir = Path(audio_dir)
-        if dir.exists():
-            shutil.rmtree(dir)
+        class FakeCommunicate:
+            def __init__(self, text, voice):
+                self.text = text
+
+            async def save(self, path):
+                Path(path).write_bytes(b"\xff\xf3\x64\xc4")
+
+        text_list = ["First sentence.", "Second sentence.", "Third sentence."]
+        with tempfile.TemporaryDirectory() as temp_dir, patch(
+            "utils.audio_generator.AUDIO_DIR", Path(temp_dir) / "audios"
+        ), patch("utils.audio_generator.edge_tts.Communicate", FakeCommunicate):
+            audio_dir, filenames, warning_msg = await generate_audio(
+                text_list, "en-US-TestNeural", "test"
+            )
+
+            self.assertEqual(len(text_list), len(filenames))
+            self.assertEqual(audio_dir, Path(temp_dir) / "audios" / "test")
+            self.assertTrue(audio_dir.exists())
+            self.assertEqual(warning_msg, "")
+            for i, filename in enumerate(filenames):
+                self.assertIsNotNone(filename)
+                self.assertTrue(audio_dir.joinpath(filename).exists())
+                self.assertEqual(filename.split("_")[0], str(i + 1))
+                self.assertEqual(filename.split(".")[1], "mp3")
+
+    async def test_invalid_audio_is_retried_and_keeps_alignment(self):
+        attempts = {}
+
+        class FakeCommunicate:
+            def __init__(self, text, voice):
+                self.text = text
+
+            async def save(self, path):
+                attempts[self.text] = attempts.get(self.text, 0) + 1
+                if self.text == "retry" and attempts[self.text] == 1:
+                    raise RuntimeError("temporary")
+                if self.text == "invalid":
+                    Path(path).write_bytes(b"bad!")
+                else:
+                    Path(path).write_bytes(b"\xff\xf3\x64\xc4")
+
+        with tempfile.TemporaryDirectory() as temp_dir, patch(
+            "utils.audio_generator.AUDIO_DIR", Path(temp_dir) / "audios"
+        ), patch("utils.audio_generator.edge_tts.Communicate", FakeCommunicate), patch(
+            "utils.audio_generator.asyncio.sleep", new=AsyncMock()
+        ):
+            audio_dir, filenames, warning_msg = await generate_audio(
+                ["ok", "retry", "invalid"], "en-US-TestNeural", "test"
+            )
+
+        self.assertIsNotNone(filenames[0])
+        self.assertIsNotNone(filenames[1])
+        self.assertIsNone(filenames[2])
+        self.assertEqual(attempts["retry"], 2)
+        self.assertIn("第 3 个音频生成失败", warning_msg)
 
 if __name__ == "__main__":
-    unittest.main()
+    unittest.main()