feat: smart chunking to avoid chunk boundaries mid-sentence

1 files changed, 50 insertions, 3 deletions

diff --git a/app/tests/test_hub.py b/app/tests/test_hub.py
index 90ed2d4..cbcdc8a 100644
--- a/app/tests/test_hub.py
+++ b/app/tests/test_hub.py
@@ -3529,7 +3529,8 @@ class SettingsTests(unittest.TestCase):
     # Keys _apply_settings persists; every test that triggers a real or
     # fake config write restores these afterwards.
     _SETTING_KEYS = ("AUDIO_FORMAT", "AUDIO_BITRATE", "LANGUAGE",
-                     "CHUNK_SIZE", "INPUT_DIR", "OUTPUT_DIR",
+                     "CHUNK_SIZE", "SMART_CHUNKING", "INPUT_DIR",
+                     "OUTPUT_DIR",
                      "CLONE_WAV_DIR",
                      "SPEED", "DEBUG", "STOP_SERVER_AND_EXIT",
                      "AUDIOCPP_UNLOAD_MODELS",
@@ -3604,6 +3605,7 @@ class SettingsTests(unittest.TestCase):
                         original_folders[1])
         values = {"audio_format": "ogg", "audio_bitrate": " 192k ",
                   "language": "en", "chunk_size": "300",
+                  "smart_chunking": True,
                   "input_dir": " /books ", "output_dir": "/audiobooks",
                   "clone_wav_dir": " /refs/wavs ",
                   "speed": "1.5", "debug": True,
@@ -3626,6 +3628,7 @@ class SettingsTests(unittest.TestCase):
                                    "AUDIO_BITRATE": "192k",
                                    "LANGUAGE": "English",
                                    "CHUNK_SIZE": 300,
+                                   "SMART_CHUNKING": True,
                                    "INPUT_DIR": "/books",
                                    "OUTPUT_DIR": "/audiobooks",
                                    "CLONE_WAV_DIR": "/refs/wavs",
@@ -3652,6 +3655,7 @@ class SettingsTests(unittest.TestCase):
         self.assertEqual(hub.config.AUDIO_BITRATE, "192k")
         self.assertEqual(hub.config.LANGUAGE, "English")
         self.assertEqual(hub.config.CHUNK_SIZE, 300)
+        self.assertEqual(hub.config.SMART_CHUNKING, True)
         self.assertEqual(hub.config.INPUT_DIR, "/books")
         self.assertEqual(hub.config.OUTPUT_DIR, "/audiobooks")
         self.assertEqual(hub.config.CLONE_WAV_DIR, "/refs/wavs")
@@ -3671,6 +3675,7 @@ class SettingsTests(unittest.TestCase):
         self._snapshot_settings()
         base = {"audio_format": "m4b", "audio_bitrate": "128k",
                 "language": "English", "chunk_size": "250",
+                "smart_chunking": True,
                 "input_dir": "./input", "output_dir": "./output",
                 "clone_wav_dir": "./voices",
                 "speed": "1.0", "debug": False,
@@ -3701,9 +3706,46 @@ class SettingsTests(unittest.TestCase):
                                      "audiocpp_remote_url": "not a url"})
             mk_update.assert_not_called()
 
+    def test_smart_chunking_toggle_persists(self):
+        self._snapshot_settings()
+
+        def fake_update(key, value, config_path=None):
+            setattr(hub.config, key, value)
+            return True
+
+        values = {"audio_format": "m4b", "audio_bitrate": "128k",
+                  "language": "English", "chunk_size": "250",
+                  "smart_chunking": False,
+                  "input_dir": "./input", "output_dir": "./output",
+                  "clone_wav_dir": "./voices",
+                  "speed": "1.0", "debug": False,
+                  "stop_and_exit": True, "unload_models": True,
+                  "qwen_port": "7860",
+                  "faster_port": "8000", "audiocpp_port": "8080",
+                  "sglomni_port": "8100"}
+        with patch.object(hub.common, "update_config_value", fake_update), \
+                patch.object(hub, "_sync_audiocpp_server_port"):
+            hub._apply_settings(values)
+        self.assertEqual(hub.config.SMART_CHUNKING, False)
+
+        values["smart_chunking"] = True
+        with patch.object(hub.common, "update_config_value", fake_update), \
+                patch.object(hub, "_sync_audiocpp_server_port"):
+            hub._apply_settings(values)
+        self.assertEqual(hub.config.SMART_CHUNKING, True)
+
+    def test_smart_chunking_field_defaults_on_with_help(self):
+        fields = hub._settings_fields()
+        field = next(f for f in fields if f["key"] == "smart_chunking")
+        self.assertEqual(field["kind"], "bool")
+        self.assertEqual(field["value"], hub.config.SMART_CHUNKING)
+        self.assertTrue(field["help"])
+        self.assertTrue(field["label"], "Smart chunking")
+
     def test_field_validators(self):
         self.assertIsNone(hub._validate_bitrate("128k"))
         self.assertIsNotNone(hub._validate_bitrate("   "))
+        self.assertIsNotNone(hub._validate_bitrate("   "))
         self.assertIsNone(hub._validate_language("English"))
         self.assertIsNone(hub._validate_language("en"))
         self.assertIsNotNone(hub._validate_language("Klingon"))
@@ -3794,6 +3836,7 @@ class SettingsTests(unittest.TestCase):
             captured["fields"] = fields
             return {"audio_format": "ogg", "audio_bitrate": "192k",
                     "language": "English", "chunk_size": "300",
+                    "smart_chunking": True,
                     "input_dir": "/books", "output_dir": "/audiobooks",
                     "clone_wav_dir": "/refs/wavs",
                     "speed": "1.0", "debug": False,
@@ -3817,7 +3860,8 @@ class SettingsTests(unittest.TestCase):
             hub._Hub(None).screen_settings()
         self.assertEqual([f["key"] for f in captured["fields"]],
                          ["audio_format", "audio_bitrate", "language",
-                          "chunk_size", "input_dir", "output_dir",
+                          "chunk_size", "smart_chunking", "input_dir",
+                          "output_dir",
                           "clone_wav_dir",
                           "speed", "debug", "stop_and_exit",
                           "unload_models",
@@ -3867,6 +3911,7 @@ class SettingsTests(unittest.TestCase):
                                     "audio_bitrate": "192k",
                                     "language": "English",
                                     "chunk_size": "300",
+                                    "smart_chunking": True,
                                     "input_dir": "/books",
                                     "output_dir": "/audiobooks",
                                     "clone_wav_dir": "/refs/wavs",
@@ -4020,7 +4065,8 @@ class SettingsTests(unittest.TestCase):
 
         original = {name: getattr(hub.config, name) for name in
                     ("AUDIO_FORMAT", "AUDIO_BITRATE", "LANGUAGE",
-                     "CHUNK_SIZE", "INPUT_DIR", "OUTPUT_DIR",
+                     "CHUNK_SIZE", "SMART_CHUNKING", "INPUT_DIR",
+                     "OUTPUT_DIR",
                      "CLONE_WAV_DIR",
                      "SPEED", "DEBUG", "STOP_SERVER_AND_EXIT",
                      "AUDIOCPP_UNLOAD_MODELS",
@@ -4042,6 +4088,7 @@ class SettingsTests(unittest.TestCase):
                 'LANGUAGE = "English"\n'
                 "\n"
                 "CHUNK_SIZE = 250\n"
+                "SMART_CHUNKING = True\n"
                 'INPUT_DIR = "./input"\n'
                 'OUTPUT_DIR = "./output"\n'
                 'CLONE_WAV_DIR = "./voices"\n'