feat: revamp setup wizard, fix number conversion, improve TTS reliability

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "python-voiceio"
-version = "0.3.1"
+version = "0.3.2"
 description = "Speak → text, locally, instantly."
 readme = "README.md"
 license = "MIT"

tests/test_tts.py

@@ -142,7 +142,7 @@ def test_player_empty_audio():
 
 def test_tts_config_defaults():
     cfg = TTSConfig()
-    assert cfg.enabled is False
+    assert cfg.enabled is True
     assert cfg.engine == "auto"
     assert cfg.hotkey == "ctrl+alt+s"
     assert cfg.voice == ""
@@ -155,4 +155,4 @@ def test_tts_config_in_main_config():
     cfg = Config()
     assert hasattr(cfg, "tts")
     assert isinstance(cfg.tts, TTSConfig)
-    assert cfg.tts.enabled is False
+    assert cfg.tts.enabled is True

voiceio/__init__.py

@@ -1 +1 @@
-__version__ = "0.3.1"
+__version__ = "0.3.2"

voiceio/config.py

@@ -105,7 +105,7 @@ class AutocorrectConfig:
 
 @dataclass
 class TTSConfig:
-    enabled: bool = False
+    enabled: bool = True
     engine: str = "auto"         # "auto" | "piper" | "espeak" | "edge-tts"
     hotkey: str = "ctrl+alt+s"   # "s" for speak
     voice: str = ""              # empty = engine default

voiceio/numbers.py

@@ -142,6 +142,7 @@ def convert_numbers(text: str, language: str = "en") -> str:
         # Collect consecutive number words
         if _is_number_word(low) and low != "a" and low != "and":
             num_words = []
+            last_category = None  # "ones", "tens", "scale"
             j = i
             while j < len(words):
                 w = words[j].lower().rstrip(".,;:?!")
@@ -153,6 +154,7 @@ def convert_numbers(text: str, language: str = "en") -> str:
                         # "a" at start: only if followed by scale word
                         if j + 1 < len(words) and words[j + 1].lower().rstrip(".,;:?!") in _SCALES:
                             num_words.append(w)
+                            last_category = "ones"
                             j += 1
                             continue
                         break
@@ -163,7 +165,14 @@ def convert_numbers(text: str, language: str = "en") -> str:
                             j += 1
                             continue
                         break
+                    # Two consecutive ones-words = separate numbers
+                    # e.g. "one two three" should NOT become 6
+                    # But "twenty three", "one hundred", "thirteen thousand" are valid
+                    cat = "scale" if w in _SCALES else ("tens" if w in _TENS else "ones")
+                    if cat == "ones" and last_category == "ones":
+                        break
                     num_words.append(w)
+                    last_category = cat
                     j += 1
                 else:
                     break

voiceio/tts/edge_engine.py

@@ -19,12 +19,25 @@ class EdgeEngine:
     def probe(self) -> ProbeResult:
         try:
             import edge_tts  # noqa: F401
-            return ProbeResult(ok=True)
         except ImportError:
             return ProbeResult(
                 ok=False, reason="edge-tts not installed",
                 fix_hint="pip install edge-tts",
             )
+        try:
+            import soundfile  # noqa: F401
+            return ProbeResult(ok=True)
+        except ImportError:
+            pass
+        try:
+            import pydub  # noqa: F401
+            return ProbeResult(ok=True)
+        except ImportError:
+            return ProbeResult(
+                ok=False,
+                reason="edge-tts needs soundfile or pydub to decode audio",
+                fix_hint="pip install soundfile",
+            )
 
     def synthesize(self, text: str, voice: str, speed: float) -> tuple[np.ndarray, int]:
         import asyncio

voiceio/tts/piper_engine.py

@@ -22,10 +22,11 @@ def __init__(self, model: str = ""):
     def probe(self) -> ProbeResult:
         try:
             import piper  # noqa: F401
+            from piper.download import ensure_voice_exists, get_voices  # noqa: F401
             return ProbeResult(ok=True)
-        except ImportError:
+        except ImportError as e:
             return ProbeResult(
-                ok=False, reason="piper-tts not installed",
+                ok=False, reason=f"piper-tts not fully installed: {e}",
                 fix_hint="pip install piper-tts",
             )