python-voiceio 1.15.1__tar.gz → 1.15.2__tar.gz

{python_voiceio-1.15.1/python_voiceio.egg-info → python_voiceio-1.15.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-voiceio
-Version: 1.15.1
+Version: 1.15.2
 Summary: Voice dictation for Linux. Speak → text, locally, instantly.
 Author: Hugo Montenegro
 License-Expression: MIT
@@ -146,6 +146,10 @@ Everything lives in `~/.config/voiceio/config.toml`; [config.example.toml](confi
 
 Open an issue before a large PR. [CONTRIBUTING.md](CONTRIBUTING.md) covers the architecture, conventions and the reasoning behind the model choice. Install reports and ideas are welcome on the [feedback board](https://voiceio.dev/#feedback).
 
+## How it's built
+
+I write voiceio with AI coding assistants, mostly Claude Code. I decide what goes in and I use it every day, which is how most of its bugs get found.
+
 ## Credits
 
 - Speech recognition: [faster-whisper](https://github.com/SYSTRAN/faster-whisper) and [CTranslate2](https://github.com/OpenNMT/CTranslate2), with OpenAI's [Whisper](https://github.com/openai/whisper) models; [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx) with NVIDIA's [Parakeet-TDT](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2). The in-browser demo runs [Moonshine](https://github.com/usefulsensors/moonshine) through [transformers.js](https://github.com/huggingface/transformers.js).

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/README.md

@@ -93,6 +93,10 @@ Everything lives in `~/.config/voiceio/config.toml`; [config.example.toml](confi
 
 Open an issue before a large PR. [CONTRIBUTING.md](CONTRIBUTING.md) covers the architecture, conventions and the reasoning behind the model choice. Install reports and ideas are welcome on the [feedback board](https://voiceio.dev/#feedback).
 
+## How it's built
+
+I write voiceio with AI coding assistants, mostly Claude Code. I decide what goes in and I use it every day, which is how most of its bugs get found.
+
 ## Credits
 
 - Speech recognition: [faster-whisper](https://github.com/SYSTRAN/faster-whisper) and [CTranslate2](https://github.com/OpenNMT/CTranslate2), with OpenAI's [Whisper](https://github.com/openai/whisper) models; [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx) with NVIDIA's [Parakeet-TDT](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2). The in-browser demo runs [Moonshine](https://github.com/usefulsensors/moonshine) through [transformers.js](https://github.com/huggingface/transformers.js).

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "python-voiceio"
-version = "1.15.1"
+version = "1.15.2"
 description = "Voice dictation for Linux. Speak → text, locally, instantly."
 readme = "README.md"
 license = "MIT"

{python_voiceio-1.15.1 → python_voiceio-1.15.2/python_voiceio.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-voiceio
-Version: 1.15.1
+Version: 1.15.2
 Summary: Voice dictation for Linux. Speak → text, locally, instantly.
 Author: Hugo Montenegro
 License-Expression: MIT
@@ -146,6 +146,10 @@ Everything lives in `~/.config/voiceio/config.toml`; [config.example.toml](confi
 
 Open an issue before a large PR. [CONTRIBUTING.md](CONTRIBUTING.md) covers the architecture, conventions and the reasoning behind the model choice. Install reports and ideas are welcome on the [feedback board](https://voiceio.dev/#feedback).
 
+## How it's built
+
+I write voiceio with AI coding assistants, mostly Claude Code. I decide what goes in and I use it every day, which is how most of its bugs get found.
+
 ## Credits
 
 - Speech recognition: [faster-whisper](https://github.com/SYSTRAN/faster-whisper) and [CTranslate2](https://github.com/OpenNMT/CTranslate2), with OpenAI's [Whisper](https://github.com/openai/whisper) models; [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx) with NVIDIA's [Parakeet-TDT](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2). The in-browser demo runs [Moonshine](https://github.com/usefulsensors/moonshine) through [transformers.js](https://github.com/huggingface/transformers.js).

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/tests/test_feedback_stream.py

@@ -1,11 +1,11 @@
 """The persistent cue stream: buffer size and cue timing."""
 from __future__ import annotations
 
-import pytest
+from types import SimpleNamespace
 
-pytest.importorskip("sounddevice")
+import pytest
 
-from voiceio import feedback  # noqa: E402
+from voiceio import feedback
 
 
 class FakeStream:
@@ -24,7 +24,7 @@ class FakeStream:
 @pytest.fixture
 def fake_stream(monkeypatch):
     FakeStream.opened = []
-    monkeypatch.setattr(feedback.sd, "OutputStream", FakeStream)
+    monkeypatch.setattr(feedback, "sd", SimpleNamespace(OutputStream=FakeStream))
     monkeypatch.setattr(feedback, "_stream", None)
     monkeypatch.setattr(feedback, "_sounds", {})
     return FakeStream

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/tests/test_idle_audio.py

@@ -3,17 +3,16 @@ from __future__ import annotations
 
 import threading
 import time
+from types import SimpleNamespace
 from unittest.mock import MagicMock, patch
 
 import numpy as np
 import pytest
 
-pytest.importorskip("sounddevice")
-
-from voiceio import feedback, idle_audio  # noqa: E402
-from voiceio.config import AudioConfig, Config  # noqa: E402
-from voiceio.idle_audio import IdleAudio  # noqa: E402
-from voiceio.recorder import AudioRecorder  # noqa: E402
+from voiceio import feedback, idle_audio
+from voiceio.config import AudioConfig, Config
+from voiceio.idle_audio import IdleAudio
+from voiceio.recorder import AudioRecorder
 
 
 class FakeInput:
@@ -50,8 +49,9 @@ class FakeOutput(FakeInput):
 @pytest.fixture
 def streams(monkeypatch):
     FakeInput.opened = FakeOutput.opened = 0
-    monkeypatch.setattr("voiceio.recorder.sd.InputStream", FakeInput)
-    monkeypatch.setattr(feedback.sd, "OutputStream", FakeOutput)
+    # Whole fakes: touching the real module would initialize PortAudio.
+    monkeypatch.setattr("voiceio.recorder.sd", SimpleNamespace(InputStream=FakeInput))
+    monkeypatch.setattr(feedback, "sd", SimpleNamespace(OutputStream=FakeOutput))
     monkeypatch.setattr(feedback, "_stream", None)
 
 

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/tests/test_postprocess.py

@@ -88,6 +88,20 @@ class TestStripDisfluencies:
         assert strip_disfluencies("uh-huh right") == "right"
         assert strip_disfluencies("uh huh yes") == "yes"
 
+    def test_mm_hmm_goes_whole(self):
+        # Parakeet ends about 1 clip in 4 with an invented "Mm-hmm."; cutting
+        # only its "hmm" typed a stray "Mm-." after real dictation.
+        assert strip_disfluencies("I sent it. Mm-hmm.") == "I sent it."
+        assert strip_disfluencies("Are you sure? Mm-hmm.") == "Are you sure?"
+        assert strip_disfluencies("Mm-hmm.") == ""
+        assert strip_disfluencies("Fine. Mm-hmm. So it works.") == "Fine. So it works."
+        assert strip_disfluencies("mmhmm that works") == "that works"
+
+    def test_fillers_inside_hyphenated_words_survive(self):
+        assert strip_disfluencies("a hmm-hmm sound") == "a hmm-hmm sound"
+        assert strip_disfluencies("she said uh-uh twice") == "she said uh-uh twice"
+        assert strip_disfluencies("the 5 mm drill") == "the 5 mm drill"
+
     def test_preserves_emphatic_short_repeat(self):
         # Short repeats are emphasis, not a re-decode artifact — keep them.
         assert strip_disfluencies("No. No.") == "No. No."

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/tests/test_robustness.py

@@ -936,3 +936,20 @@ class TestZombieStreamRecovery:
         # any real signal → not a zombie
         rec._ring.append(np.full(rec._ring._max, 0.1, dtype=np.float32))
         assert rec.is_zombie() is False
+
+
+def test_app_imports_when_portaudio_cannot_initialize(tmp_path):
+    # ubuntu-26.04 runners: libportaudio2 pulls libpulse0, and with no pulse
+    # server `import sounddevice` raises; importing voiceio.app must not.
+    import subprocess
+    import sys
+    (tmp_path / "sounddevice.py").write_text(
+        "class PortAudioError(Exception): pass\n"
+        "raise PortAudioError('PulseAudio_Initialize: Can\\'t connect to server')\n")
+    code = ("import voiceio.app, voiceio.recorder as r\n"
+            "try:\n    r.sd.InputStream\nexcept Exception as e:\n"
+            "    print(type(e).__name__)\n")
+    out = subprocess.run([sys.executable, "-c", code], capture_output=True, text=True,
+                         env={**__import__("os").environ, "PYTHONPATH": str(tmp_path)})
+    assert out.returncode == 0, out.stderr
+    assert out.stdout.strip() == "PortAudioError"

python_voiceio-1.15.2/voiceio/__init__.py

@@ -0,0 +1 @@
+__version__ = "1.15.2"

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/voiceio/feedback.py

@@ -20,11 +20,9 @@ from pathlib import Path
 
 import numpy as np
 
-try:
-    import sounddevice as sd
-except ImportError as _e:
-    from voiceio.platform import desktop_dep_hint
-    raise ImportError(desktop_dep_hint("sounddevice")) from _e
+from voiceio.platform import LazySounddevice
+
+sd = LazySounddevice()
 
 log = logging.getLogger(__name__)
 

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/voiceio/platform.py

@@ -15,6 +15,22 @@ log = logging.getLogger(__name__)
 DESKTOP_EXTRA_INSTALL = "pipx install --force 'python-voiceio[desktop]'"
 
 
+class LazySounddevice:
+    """`sounddevice`, imported on first use. Its import initializes PortAudio,
+    which raises when libpulse is installed but no sound server runs (CI,
+    headless boxes): imported eagerly, that broke `import voiceio.app` and
+    every command behind it."""
+
+    def __getattr__(self, name: str):
+        if name.startswith("_"):  # introspection (mock, inspect) must not import it
+            raise AttributeError(name)
+        try:
+            import sounddevice
+        except ImportError as e:
+            raise ImportError(desktop_dep_hint("sounddevice")) from e
+        return getattr(sounddevice, name)
+
+
 def desktop_dep_hint(package: str) -> str:
     """One-line message for a desktop dependency that failed to import."""
     return (

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/voiceio/postprocess.py

@@ -19,8 +19,9 @@ _NO_CASE_LANGUAGES = frozenset({"zh", "ja", "ko", "ar", "he", "th", "hi", "bn",
 #     own line doesn't swallow the paragraph/list break around it.
 # Word repetitions ("had had" is valid English) and filler "like"/"you know"
 # need judgment and are left to the LLM layer. Order matters: multi-token
-# "uh-huh" before "[Uu]h+" so it isn't clipped to a stray "-huh".
-_FILLER = r"\b(?:[Uu]h[-\s]?huh|[Mm]hm|[Uu]h+m*|[Uu]m+|[Hh]m{2,})\b"
+# "uh-huh" and "mm-hmm" (Parakeet invents "Mm-hmm." on a clip's quiet tail)
+# before the bare forms. Never a piece of a hyphenated word: that left "Mm-.".
+_FILLER = r"(?<!-)\b(?:[Uu]h[-\s]?huh|[Mm]m-?[Hh]m+|[Mm]hm|[Uu]h+m*|[Uu]m+|[Hh]m{2,})\b(?!-)"
 # A run of fillers and the comma/space around it; the lowercase letter after it, when
 # it opened a sentence, gets capitalized ("those. Um so I" -> "those. So I"),
 # or the sentence end before it would read as a false stop and become a comma.

{python_voiceio-1.15.1 → python_voiceio-1.15.2}/voiceio/recorder.py

@@ -8,19 +8,15 @@ from typing import TYPE_CHECKING, Callable
 
 import numpy as np
 
-try:
-    import sounddevice as sd
-except ImportError as _e:
-    from voiceio.platform import desktop_dep_hint
-    raise ImportError(desktop_dep_hint("sounddevice")) from _e
-
 from voiceio.audio_source import AudioBuffer
+from voiceio.platform import LazySounddevice
 
 if TYPE_CHECKING:
     from voiceio.config import AudioConfig
     from voiceio.vad import VadBackend
 
 log = logging.getLogger(__name__)
+sd = LazySounddevice()
 
 # Peak amplitude below this is treated as no signal at all (a muted mic emits
 # exact zeros; a working mic's "silence" still noises above ~1e-4). Matches the

python_voiceio-1.15.1/voiceio/__init__.py

@@ -1 +0,0 @@
-__version__ = "1.15.1"