feat: sovereign voice loop — timmy voice command

Adds fully local listen-think-speak voice interface. STT: Whisper, LLM: Ollama, TTS: Piper. No cloud, no network. - src/timmy/voice_loop.py: VoiceLoop with VAD, Whisper, Piper - src/timmy/cli.py: new voice command - pyproject.toml: voice extras updated - 20 new tests
2026-03-14 13:58:56 -04:00
parent d770d66150
commit dbadfc425d
4 changed files with 696 additions and 1 deletions
--- a/tests/timmy/test_voice_loop.py
+++ b/tests/timmy/test_voice_loop.py
@@ -0,0 +1,273 @@
+"""Tests for the sovereign voice loop.
+
+These tests verify the VoiceLoop components without requiring a microphone,
+Whisper model, or Piper installation — all I/O is mocked.
+"""
+
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+
+from timmy.voice_loop import VoiceConfig, VoiceLoop
+
+# ── VoiceConfig tests ──────────────────────────────────────────────────────
+
+
+class TestVoiceConfig:
+    def test_defaults(self):
+        cfg = VoiceConfig()
+        assert cfg.whisper_model == "base.en"
+        assert cfg.sample_rate == 16000
+        assert cfg.silence_threshold == 0.015
+        assert cfg.silence_duration == 1.5
+        assert cfg.min_utterance == 0.5
+        assert cfg.max_utterance == 30.0
+        assert cfg.session_id == "voice"
+        assert cfg.use_say_fallback is False
+
+    def test_custom_values(self):
+        cfg = VoiceConfig(
+            whisper_model="tiny.en",
+            silence_threshold=0.02,
+            session_id="custom",
+            use_say_fallback=True,
+        )
+        assert cfg.whisper_model == "tiny.en"
+        assert cfg.silence_threshold == 0.02
+        assert cfg.session_id == "custom"
+        assert cfg.use_say_fallback is True
+
+
+# ── VoiceLoop unit tests ──────────────────────────────────────────────────
+
+
+class TestVoiceLoopInit:
+    def test_default_config(self):
+        loop = VoiceLoop()
+        assert loop.config.whisper_model == "base.en"
+        assert loop._running is False
+        assert loop._speaking is False
+
+    def test_custom_config(self):
+        cfg = VoiceConfig(whisper_model="tiny.en")
+        loop = VoiceLoop(config=cfg)
+        assert loop.config.whisper_model == "tiny.en"
+
+
+class TestPiperFallback:
+    def test_falls_back_to_say_when_no_voice_file(self):
+        cfg = VoiceConfig(piper_voice=Path("/nonexistent/voice.onnx"))
+        loop = VoiceLoop(config=cfg)
+        loop._ensure_piper()
+        assert loop.config.use_say_fallback is True
+
+    def test_keeps_piper_when_voice_exists(self, tmp_path):
+        voice_file = tmp_path / "test.onnx"
+        voice_file.write_bytes(b"fake model")
+        cfg = VoiceConfig(piper_voice=voice_file)
+        loop = VoiceLoop(config=cfg)
+        loop._ensure_piper()
+        assert loop.config.use_say_fallback is False
+
+
+class TestTranscribe:
+    def test_transcribes_audio(self):
+        """Whisper transcription returns cleaned text."""
+        loop = VoiceLoop()
+
+        mock_model = MagicMock()
+        mock_model.transcribe.return_value = {"text": "  Hello Timmy  "}
+        loop._whisper_model = mock_model
+
+        audio = np.random.randn(16000).astype(np.float32)
+        result = loop._transcribe(audio)
+
+        assert result == "Hello Timmy"
+        mock_model.transcribe.assert_called_once()
+
+    def test_transcribes_empty_returns_empty(self):
+        loop = VoiceLoop()
+        mock_model = MagicMock()
+        mock_model.transcribe.return_value = {"text": "   "}
+        loop._whisper_model = mock_model
+
+        audio = np.random.randn(16000).astype(np.float32)
+        result = loop._transcribe(audio)
+        assert result == ""
+
+
+class TestThink:
+    @patch("timmy.voice_loop.asyncio")
+    def test_think_returns_response(self, mock_asyncio):
+        mock_asyncio.run.return_value = "I am Timmy."
+        loop = VoiceLoop()
+        result = loop._think("Who are you?")
+        assert result == "I am Timmy."
+
+    @patch("timmy.voice_loop.asyncio")
+    def test_think_handles_error(self, mock_asyncio):
+        mock_asyncio.run.side_effect = RuntimeError("Ollama down")
+        loop = VoiceLoop()
+        result = loop._think("test")
+        assert "trouble" in result.lower()
+
+
+class TestSpeakSay:
+    @patch("subprocess.Popen")
+    def test_speak_say_calls_subprocess(self, mock_popen):
+        mock_proc = MagicMock()
+        mock_proc.wait.return_value = 0
+        mock_popen.return_value = mock_proc
+
+        cfg = VoiceConfig(use_say_fallback=True)
+        loop = VoiceLoop(config=cfg)
+        loop._speak_say("Hello")
+
+        mock_popen.assert_called_once()
+        args = mock_popen.call_args[0][0]
+        assert args[0] == "say"
+        assert "Hello" in args
+
+    @patch("subprocess.Popen", side_effect=FileNotFoundError)
+    def test_speak_say_handles_missing(self, mock_popen):
+        cfg = VoiceConfig(use_say_fallback=True)
+        loop = VoiceLoop(config=cfg)
+        # Should not raise
+        loop._speak_say("Hello")
+
+
+class TestSpeakPiper:
+    @patch("timmy.voice_loop.VoiceLoop._play_audio")
+    @patch("subprocess.run")
+    def test_speak_piper_generates_and_plays(self, mock_run, mock_play):
+        mock_run.return_value = MagicMock(returncode=0, stderr="")
+
+        voice_path = Path("/tmp/test_voice.onnx")
+        cfg = VoiceConfig(piper_voice=voice_path)
+        loop = VoiceLoop(config=cfg)
+        loop._speak_piper("Hello from Piper")
+
+        # Piper was called
+        mock_run.assert_called_once()
+        cmd = mock_run.call_args[0][0]
+        assert cmd[0] == "piper"
+        assert "--model" in cmd
+
+        # Audio was played
+        mock_play.assert_called_once()
+
+    @patch("timmy.voice_loop.VoiceLoop._speak_say")
+    @patch("subprocess.run")
+    def test_speak_piper_falls_back_on_error(self, mock_run, mock_say):
+        mock_run.return_value = MagicMock(returncode=1, stderr="model error")
+
+        cfg = VoiceConfig(piper_voice=Path("/tmp/test.onnx"))
+        loop = VoiceLoop(config=cfg)
+        loop._speak_piper("test")
+
+        # Should fall back to say
+        mock_say.assert_called_once_with("test")
+
+
+class TestHallucinationFilter:
+    """Whisper tends to hallucinate on silence/noise. The loop should filter these."""
+
+    def test_known_hallucinations_filtered(self):
+        hallucinations = [
+            "you",
+            "thanks.",
+            "Thank you.",
+            "Bye.",
+            "Thanks for watching!",
+            "Thank you for watching!",
+        ]
+        for text in hallucinations:
+            assert text.lower() in (
+                "you",
+                "thanks.",
+                "thank you.",
+                "bye.",
+                "",
+                "thanks for watching!",
+                "thank you for watching!",
+            ), f"'{text}' should be filtered"
+
+
+class TestExitCommands:
+    """Voice loop should recognize exit commands."""
+
+    def test_exit_commands(self):
+        exits = ["goodbye", "exit", "quit", "stop", "goodbye timmy", "stop listening"]
+        for cmd in exits:
+            assert cmd.lower().strip().rstrip(".!") in (
+                "goodbye",
+                "exit",
+                "quit",
+                "stop",
+                "goodbye timmy",
+                "stop listening",
+            ), f"'{cmd}' should be an exit command"
+
+
+class TestPlayAudio:
+    @patch("subprocess.Popen")
+    def test_play_audio_calls_afplay(self, mock_popen):
+        mock_proc = MagicMock()
+        mock_proc.poll.side_effect = [None, 0]  # Running, then done
+        mock_popen.return_value = mock_proc
+
+        loop = VoiceLoop()
+        loop._play_audio("/tmp/test.wav")
+
+        mock_popen.assert_called_once()
+        args = mock_popen.call_args[0][0]
+        assert args[0] == "afplay"
+
+    @patch("subprocess.Popen")
+    def test_play_audio_interruptible(self, mock_popen):
+        mock_proc = MagicMock()
+        # Simulate running, then we interrupt
+        call_count = 0
+
+        def poll_side_effect():
+            nonlocal call_count
+            call_count += 1
+            return None  # Always running
+
+        mock_proc.poll.side_effect = poll_side_effect
+        mock_popen.return_value = mock_proc
+
+        loop = VoiceLoop()
+        loop._interrupted = True  # Pre-set interrupt
+        loop._play_audio("/tmp/test.wav")
+
+        mock_proc.terminate.assert_called_once()
+
+
+class TestStopMethod:
+    def test_stop_sets_running_false(self):
+        loop = VoiceLoop()
+        loop._running = True
+        loop.stop()
+        assert loop._running is False
+
+
+class TestSpeakSetsFlag:
+    @patch("timmy.voice_loop.VoiceLoop._speak_say")
+    def test_speaking_flag_set_during_speech(self, mock_say):
+        cfg = VoiceConfig(use_say_fallback=True)
+        loop = VoiceLoop(config=cfg)
+
+        # Before speak
+        assert loop._speaking is False
+
+        # Mock say to check flag during execution
+        def check_flag(text):
+            assert loop._speaking is True
+
+        mock_say.side_effect = check_flag
+        loop._speak("Hello")
+
+        # After speak
+        assert loop._speaking is False