refactor: remove all local STT/TTS dependencies and code

- Remove openai-whisper and piper-tts from dependencies (STT/TTS handled by Home Assistant)
- Remove gradio from dependencies (not needed for Reachy Mini app)
- Remove opencv-python and mediapipe from core dependencies (moved to optional vision)
- Remove stt_engine, stt_model, tts_engine, tts_model from default config
- Remove gradio config section
- Keep only essential dependencies for core functionality
- This fixes the protobuf version conflict between aioesphomeapi and mediapipe
- Users can install vision features with: pip install -e .[vision]

pyproject.toml

@@ -72,6 +72,8 @@ vision = [
     "pollen-vision",
     "torch>=2.0.0",
     "transformers>=4.30.0",
+    "opencv-python>=4.8.0",
+    "mediapipe>=0.10.0",
 ]
 
 dev = [

src/reachy_mini_ha_voice/config/manager.py

@@ -56,11 +56,7 @@ class ConfigManager:
             },
             "voice": {
                 "wake_word": "okay_nabu",
-                "wake_word_dirs": ["wakewords"],
-                "stt_engine": "whisper",
-                "stt_model": "base",
-                "tts_engine": "piper",
-                "tts_model": "en_US-lessac-medium"
+                "wake_word_dirs": ["wakewords"]
             },
             "motion": {
                 "enabled": True,
@@ -76,11 +72,6 @@ class ConfigManager:
                 "host": "localhost",
                 "wireless": False
             },
-            "gradio": {
-                "enabled": False,
-                "host": "127.0.0.1",
-                "port": 7860
-            },
             "logging": {
                 "level": "INFO",
                 "file": "reachy_mini_ha_voice.log"