Mac M1 optimizations, fix train pipeline, add Hey Cosmo wake word model

- Fix install_mac.sh: use venv + Python 3.12 (3.14 incompatible with ML libs)
- Fix run_mac.sh: activate venv, add CPU thread optimization env vars
- Fix agent.py: remove f-string from SYSTEM_PROMPT template (NameError on import)
- Add missing deps: sounddevice, pydub, imageio-ffmpeg, omegaconf
- Optimize for M1: torch.inference_mode, set_num_threads, OMP/MKL tuning
- Switch to qwen2.5:3b for faster LLM responses on Mac
- Switch Whisper to medium model with auto compute (small+int8 had poor Russian)
- Add initial_prompt for better Russian transcription
- Add open_app tool for native macOS app launching
- Fix TTS: sanitize Latin text to Cyrillic for Silero compatibility
- Fix wake word echo: add cooldown after TTS, reset model state, raise threshold
- Make "Слушаю" TTS synchronous to avoid mic interference
- Fix train Dockerfile: remove tensorflow/onnx2tf (only ONNX needed), fix deps
- Fix train.sh: use wget for dataset download, add --shm-size=2g
- Add trained hey_cosmo.onnx wake word model
- Add TODO section to CLAUDE.md (ChatterBox TTS, Ollama Modelfile ideas)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

train_wakeword/train.sh

@@ -51,24 +51,17 @@ NEGATIVE_FEATURES="$DATA_DIR/openwakeword_features_ACAV100M_2000_hrs_16bit.npy"
 VALIDATION_FEATURES="$DATA_DIR/validation_set_features.npy"
 
 if [ ! -f "$NEGATIVE_FEATURES" ]; then
-    echo "[2/4] Скачиваю негативный датасет (~20 GB, один раз)..."
+    echo "[2/4] Скачиваю негативный датасет (~17 GB + ~500 MB, один раз)..."
     echo "      Это займёт время в зависимости от скорости интернета."
-    docker run --rm \
-        -v "$DATA_DIR:/data" \
-        cosmo-wakeword-trainer \
-        python -c "
-from datasets import load_dataset
-import numpy as np, os
-print('Скачиваю ACAV100M features...')
-ds = load_dataset('davidscripka/openwakeword_features', 'ACAV100M_2000_hrs_16bit', split='train')
-arr = np.array(ds['features'])
-np.save('/data/openwakeword_features_ACAV100M_2000_hrs_16bit.npy', arr)
-print('Скачиваю validation features...')
-ds_val = load_dataset('davidscripka/openwakeword_features', 'validation_set', split='train')
-arr_val = np.array(ds_val['features'])
-np.save('/data/validation_set_features.npy', arr_val)
-print('Датасет скачан.')
-"
+    echo ""
+    echo "      Скачиваю ACAV100M features (~17 GB)..."
+    wget -q --show-progress \
+        -O "$NEGATIVE_FEATURES" \
+        "https://huggingface.co/datasets/davidscripka/openwakeword_features/resolve/main/openwakeword_features_ACAV100M_2000_hrs_16bit.npy"
+    echo "      Скачиваю validation features (~500 MB)..."
+    wget -q --show-progress \
+        -O "$VALIDATION_FEATURES" \
+        "https://huggingface.co/datasets/davidscripka/openwakeword_features/resolve/main/validation_set_features.npy"
     echo "      Датасет готов."
 else
     echo "[2/4] Негативный датасет уже скачан. Пропускаю."
@@ -86,6 +79,7 @@ if [ -d "$POSITIVE_DIR" ] && [ -n "$(ls "$POSITIVE_DIR"/*.wav 2>/dev/null)" ]; t
 fi
 
 docker run --rm \
+    --shm-size=2g \
     -v "$SCRIPT_DIR/cosmo_config.yaml:/app/cosmo_config.yaml" \
     -v "$DATA_DIR:/data" \
     -v "$MODELS_DIR:/output" \