# ─── Core ML / AI ───────────────────────────────────────────────────────────── torch>=2.0.0,<3.0.0 transformers>=4.40.0 faster-whisper>=1.0.0 # Whisper ASR via CTranslate2 - MIT silero-vad>=5.1 # Server-side VAD before ASR - MIT sentencepiece>=0.2.0 # Tokenizer dependency # ─── GGUF / LLM inference (CPU-native int4) ─────────────────────────────────── # Pre-built CPU wheel; no CUDA / Metal compilation needed llama-cpp-python>=0.2.90 # ─── Web / API ──────────────────────────────────────────────────────────────── fastapi>=0.100.0 uvicorn[standard]>=0.24.0 pydantic>=2.0.0 edge-tts>=7.2.3 # ─── Audio & Signal Processing ──────────────────────────────────────────────── soundfile>=0.12.1 numpy>=1.24.0 scipy>=1.11.0 # ─── Utilities ──────────────────────────────────────────────────────────────── tqdm>=4.66.0 huggingface-hub>=0.20.0