api/requirements.txt

# Voxtral local inference + FER (facial emotion recognition)
fastapi>=0.115.0
uvicorn[standard]>=0.32.0
python-multipart>=0.0.9
librosa>=0.10.0
soundfile>=0.12.0
numpy>=1.24.0
# Voxtral model
torch>=2.4.0
transformers==4.56.0
peft>=0.13.0
accelerate>=1.0.0
mistral-common
safetensors
sentencepiece
# FER inference — model uses ONNX IR v10, requires ort>=1.19.0
onnxruntime==1.24.2
opencv-python-headless>=4.8.0