| pip>=23.0 | |
| setuptools>=68.0 | |
| wheel | |
| fastapi>=0.100 | |
| uvicorn>=0.34.0 | |
| # gradio==5.29.0 | |
| Pillow>=11.0.0 | |
| numpy==2.2.5 | |
| openai-whisper==20240927 | |
| ultralytics>=8.3.40 | |
| torch>=2.6.0 | |
| pydantic>=2.11.3 | |
| python-multipart>=0.0.20 | |
| opencv-python>=4.10.0.84 | |
| transformers>=4.51.3 | |
| sentencepiece>=0.2.0 | |
| scikit-image>=0.25.2 | |
| scipy>=1.15.2 | |
| googletrans==4.0.2 | |
| httpcore==1.0.9 | |
| roboflow==1.1.63 | |
| inference-gpu[yolo-world]==0.48.1 # Commented out due to numpy version conflicts | |
| git+https://github.com/ultralytics/CLIP.git | |
| faster-whisper>=1.0.3 | |