# Core fastapi>=0.115.0 uvicorn[standard]>=0.32.0 pydantic>=2.9.0 pydantic-settings>=2.6.0 python-multipart>=0.0.12 python-dotenv>=1.0.0 # Image / audio Pillow>=10.4.0 numpy>=1.26.0 # Audio (Whisper) faster-whisper>=1.0.3 # Note: faster-whisper needs ctranslate2 which has prebuilt wheels for win/amd64 # Vision / text models (HF transformers) # Pin transformers <4.49 to avoid Florence-2/Moondream2 breaking changes torch>=2.4.0 transformers>=4.45.0,<4.49.0 accelerate>=1.0.0 sentencepiece>=0.2.0 protobuf>=5.28.0 einops>=0.8.0 timm>=1.0.0 # Optional: Ollama HTTP client (already in stdlib via httpx) httpx>=0.27.0 # Retry / robustness tenacity>=9.0.0 # Testing pytest>=8.3.0 pytest-asyncio>=0.24.0 httpx>=0.27.0 # Optional: Tesseract OCR (no HF download needed) # pip install pytesseract # Windows: also install tesseract binary from https://github.com/UB-Mannheim/tesseract/wiki pytesseract>=0.3.13