[build-system]
requires = ["setuptools>=69", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "moda-fashionsiglip-multiview"
version = "0.1.0"
description = "Zero-additional-parameter multi-view text-to-image retrieval with FashionSigLIP"
readme = "README.md"
requires-python = ">=3.10"
license = { text = "MIT" }
authors = [
  { name = "MODA contributors" },
]
dependencies = [
  "huggingface_hub>=0.24",
  "open_clip_torch>=2.24",
  "Pillow>=10",
  "safetensors>=0.4",
  "torch>=2.2",
  # The FashionSigLIP checkpoint declares hf_tokenizer_name
  # "timm/ViT-B-16-SigLIP", so open_clip builds an HFTokenizer, which needs
  # transformers and the SigLIP sentencepiece vocabulary at load time.
  "transformers>=4.40",
  "sentencepiece>=0.1.99",
]
keywords = [
  "fashion",
  "FashionSigLIP",
  "multimodal retrieval",
  "text-to-image retrieval",
]

[project.urls]
Homepage = "https://huggingface.co/HopitAI/moda-fashionsiglip-multiview-203m"
Demo = "https://huggingface.co/spaces/HopitAI/moda-fashionsiglip-multiview-demo"
Base-Model = "https://huggingface.co/Marqo/marqo-fashionSigLIP"

[project.scripts]
moda-fashion-search = "moda_fashionsiglip_multiview.cli:main"

[tool.setuptools]
package-dir = { "" = "src" }

[tool.setuptools.packages.find]
where = ["src"]

[tool.setuptools.package-data]
moda_fashionsiglip_multiview = ["recipe.json"]

[tool.setuptools.exclude-package-data]
"*" = ["__pycache__/*", "*.pyc"]