Add support to HuggingFace to provide Vision and Embedding models

v0.6.2

Version 0.6.2: SplitterMR now supports HuggingFace as a backend for both embedding and vision models:

• HuggingFaceEmbedding: Use any Sentence Transformers model (local or from Hugging Face Hub) for fast, local, or cloud embeddings.
• HuggingFaceVisionModel: Leverage Hugging Face’s vision-language models for image-to-text and image captioning.

To use HuggingFace, Gemini, Claude or Grok models, you must install SplitterMR with the multimodal extra:

pip install "splitter-mr[multimodal]"

Add HuggingFace Model and Embedding support.

Features

• Add HuggingFaceVisionModel class. Note that the support is limited until now.
• Add HuggingFaceEmbedding class.

Documentation

• Add HuggingFaceVisionModel to documentation.
• Update architecture diagram.
• Update README.md.

Developer features

• Add new dependencies to multimodal group.