All Tools
U
RAGFreeOpen Source
UFORM
Tiny multimodal embeddings for text and images
Apache-2.0
ABOUT
CLIP-class models are too large for many CPU and edge retrieval setups. UForm provides compact multimodal encoders for text and images so you can build multilingual visual search and RAG over screenshots, product photos, and captions without a GPU.
INSTALL
pip install uformINTEGRATION GUIDE
1. Embed product images and captions into one space for visual search
2. Run multilingual image-text retrieval on CPU or edge hardware
3. Index screenshots and UI images for multimodal RAG
4. Replace a heavy CLIP checkpoint with a smaller UForm encoder
TAGS
pythonembeddingsmultimodalclipragopen-source