Fetching metadata from the HF Docker repository...
-
Switch to Google Vision only - remove CLIP/transformers/torch
-
Improve CLIP accuracy: photo prompts, expanded categories, confidence threshold
-
Add Product Vision AI tab with Google Vision + CLIP fallback
-
Add PDF support with per-page OCR extraction
-
Fix: lazy-load OCR model and smart URL auto-detection
-
Add Dockerfile and secure API Key for HuggingFace Spaces
-
Fix Render port and python-bidi install error
-
first commit