perf: reduce per-query latency — shared model, single encode, float32 LLM

fix: add api_name=False to all event handlers to bypass Gradio schema crash

fix: remove monkeypatch to restore standard HF auth handshake

fix: remove unsupported launch arguments to fix TypeError

fix: disable share and ssr to resolve Hugging Face internal routing conflict

fix: explicit chatbot type='messages' for Gradio 5 compatibility

fix: move css and theme to gr.Blocks constructor for Gradio 5 compatibility

fix: nuclear monkeypatch for Gradio HfFolder ImportError and bump sdk to 5.13.0

fix: Map missing Unsplash images and replace 404s

feat: Add reset button and optimize LLM inference speed/accuracy

feat: Add health/beauty products and refine retriever logic

feat: Complete RetailMind overhaul — hybrid retrieval, EWMA drift detection, self-healing adapter, premium UI, tests & CI