Update model from decommissioned llama3-70b-8192 to llama-3.3-70b-versatile

AI now picks relevant stores per item instead of hardcoded Amazon/Etsy

Improve system prompt to request precise item names

Use Wikipedia API for images, remove image URLs from prompt

Add auto image preview for high-confidence results

Fix white margins in dark mode by setting html/body background

Center question mark in magnifying glass circle

Center question mark in magnifying glass circle

Update LogoMark SVG with C2PA metadata and embedded image

Switch to OpenAI SDK with IPAI_SK endpoint

Use @huggingface/inference library for proper API routing

Add health check endpoint, fetch timeouts, switch to zephyr-7b-beta

Fix HF API: use model in URL path, switch to Phi-3, improve error handling

Add custom DuckDuckGo web search tool

Switch from Anthropic to HF Inference API

Initial commit