Add dual recipe backends with OpenAI as the default.

Switch to Qwen2.5-3B-Instruct with CPU inference support.

Switch recipe generation to Qwen2.5-7B-Instruct.

Load gpt-oss-20b directly from the Hugging Face Hub.

Switch recipe generation to local gpt-oss via Transformers.

Add Epicure paper and model citations to the UI and README.

Switch recipe generation to OpenAI gpt-4o-mini.

Switch default recipe model to Arch-Router-1.5B.

Add HF Inference recipe generation and wire it to Generate.

Integrate Epicure Core nearest-neighbor shopping list suggestions.

initial commit