use gpt-4.1-nano for response validation instead of Gemini

replace sentence-transformers with OpenAI embeddings — fixes slow HF build

add HF Space config frontmatter to README

add HF Space config frontmatter to README

initial deployment