debug: add internet diagnostics to test httpbin

feat: add support for custom OpenAI-compatible endpoints in config UI

debug: add env diagnostics to print all env keys

fix: pass user overrides token and model to inference_generate

debug: add startup diagnostics to test various models and endpoints

fix: handle None genre from Gradio API input

feat: 3-call LLM generation per turn with funny health commentary

feat: LLM-driven infinite adventure loop