README: update WebSocket path, config, model selection, prompts

OpenAI Realtime API: fix session/response params, remove temp docs

Post-merge: doc update, test coverage, security fixes

Remove X.AI, WebRTC, Gemini transcription; keep GA migration, gpt-realtime-1.5, IndexedDB replay, model selection

Merge dev into master: GA migration, gpt-realtime-1.5, IndexedDB replay, model selection, x.ai, WebRTC, Gemini (cleanup to follow)

realtime: remove type param from create_realtime_session (sessions API does not accept it)

chore: remove obsolete doc files

realtime: replace gpt-realtime with gpt-realtime-1.5 in model options

realtime: migrate from Beta to GA API

Feature/xai support (#15)

Add debug logging for provider/model selection and XAI client initialization

Fix XAI realtime client to disable audio output when modalities is text-only

Fix OpenAI client modalities parameter error and improve connection state management

Fix: Robust audio resource handling and static file serving

Fix: Properly cleanup audio resources on WebSocket close