- fix for stt-server (got broken by webservers dependency upgrade because of an api change)
- added initial_prompt_realtime to AudioToTextRecorder to be able to give different prompts to final and realtime model
- added new parameters to client/server (download root, batch sizes)