MVP add voice viz add shadcn based ui with easy brand overrides add varying greeting texts add saving of transcriptions and recordings of conversations
add some screener? or keep it panel side?
add some sort of sampling capture (ie from Cint panels) and redirect back to panel
at end of interview ask some contact details?
add prompt generator to design proper prompts (based on cli for devs?)
input from product interview sources
use openai reference prompt
Some max session length limiter to avoid excess costs
Add some noise reduction worker & plugin (ie krisp or something open source)
Add some timeout if there too much silence from client side, disconnect and / or pause the session server side to not use openai tokens
On mobile, simplify the mic & speaker selection somehow?
Later
fix the bug with clientside tool use and server not allowing sessionupdates
- Add authentication to the relay endpoint
- Add rate limiting
- Add error handling and retries for WebSocket connections
- Add monitoring and logging
- Add tests
- Add documentation for deployment and configuration
- Add optional Sentry observability
- Add optional PostHog analytics