Voice in, text out. Users speak to an agent which helps them draft a document.
Based on the OpenAI Realtime agents demo
Possible future directions
- Diagram building - e.g. using Mermaid diagrams
- (Gratitude) journalling - easy way to add entries
- Forecasting - help users build 'guesstimate' style forecasts by mapping out different option spaces and assigning probabilities
Tech improvements
- Persist conversations and documents
- 1 - just with cookies
- 2 - login w supabase
- Instead of editing full document, provide a (git style?) diff, or find and replace with sed
- Try out Gemini offerings - 2.5 voice is pretty good
- Make it responsive - work on mobile