Die Trying · 06

What's next

  • Cut over to the new interface.
  • Journal drafts. Dated entries drafted from the encounters a character was in, one event per prompt, filtered on the server before the model sees anything.
  • Voice dictation. Short Nova-3 clips into the journal, with the campaign's keyterms.
  • Session summaries. The hard part is the transcript, not the summary. Four hours at the table is part game and part everything else, so a model first has to sort the play from the chatter. The recap should cover the dragon, not the pizza order.
  • Creative help for the DM. Ideas for NPCs, places and adventure prep, inside the notes the DM is already writing. The most optional thing in the app: off unless the DM turns it on, with a bring-your-own-key option.
  • Speaker separation. pyannote, already coded, in a self-hosted container that still has to be built.
  • AI Gateway. In front of every model call.
Diagram: the planned architecture. The SvelteKit Worker becomes the interface, calling the Python API Worker with a JWT. The API keeps its D1, R2 and Email Service bindings. AI Gateway is added in front of Workers AI (Nova-3, Llama 3.2 Vision, Qwen2.5-Coder). A self-hosted pyannote container, coded but not built, takes session audio and returns speaker turns matched against player voiceprints.
Where it goes next: the new interface in front, AI Gateway in front of the models, and a pyannote container for who said what.