How it works
To SillyTavern, SmutWriter Model looks like any Custom (OpenAI-compatible) endpoint. The difference happens on our side, between your message and the reply.
1. We recognise the story
Each story is matched by your account and its character card. Start a new chat with the same card and its memory starts over. Want two runs of one card at once? Add a story_id (see Docs).
2. We remember in the background
After each reply, older messages are folded into a short story so far and a ledger of lasting facts: names, looks, relationships, places, events, and any instruction you gave about how to write. Swiped-away replies are never remembered, only the one you keep.
3. We rebuild the prompt
Each request is assembled fresh, up to 32K tokens: your card and World Info, then the story so far, then the facts that matter for this scene, then as many recent messages as fit. Messages above SillyTavern's cutoff line are covered by the memory, not lost.
4. We nudge the writing
A short set of writing notes rides along: concrete prose, continuity, staying in character, and never writing your character's lines for you. You are never billed for them.
5. Hard limits
Fiction between adults only. Anything sexual involving a person under 18 is blocked on the way in and on the way out, however it's framed.