Draft your prologue setting, formulate an initial command below, or initialize the journey with a custom Story Seed.
This is the compressed continuity state for older story history after compaction. Edit only when compaction missed important continuity, preserved wrong facts, or became too verbose. Over-editing can reduce story quality and texture.
Use the guided forms for normal edits. The raw editor is kept only for repairing model output, import cleanup, or emergency recovery.
You don't always need to type a prompt. Scribe & Saga supports dynamic auto-advancements:
"start the story naturally"
message to establish your story in motion.
"(continue)"
command, prompting the AI storyteller to move the living world
forward naturally.
The engine parses LLM outputs in real-time, giving standard manuscript rules beautiful typographic distinction:
<think> blocks
are structured into reasoning modules (great for DeepSeek R1
models).
Hovering or tapping a manuscript block reveals modification tools. Long generated passages show controls near both the top and bottom so you do not need to scroll back up after reading.
The Memory Ledger stores durable continuity anchors that should survive long stories and automatic compaction. It is separate from the visible manuscript and is sent to the model every turn.
When older story history grows too large, Scribe & Saga compacts earlier raw messages into a Chronicle Summary while keeping recent messages raw for scene flow.
Web browsers isolate script requests from online sandboxes to local hostnames. To authorize the engine to speak with your local Ollama port, launch Ollama with a cross-origin directive:
set OLLAMA_ORIGINS="*" ollama serve
OLLAMA_ORIGINS="*" open -a Ollama(Quit the running menu-bar app first before running this terminal launcher!)
sudo systemctl edit ollama.serviceAdd these rows under the Service section, save, and restart:
[Service] Environment="OLLAMA_ORIGINS=*"