Inside the StoryGraph
A prose-first tool holds your plot in a paragraph and hopes the model remembers it. Storyum holds it as a graph — beats wired by typed connections, each carrying its own layer of who-knows-what. That's what makes a plot computable: change one node and the whole story re-derives, coherently. Here's how the graph is built.
Below is an example chapter as a graph — a short original mystery, built to show every part at once. The investigator works down the centre spine; the concealed truth runs on the left, ahead of him. Pan, zoom, and click any node to see what it is and what it reveals — to whom. (For famous works rebuilt this way, see the case studies.)
Nodes — the beats
Every node is a beat with a type that tells the engine what job it does. A reveal is not a complication; a decision is not a climax. Typed beats let the structure be read, checked and recomputed — not just written and hoped over.
Connections — the typed wires
Nodes are joined by connections, and every connection has a channel — a colour that says what it carries. One beat can feed the next along several channels at once: information down one wire, rising tension down another, a shifting relationship down a third. This is how a plot stays coherent on every axis, not just the sequence of events.
The knowledge layer
This is the part no prose-first tool has. Each node carries a small epistemic layer — a record of what it does to knowledge:
Because reveal records to whom, you can design the exact gap that makes a mystery work: in the graph above, the reader learns the witness is lying long before the detective does. You aren't hoping the model keeps the secret — you're engineering who holds it, and for how long.
Operators — shaping the signal
Sometimes a plain wire isn't enough — you don't just want a beat to carry tension forward, you want to hold it, invert it, or make it pay off later. Operators are transformers you drop onto a connection. They change what the wire carries downstream, which is how you compose the subtler moves — timing, contrast, setups that land scenes later. Here is the full set, by family.
Why a graph, not a document
Because the plot is a structure, not a paragraph, the engine can compute the book to it — and recompute when you change one node. Move a reveal earlier, add a complication, flip who knows what, and everything downstream re-derives to stay coherent. A document you'd have to rewrite by hand; a graph re-solves itself. That's the moat: structure reaches the text.