StoryGraph

In production

A screenplay, turned into the graph of its own story.

Structural dramaturgy analysis: a script becomes acts, scenes, characters and storylines, and a diagnostic engine says which scenes are not carrying their weight.

Input
Screenplay text · Fountain export
Model
Acts · scenes · characters · storylines · want/need
Engine
Structural Diagnostic Engine
Processing
Asynchronous — Celery task queue

Overview

Script development runs on notes — expensive, slow, and shaped by whoever gave them. StoryGraph makes the structure explicit first: it parses a screenplay into a story graph of acts, scenes, characters and narrative lines, including what each character wants and what they need.

On top of that graph runs a Diagnostic Engine that produces specific, locatable findings rather than impressions — the kind that reads "scenes 3–5 do not advance any active storyline." The graph is the argument; the diagnosis points at it.

The loop is verified end-to-end against real models: create a project, upload a script, parse asynchronously, inspect the structure, run diagnostics, export to Fountain. Transient model failures retry with backoff and surface in the interface if they exhaust rather than failing silently.

Capabilities

Structural parsing

A screenplay decomposed into an addressable graph instead of a wall of text.

Diagnostic findings

Concrete cards pointing at concrete scenes — dead stretches, abandoned lines, unmotivated turns.

Character arcs

Want versus need tracked per character across the graph, so a flat arc becomes visible rather than felt.

Provider-agnostic

Runs against multiple LLM providers behind one interface, with retry and backoff on transient failures.