English · 简体中文
Show Me The Story
User guide · Download releases
A local application for writing long fiction with AI. One executable and a browser interface connect to an OpenAI-compatible model service to manage settings, plan batches of chapters, draft and revise prose, and export your work.
The complete application ships as a single executable under 5 MB—small enough to download, move, and keep anywhere.
You decide the story’s rules and direction. Those rules may be realistic or entirely fictional.
First run
- Download a release for your system, extract it into a directory you intend to keep, and launch it.
- Open
http://localhost:48090 and create a Chinese or English project.
- In Configuration, enter the API address, model identifier, and API key. Test the connection and save. Use the exact model identifier supplied by your provider.
- Save the genre, target chapter length, style, and point of view. Add essential characters and worldview entries.
- In Outline, enter a synopsis and chapter count (1–36) for the first batch. Generate and review it. If the page offers Confirm outline, confirm before writing.
- In Writing, generate one chapter, read and revise it, then accept it before continuing.
Start with a small batch and auto-confirm disabled to check the voice, characters, and model behavior.
The first-story walkthrough explains the input, expected result, and next step for each stage.
Find the right workflow
Features
- Multiple projects; Chinese and English writing with independently switchable UI language.
- Batch outlines, long-term direction, ending intent, and planning reviews.
- Characters, worldview, organizations, relationship graphs, and story knowledge.
- Chapter drafting, review, paragraph editing, targeted revision, and optional auto-confirm.
- Extracted facts with source references, setting-change proposals, and foreshadow tracking.
- Save knowledge and revise a chapter: selected entries enter that request in full; later writing retrieves relevant entries.
- Existing-text import, optional writing/polishing skills, and a conversational assistant.
- Separate final proofreading, anchored reports, per-chapter proofreading undo, and continuation projects.
- Streaming output, logs, cancellation, local persistence, and text/outline/report exports.
Consistency checks primarily use your story context; they are not online research. Authors still need to review model output.
Running and storing data
The default data directory is the working directory at launch. You can pass an existing directory:
./show-me-the-story
./show-me-the-story /path/to/existing/novels
On Windows, double-click the executable or run:
.\show-me-the-story.exe "D:\Novels"
Check the project directory printed at startup. An invalid directory argument falls back to the working directory. The default port is 48090; override it with the PORT environment variable.
Projects live in storys/<project>/. Stop tasks and close the program before copying the entire data directory for backup or migration. This release only opens v4 projects. Use the version recommended by the project list for older formats; do not edit the format number manually.
After tasks finish, use Back up ZIP in the project list to download a complete project. Restore a project backup restores it as a new project without overwriting the original. Select the restored project before asking the assistant about its entries; search_project searches stored characters, worldview, organisations, foreshadows, and outlines by keyword. Global API configuration and user Skills are excluded; preserve them separately when migrating. See backups and restoration.
After an interrupted prose/progress save, reopening the project attempts to restore the last complete save. Damaged chapter files or failed recovery prevent the project from opening and display diagnostics. Preserve the entire project directory, including progress.json.rollback if present, then resolve file access problems or restore a backup. The rollback journal is not version history or a replacement for regular backups; do not run multiple writers against the same project.
Files are stored locally, but AI operations send relevant prose, settings, and prompts to your configured model service, which may charge for usage. API keys are stored in api.json; inspect shared files and logs for sensitive information. Use the application in a trusted local environment, not directly exposed to the public internet.
If Windows reports an unknown publisher, verify the download source. After deciding to trust the executable, use More info → Run anyway.
Development
Go 1.25.1 with the standard library; Vite 5, Svelte 4, Tailwind CSS 4, DaisyUI 5, and @xianii/design-system. Frontend assets and built-in skills are embedded in the executable.
Install Go and Node.js, optionally Task:
task build
Or build manually:
cd frontend
npm install
npm run build
cd ..
go build -o show-me-the-story .
Use task dev for the backend and task dev:frontend for the frontend development server (5173, proxying API requests to 48090). Run task screenshots with Google Chrome installed to rebuild the README images from fixed offline sample data.
Checks:
go build ./...
go test ./...
go vet ./...
node frontend/src/lib/projectRestore.check.js
node frontend/src/lib/forceGraphLayout.check.js
See AGENTS.md for architecture and contribution constraints, and the user guide for workflows.
License
MIT
评论