Writingway 2

AI-assisted creative writing, scene planning, and worldbuilding in a local-first app.

Writingway 2 is a browser-based writing tool built for drafting fiction, organizing scenes, keeping worldbuilding notes close at hand, and working with either cloud AI providers or a local GGUF model.

Discord: https://discord.gg/HyRmNKe5QA

What Writingway does

Writingway is organized around projects, chapters, scenes, and compendium entries. It gives you:

• A scene-first editor for drafting and revising
• Chapter and scene organization with reordering
• A compendium for characters, locations, lore, items, and other story notes
• AI-assisted drafting, rewriting, brainstorming, and workshop chat
• Writingway 1 project import
• Local project save/export tools
• Optional backup flows
• Optional local GGUF inference through llama.cpp

Highlights

• Local-first writing workflow Your projects live in IndexedDB while you work, and can also be saved to disk as project files.

• Flexible AI setup Use OpenRouter, Anthropic, OpenAI, Google, NanoGPT, LM Studio, custom OpenAI-compatible endpoints, or a local GGUF model via llama.cpp.

• Built-in local GGUF setup flow If Writingway detects a .gguf file in models/ but no llama.cpp server, it can offer an in-app setup wizard to install llama.cpp for you.

• Backups GitHub Gist backup is supported, and local versioned backups are supported through the app server. OneDrive and Google Drive are listed in the UI but are not implemented yet.

• In-app update staging Writingway can detect newer builds, download an update, stage it locally, and apply it the next time you restart from the launcher.

Requirements

All platforms

• Python 3
• A modern browser

Optional for local GGUF mode

• A .gguf model placed in models/
• llama.cpp server files in llama/ Or let the app install them through the setup wizard when supported.

Quick start

Windows

1. Download and extract the project.
2. Double-click start.bat.
3. Open Writingway in the browser window it launches.

macOS / Linux

1. Download and extract the project.
2. Run:

chmod +x start.sh
./start.sh

1. Open Writingway in the browser window it launches.

First-run local AI flow

If you already placed a .gguf model in models/:

• Writingway checks whether llama.cpp is installed
• If it is missing, Writingway can show a setup wizard
• The wizard can install a supported llama.cpp build
• After installation, restart Writingway from start.sh or start.bat

If you do not want local AI, just skip the wizard and use an API provider instead.

AI modes

API / Local API

Use this for:

• OpenRouter
• Anthropic
• OpenAI
• Google AI
• NanoGPT
• LM Studio
• Custom OpenAI-compatible endpoints
• Ollama if exposed through a compatible API layer

This is the best choice if you want the simplest setup.

Local GGUF Model

Use this only when both are true:

• You have at least one .gguf file in models/
• llama.cpp server files are installed in llama/

Writingway hides this option when the local backend is not actually available, so users are less likely to end up in a broken configuration.

Launchers and local services

The launchers do a few important things for you:

• Start the Writingway app server on http://127.0.0.1:8000
• Start the updater service on http://127.0.0.1:8001
• Start llama.cpp on http://127.0.0.1:8080 when local GGUF mode is available
• Apply staged updates on the next start

Use the launcher scripts instead of opening main.html directly.

Saving and backups

Manual project save

The disk save button writes the current project snapshot to the projects/ folder through the local app server.

Local versioning backup

Writingway can create timestamped JSON backups in:

project-backups/

This gives you local restore points without needing a cloud account.

GitHub Gist backup

Writingway can back up a project to a private GitHub Gist if you provide a GitHub token with gist scope.

Not implemented yet

These appear in the backup provider selector, but are not functional yet:

• OneDrive
• Google Drive

Updates

Writingway compares the latest GitHub commit date on main with the local build date in src/update-checker.js.

If a newer build is available:

• Writingway can download and stage the update
• You restart Writingway manually
• The launcher applies the staged update on startup

On Windows and Linux/macOS, the staged update is applied by start.bat or start.sh on the next launch.

Writingway 1 import

Writingway includes a Writingway 1 importer for older projects. It can bring over project structure and content so you can continue working in Writingway 2.

Project structure

A few important folders:

models/           Optional GGUF model files
llama/            Optional llama.cpp server files
projects/         Manual project saves written by the app server
project-backups/  Local versioned backups
tools/            Local Python services
src/              App source

Development notes

This repo includes a small test setup in package.json. Available scripts:

npm run smoke
npm run unit
npm run ui
npm test

Current status

What is working now:

• Writing and scene management
• Compendium/worldbuilding
• AI provider configuration
• LM Studio integration
• Local GGUF mode through llama.cpp
• In-app llama.cpp setup flow on supported platforms
• Manual project save to disk
• Local versioned backups
• GitHub Gist backup
• Update detection and staged update download

What is intentionally incomplete:

• OneDrive backup
• Google Drive backup
• Fully automatic restart/apply after update download
• Broader local installer coverage for every llama.cpp release variant

Troubleshooting

Writingway opens but local GGUF mode is unavailable

Check that:

• A .gguf file exists in models/
• llama.cpp is installed in llama/
• You restarted the launcher after installation

The browser says it cannot connect on startup

Use the launcher scripts, not main.html directly. The launchers wait for the local app server before opening the browser.

Update downloaded but nothing changed

Restart Writingway using start.sh or start.bat. The staged update is applied by the launcher on startup.

Backups are enabled but cloud providers are missing

Only GitHub Gist and Local Versioning are currently implemented. OneDrive and Google Drive are placeholders in the UI for future work.