Settings¶
Storywright's settings let you configure your AI provider, choose models, and customize the app to your workflow. Changes are saved immediately — there's no save button to forget.
Accessing Settings¶
Click the Settings tab (gear icon) in the navigation sidebar, rail, or bottom bar depending on your platform and screen size. Every change you make takes effect right away.
Settings are organized into four groups:
- Connection — Providers and Models
- Creative — Prompts, Structures, Content
- System — Appearance, Storage, Memory, Debug
- Info — About (version info, re-run wizard, take a tour)
Providers¶
Storywright calls LLM APIs directly from your device. There is no Storywright backend — your API key and data never leave your machine.
Providers¶
Storywright supports multiple LLM providers simultaneously. Each provider has its own API key, base URL, and model list.
| Provider | Description | API Key Required |
|---|---|---|
| NanoGPT (default) | 100+ models, one key, pay-per-token | Yes |
| OpenAI | GPT-4.1, o4-mini, embeddings | Yes |
| Google Gemini | Gemini Flash, Gemini Pro — native Google AI API | Yes |
| Ollama | Local models, offline capable | No |
| LM Studio | Local models with GUI | No |
| Custom | Any OpenAI-compatible endpoint | Varies |
Web users: Local providers (Ollama, LM Studio) require CORS to be enabled. In LM Studio, enable the CORS toggle in the Developer tab. For Ollama, set
OLLAMA_ORIGINS=*before starting, or use Ollama 0.2.0+.
Add providers in Settings → Providers. Each gets its own card where you configure the API key, base URL, and test the connection.
You're not locked in. Use one provider or several. NanoGPT is just a convenient default because it gives you access to many models without managing multiple accounts.
Test Connection¶
After configuring a provider, click Test Connection to verify everything works. You'll see either a success confirmation or specific error details (invalid key, unreachable URL, etc.). Always test after making changes.
Default Provider: NanoGPT¶
Storywright ships with NanoGPT as the recommended default. NanoGPT aggregates many AI models under a single API key:
- Wide model selection — GPT-4.1, Claude, Gemini, Llama, Mistral, and many others, all accessible with one key.
- Pay-per-token — no subscription required. You pay only for what you use.
- Easy setup — sign up at https://nano-gpt.com, grab your API key, and paste it into Storywright.
You're not locked in. Any OpenAI-compatible provider works. nano-gpt is just a convenient default because it gives you access to many models without managing multiple accounts.
Model Selection¶
Storywright uses three model roles. Each one is configurable independently so you can balance quality, speed, and cost.
Writing Model¶
Used for scene generation and revision. This is where quality matters most — the writing model produces the actual prose your readers will see.
Recommended: Use the best model you can afford. Examples: GPT-4.1, Claude Sonnet 4, Gemini 2.5 Pro, Euryale 70B.
Planning Model¶
Used for scene planning and plan refinement. Needs good reasoning ability to break down story beats, but doesn't need to produce polished prose.
Recommended: Can be the same as your writing model, or a slightly cheaper alternative to save cost. Examples: o4-mini, Claude Opus 4, Gemini 2.5 Pro.
Extraction Model¶
Used for summaries, memory extraction, and continuity updates. These are structured extraction tasks — the model follows a schema and pulls facts from text. Creativity isn't important here.
Recommended: A fast, cheap model works great. Examples: GPT-4.1-nano, GPT-4.1-mini, Claude Haiku 3.5, Gemini 2.0 Flash.
How Model Selection Works¶
- Models are selected from a dropdown populated by querying your API for available models. If you don't see a model you expect, verify your API key has access to it.
- Per-story overrides: You can set different models for individual stories in the story workspace. This is useful if you want to use a premium model for your main project and a cheaper one for experiments.
Writing Styles (Named Prompts)¶
Writing styles are system prompt templates that shape the AI's tone, pacing, and prose style. Each style auto-appends relevant prompt blocks (writing craft guidelines, banned phrases, etc.) to keep the output consistent.
| Style | Best For |
|---|---|
| literary-fiction | Rich worldbuilding, dreamy atmospherics. Flexible POV. |
| genre-fiction | Plot-driven, pace and tension. Thriller, mystery, sci-fi, fantasy. |
| sci-fi | Ideas-driven science fiction. Worldbuilding rigor, protagonist competence, human cost of the speculative concept. |
| mystery-thriller | Propulsive mystery and thriller. Fair-play clues, escalating stakes, information revealed at the right moment. |
| fantasy | Immersive fantasy. Magic with rules, worldbuilding through character action, sense of a world beyond the page. |
| humor | Comedy-forward. Sharp dialogue, absurd situations, comic timing. |
| romance | Character-driven. Emotional depth, tension, and connection. |
| horror | Atmospheric dread, sensory unease, controlled revelation. |
| minimalist | Sparse prose, short sentences. Carver-Hemingway style. |
| audio-storytelling | Written for spoken audio. Dreamy, sleep-appropriate. Includes SFX/AMBIENT tags for audio pipelines. |
Writing styles are selected per-story in the story workspace's Story Setup section — not globally in Settings. The styles themselves are defined here for reference, and you can preview what each one does.
Preferences¶
UI Scale¶
Adjust the interface size to your comfort. Useful on high-DPI displays or if you prefer larger text.
Content Level¶
Control what content is visible in the app and what the AI is allowed to generate. Storywright uses a tag-based content level system with four built-in levels:
| Level | What's Allowed | Emoji |
|---|---|---|
| 🌈 Kids | Only untagged content. Lighthearted themes only. | |
| 📖 General | Action, drama, romance, tension. Nothing explicit. | |
| 🌙 Mature | Adult themes, darker subjects, intense situations. Not explicit. | |
| 🔓 Unrestricted | No content restrictions. Full creative freedom. |
How it works:
- Stories, characters, lorebooks, and prompts can be tagged with content descriptors (violence, horror, romance, dark, profanity, mature, explicit).
- Your active content level determines which tags are allowed.
- Entities with disallowed tags are hidden from the UI (not deleted — just hidden until you change levels).
- Generation prompts automatically include content boundary instructions based on which tags are disallowed at your level.
Changing levels: Settings → Content Level. Changes take effect immediately. You can also select your level during the onboarding wizard.
Custom levels: Toggle individual content tags on/off to create a custom configuration. Save it with a name to reuse later.
Storage Path¶
Where your stories, characters, lorebooks, and other data are saved on disk.
| Platform | Default Location |
|---|---|
| macOS | ~/Documents/Storywright/ |
| Windows | %USERPROFILE%\Documents\Storywright\ |
| iOS / Android | App sandbox (managed automatically) |
| Web | Browser local storage (not persistent across browsers — export regularly) |
On desktop, you can change this to any folder you like — for example, a Dropbox or iCloud folder for cross-device sync.
Re-running Onboarding¶
If you need to redo the initial setup wizard, go to Settings → About → Re-run Setup Wizard. To replay the guided workspace tour, use Settings → About → Take a Tour.
This is useful if you:
- Switched to a different API provider and want the guided setup flow.
- Want a refresher on Storywright's features and workflow.
- Accidentally skipped a step during first setup.
Prompts¶
Browse and preview the available writing style prompts. Each style auto-appends relevant prompt blocks (writing craft guidelines, banned phrases, etc.) to keep the output consistent. Styles are selected per-story in the story workspace — this page is for reference and preview.
Structures¶
Browse story structure templates (e.g., Three Act, Hero's Journey) and their beat descriptions. Structure templates guide the planning model when it generates a scene outline.
Debug¶
Toggle diagnostic settings for troubleshooting:
- File logging — write structured logs to disk.
- Verbose LLM logging — log full request/response payloads for LLM calls.
- View log files — open the logs directory directly.
Log files are saved to your storage directory under logs/.
About¶
Displays the current app version and build number. Also provides:
- Re-run Setup Wizard — relaunch the onboarding wizard.
- Take a Tour — replay the guided workspace tour.
Tips¶
- Use a fast, cheap model for extraction. It just needs to follow instructions and extract structured data — it doesn't need to be creative.
- Slow generation? Try a faster writing model or reduce your scene length targets in the story workspace.
- Always test your connection after changing the API key or base URL.
- Using a local model? Set up an Ollama or LM Studio provider in Settings → Providers. The base URL defaults to the right localhost port.
- Per-story overrides let you use a premium model for your main project and a budget model for brainstorming — without changing global settings each time.
Related Guides¶
- Getting Started — initial API setup walkthrough
- Generation — how models are used during generation
- Stories — per-story model overrides and writing style selection