Settings & AI Configuration
Story Teller’s settings are where you choose the AI model it uses and tune how the app behaves. The model choice matters most — it determines both the quality of the writing help and, as covered in Search, Memory & Privacy , whether anything leaves your machine.
Choosing a model provider
Story Teller can talk to several kinds of language model:
- Local — a model running on your own machine (for example via LM Studio or a local vLLM server). Everything stays on-device.
- OpenAI — OpenAI’s hosted models.
- Azure AI — models hosted on Azure.
For each provider you supply the details it needs (an endpoint address, and an API key for the hosted providers). A Test button checks the connection and detects whether the model supports the tool-calling the assistant relies on.
Mapping think levels to models
The Chat Assistant offers four think levels — Quick, Story, Thoughtful, and Research — and you decide which model answers at each. You might map Quick to a small fast local model and Research to a larger one, for example. There is also a Default Tools Model used as a fallback when a task needs tool-calling.
Embeddings for meaning-based search
To turn on meaning-based search and richer context for the assistant, configure an embedding model (an Embed category alongside the think levels). Without it, search falls back to matching words — see Search, Memory & Privacy .
Other settings
| Setting | What it does |
|---|---|
| License | Register your license key (see Licensing & Trial ) |
| Theme | Choose from several color themes |
| Backlog threshold | How many chat exchanges accumulate before the assistant offers to organize them — see The Backlog |
| Page size | How many items lists show at once |
| Search backend | Which backend stores your search index (see the note in Search, Memory & Privacy ) |
| Rebuild Topics | Regenerate all topics and timelines from your project history |
Rebuild Topics is destructive and cannot be undone — it deletes your existing topics, memoranda, and timeline events and regenerates them from scratch. Use it only when you deliberately want a clean rebuild.
Related pages
- Search, Memory & Privacy — local versus cloud models
- The Chat Assistant — the think levels you map to models
- Licensing & Trial — registering your key