Story Teller Docs / Search, Memory & Privacy

Search, Memory & Privacy

Story Teller is local-first: your story lives on your own computer, and the app is built to keep it that way. This page explains how it searches your material and — plainly — what stays on your machine and what does not.

Searching your story

Story Teller offers two kinds of search over your libraries and notes:

  • Full-text search matches the words you type. It is always available and powers the Search panel, so you can find any passage across your documents and notes.
  • Meaning-based (semantic) search finds material that is about what you asked, even when the words differ. It supports the assistant’s ability to pull in relevant context (the RAG feature). Meaning-based search depends on embeddings, which you enable by configuring an embedding model in Settings .

Story Teller can store your search index using one of two backends. The default backend is fully supported. A second, LanceDB-based vector backend is offered as an option but its meaning-based search is still being completed — if you want semantic search today, use the default backend. Full-text search works regardless of the backend you choose.

What stays on your machine

Your story data is local. The manuscript, your notes, characters, world, timeline, and search index are all stored on your own computer:

  • Your manuscript is plain Markdown files on disk — you can open, back up, and move them yourself.
  • Everything else lives in a local database in the app’s data folder.
  • The local engine that does the memory and search work listens only to the app, on your own machine. It is not reachable over the network.
  • Your license key is stored encrypted locally, and license checks are done offline — the app does not phone home to validate it.

What leaves your machine

There is one thing that can leave your computer: the text you send to the AI model. Story Teller needs a language model to help you write, and you choose which one in Settings :

  • If you point it at a model that runs on your own machine (a local provider), nothing leaves your computer — prompts, prose, and notes all stay on-device.
  • If you point it at a cloud provider, the content of your prompts (which can include your prose and relevant notes) is sent to that provider to generate a response, subject to their terms.

So local-first storage does not automatically mean nothing leaves the machine. If keeping your manuscript entirely on-device matters to you, choose a local model. The choice is yours, and you can change it at any time.

← PreviousThe Backlog & Topic Promotion