# Ollama and models

Curator's AI features run entirely on your Mac, through a program called **Ollama** and the AI models it runs.
This page explains both.

## What an LLM is

An LLM, or large language model, is a program that understands and generates text.
It is trained on a very large amount of written material, which is how it learns to
read a document, answer a question, or write a short summary. When Curator describes
a file or answers a question in **Chat**, an LLM is doing that work.

You do not need to know how a model is built to use it. You just need one running
somewhere Curator can reach. On your Mac, that is Ollama.

## What Ollama is

**Ollama** is a free, open-source app that runs AI models directly on your Mac.
Curator depends on it for every AI feature.

Your files and your questions stay on your machine. Nothing is uploaded, there are no API keys to manage, and there
is no per-use cost. The only network call Curator makes for AI is to Ollama running on
your own Mac, or to one you point it at on your own network.

Curator can install Ollama for you in one click during setup, so you do not have to
set it up by hand.

## What a model is

A **model** is an AI model that runs inside Ollama. Different models are good
at different things, and they come in different sizes.

Before Curator can use a model, you **pull** it, which simply means download it
through Ollama. Each model takes up disk space while it sits on your Mac, and it uses
RAM while it is actively running. Bigger models are more capable but need more RAM, so
Curator recommends models that fit your Mac.

## A model for each job

Curator does not use one model for everything. It assigns a model to each kind of job,
so the right tool handles the right task.

| Job | What it does |
| --- | --- |
| **Vision / Analysis** | Reads images and scanned PDFs. |
| **OCR** | Pulls text out of scanned pages and images. |
| **Embeddings** | Powers search by meaning. |
| **Chat** | Answers your questions about your files. |
| **File Processing** | Writes the description and details for each file. |
| **Organization** | Builds the smart folders on the **Organize** screen. |

During setup, Curator detects how much RAM your Mac has and recommends models sized to
fit, with a one-line description and download size for each. You can change these
assignments later in **Settings** under **AI & Models**.

> [!IMPORTANT]
> AI features stay off until the models they need are downloaded. You can use the file
> browser and search by filename right away, but search by meaning, chat, and
> organization wait until at least the needed models are pulled.

## Learn more

[Model assignments](/reference/models/)
  [Set up Curator](/get-started/setup/)