Models

Settings → Models is where you choose the AI models OrbySystems uses. See LLM models for the ideas behind them; this page covers the settings.

Trusted Models

OrbySystems Small, OrbySystems Medium and OrbySystems Large are the models OrbySystems is built and tested for. They run on your computer. The one that suits your computer's memory is recommended, and it's downloaded the first time it's needed. See Hardware requirements for which one your computer gets, and Install OrbySystems for what to expect the first time.

Open Source and Cloud Models

If you want a different chat model, choose one here: a larger open-source model that runs on your computer, or a model you connect through a local server (such as LM Studio) or a cloud provider. The model you choose reads your data just as the OrbySystems models do, and a cloud model receives every message you send. OrbySystems isn't tested with these models, so its answers can be less reliable. This is entirely optional: OrbySystems works fully offline.

Embedding Model

OrbySystems Embedding indexes your documents and memory so OrbySystems can search and recall them. It runs on your computer, alongside the chat model.

Open Source and Cloud Embedding Models

You can choose a different embedding model: an open-source one that runs on your computer, or one you connect through a local server or a cloud provider. A cloud one is sent the text it indexes from your documents and transactions.

Before you rely on a model

Use Model Test to confirm a newly configured model works correctly with OrbySystems.