Introduction to AI Assistant
Nutrient iOS SDK supports two AI Assistant variants: AI Assistant server and on-device AI Assistant. Both variants enable users to ask questions about documents in a chat interface.
The main difference is where each variant processes document content:
- AI Assistant server processes documents on your server instance. Refer to the AI Assistant overview guide for server product details. You can provide your own large language model (LLM) or use a hosted model from a provider like OpenAI. Refer to the LLM hosting strategy guide to compare those options.
- On-device AI Assistant processes documents on users’ devices. By default, the LLM comes from Apple’s Foundation Models(opens in a new tab) framework, which powers Apple Intelligence. Starting with iOS 27, you can use a custom model that conforms to Foundation Models’
LanguageModel(opens in a new tab) protocol. Nutrient iOS SDK’s retrieval pipeline searches large documents without loading the entire document into the model’s context window. Refer to the on-device AI Assistant guide for setup details.
Choosing between AI Assistant server and on-device AI Assistant
Use this table to choose the variant that matches your document storage, privacy, and platform requirements.
| AI Assistant server | On-device AI Assistant | |
|---|---|---|
| AI capability | Uses your choice of server-scale model. The server pipeline helps find information and reason over document content. | Uses Apple’s Foundation Models by default. Starting with iOS 27, it also supports custom Foundation Models language models. Nutrient’s retrieval pipeline supports large documents. |
| Integration setup | Requires AI Assistant server and your backend for authentication. | Uses Apple’s system model. Custom models require Foundation Models integration on iOS 27 or later. |
| Document ownership | Choose this option when your backend stores documents. | Choose this option when users provide their own documents. |
| Privacy | You’re responsible for your users’ data. Documents are sent to your AI Assistant server instance. If you use a hosted model, questions and relevant document content go to the model provider. | Documents and questions stay on the user’s device when you use Apple’s system model. Custom models may run on-device or call a server. Document contents are copied into a local database to improve load time. |
| Network | Requires a connection to AI Assistant server. | Works offline with Apple’s system model. A custom model may need a network connection, depending on its implementation. |
| Cross-platform compatibility | Works across Web, Android, iOS, macOS, and visionOS. | Works only on iOS, macOS, and visionOS. |
| Apple compatibility | Works on any device supported by Nutrient iOS SDK. | Apple’s system model requires a device compatible with Apple Intelligence(opens in a new tab) on iOS 26 or later with Apple Intelligence enabled. Requirements for custom models depend on their implementation. |
| Nutrient licensing | Requires licenses for AI Assistant and Nutrient iOS SDK. The iOS client doesn’t need an extra license feature. | Requires the on-device AI Assistant license feature for Nutrient iOS SDK. |
Integrating AI Assistant with Nutrient iOS SDK
Choose the setup guide for the variant you want to integrate.
- Refer to the getting started with AI Assistant guide to set up AI Assistant server.
- Refer to the on-device AI Assistant guide to configure on-device AI Assistant.