Most Drupal AI integrations send content to an external API. But what if the language model could run directly inside the user’s browser?
In this session, I will demonstrate Offline AI, an experimental Drupal module that uses WebLLM and WebGPU to perform AI inference on the user’s device. After the initial model download, prompts and Drupal content can be processed locally without an API key or sending the content to an external AI provider.
Through a live demonstration, attendees will see how Drupal can provide local chat, content summarization, rewriting, question answering, and retrieval-augmented generation using selected content.
The session will explore where this approach can be useful, including:
- Privacy-sensitive content workflows
- Internal knowledge assistants
- Educational and research websites
- Low-connectivity or offline environments
- AI experimentation without API credentials or usage charges
Browser-based AI also has important limitations. Model downloads can be large, performance depends on the user’s hardware, WebGPU support varies, and smaller local models cannot match every cloud model.
Attendees will leave with a practical understanding of how browser-based AI works in Drupal and a simple framework for deciding when to use local, self-hosted, or cloud-based inference.