James Abrahams Demonstrates Drupal Chrome Browser AI with Local Processing Workflows

Developer James Jackson Abrahams has published a video walkthrough of a Drupal demonstration built by Marcus Johansson that uses Chrome’s Prompt API to run AI-assisted workflows locally in the browser. The announcement was shared in a LinkedIn post describing the demo as an exploration of local, browser-based AI patterns within Drupal.

The walkthrough presents a high-level overview of features available on the demo page, including alt text generation, CKEditor assistants, summarisation, and translation. These functions are executed in the browser rather than via external model APIs.

The implementation aligns with the author's description of “sovereign AI” patterns, in which processing and prompts remain on the user’s device. This approach may be relevant for teams exploring privacy-sensitive or infrastructure-controlled AI use cases.

The demo is supported by the experimental Browser Provider module, which integrates Chrome’s Prompt API into Drupal. According to the project documentation, the module executes prompts locally within the user’s browser session and is intended for experimentation and prototyping.

The module requires a compatible version of Google Chrome with experimental Prompt API features enabled, depends on the browser remaining active during processing, and may require significant system resources. The documentation also notes that the approach relies on evolving browser capabilities and is not yet suitable for production use.

The repository associated with the demo provides a local development setup using DDEV and links to a hosted test environment for evaluation. A public demo environment is also available via Drupal Forge, where users can test the Chrome Prompt API integration directly in a browser-based Drupal instance.

In a follow-up comment on the LinkedIn post, Abrahams credited Marcus Johansson for building the demo, along with contributors who hosted and supported the project.

Further details are available through the video walkthrough and project repository.