Datasette-apps 0.2a0 Released with Hidden Iframe Testing for AI Agents
Simon Willison has released datasette-apps 0.2a0, introducing new tools that allow AI agents to list, edit, and test hosted applications. Specifically, it adds the `app_debug()` tool for invisible testing and the `app_list()` tool for listing editable apps. This release enables AI agents to perform automated smoke testing and UI measurements on web applications within a sandboxed environment. It highlights a growing trend of LLM-powered agents interacting directly with and verifying the code they generate or edit. The `app_debug()` tool runs agent-provided JavaScript inside a sandboxed, invisible iframe using `opacity: 0` and `pointer-events: none`. This functionality relies on the `context.browser_task()` mechanism introduced in datasette-agent 0.4a0.
## BACKGROUND
Datasette is an open-source tool designed for exploring, analyzing, and publishing data stored in SQLite databases. Datasette Apps is a plugin that allows users to host and edit single-file HTML, CSS, and JavaScript applications directly inside Datasette, while Datasette Agent is an LLM-powered assistant that helps users interact with their data and applications.