askable

It lets a language‑model receive the exact UI element a user is looking at by adding a single `data‑askable` attribute to any meaningful DOM node. When the element gains focus—via hover, click, or other interaction—the library captures the attached JSON metadata and makes it available through `askable.toPromptContext()`. This context is then injected at the AI boundary, so the model works with the user’s visual focus instead of the whole page.

The tool is framework‑agnostic and provides lightweight adapters for React, Vue, Svelte, and plain HTML, allowing developers to annotate existing components without rewriting them. Because the metadata is already part of the UI rendering logic, no extra data pipelines, prompt templates, or mapping layers are required. It is suited for dashboards, forms, tables, and other interactive interfaces where the assistant should respond to the element the user is inspecting.

Its distinctive approach is to tie the model’s input directly to the user’s attention, delivering a concise JSON payload that reflects the currently focused element. This reduces the need for full‑page serialization and enables real‑time, context‑aware AI interactions across any front‑end stack.