@beadkit

Shared card shell and xAPI telemetry. Every learning component depends on this; installing a component pulls it in automatically.

Inline concept check with immediate feedback and an xAPI answered statement.

Two-sided card with an xAPI experienced statement on first reveal.

Shuffled steps the learner drags back into order, with equivalent keyboard reordering, live-region announcements, and an xAPI answered statement carrying the s…

Per-skill mastery bars plus a learner self-rating, emitting an xAPI rated statement.

Parses <LEARNING_COMPONENT> payloads out of streaming model output and hydrates them inline. Pulls in every component it can render.

Shared motion styles and the upstream Beautiful UI MIT notice used by Beadkit's AI interface primitives.

Pixel-grid AI loading state with Drive, Dots, and Orbit variants, a shimmer label, elapsed time, and reduced-motion support.

Expandable AI trace for step, reasoning, search, and coding progress with reusable row data and reduced-motion support.

Approved deterministic Thinking scenarios used by the Beadkit workbench.

Paged human-in-the-loop questions with single- and multi-select answers, custom input, and submission state.

Progressively revealed tool calls with expandable details and file diff summaries.

Expandable agent task rows with pending, running, failed, and completed states plus Capsule and List layouts.

Agent recommendation with confidence meter, alternatives drawer, and acceptance callback.

Retrieved context chunks with metadata and linked source attribution.

Line-streaming code surface with typed syntax tokens, static output mode, and accessible copy feedback.

Paced response text with inline citations, sources, actions, follow-ups, and reduced-motion behavior.

Rounded or Pill composer with source mentions, slash commands, attachments, model selection, and dictation state.

Async rewrite actions for selected text with custom instructions, retry, keep, and discard controls.

Tabbed AI SDK chat with a reusable prompt bar, typed learning tools, streamed Markdown, and compatibility with the learning-component protocol.

AI-proposed table changes with staged reveal and row-level or bulk accept/reject decisions.

CRM-style sortable and selectable records grid with tags, relationship status, and links.

Controlled or uncontrolled status chips that filter a responsive semantic data table.

Expanded or collapsed workspace navigation with sections, counts, actions, and quick search.

Live command search with keyboard navigation, result callbacks, clearing, and an empty state.

Paged agent insights with responsive, pointer- and keyboard-scrubbable SVG charts.

Controlled or uncontrolled design inspector for layout, dimensions, radius, opacity, and type.