Radix-style React primitives for AI chat with adapters for AI SDK, LangGraph, Mastra, and custom backends.
The full list rather than the rail: grouped by family, priced, with what each one can do.
The props-driven model picker: search, groups, and reasoning effort levels, no runtime required.
Model picker with reasoning effort levels, search, and runtime integration.
Animated number that rolls its digits odometer style when the value changes.
Digits that roll into place as a count updates in real time.
First run: three moves that teach what this assistant is actually for.
The agent asks a question with a few answers; the pick returns to it and stays as a receipt.
Calls that went out together, collapsed to one row until you want the detail.
Granting a capability rather than approving one action, with the reach spelled out.
Prompts you saved, searchable, with their variables shown before you insert one.
A few short questions asked one at a time, answered together and kept as a receipt.
How much is left, when it comes back, and the way to get more.
Select and quote message text with a floating toolbar and a composer preview.
Select a phrase in an answer and a toolbar offers to quote, explain, or rewrite it.
An answer played back, the spoken word lit as it goes, speed under your thumb.
Collapsible panel for assistant reasoning and thinking content.
How hard to think, and how much of that budget the run actually spent.
A collapsible trace that streams reasoning steps along a timeline, then settles into a summary.
The props-driven reasoning disclosure: streaming preview, fades, and variants, no runtime required.
Keeps focus on a receipt when the focused answer is replaced.
The agent proposes a change with its confidence, and waits for a yes.
Fork the same turn to a different model instead of rolling the same dice.
An outline that fills in section by section, each carrying the sources behind it.
The passages a retrieval answer stands on, scored, before the answer itself arrives.
The same diff, but each hunk is a decision: keep it, discard it, apply what survived.
Ordered commentary and tools, with a live summary and a final answer that stays visible.
A run that repeats on its own, with its cadence and how it has been doing.
A verdict with its arithmetic shown: criteria, weights, and what pulled it down.
Streaming never steals your scroll position; a pill offers the way back down.
Find inside a long thread, with every hit marked down the scrollbar.
Dropdown select with composable trigger, content, and item components.
Which servers are connected, what each one brought, and which is still waiting on you.
Model, system prompt, temperature, and what the assistant is allowed to do.
A read-only transcript someone sent you, with a way to pick it up yourself.
The props-driven Shiki code block: streaming-aware highlighting, no runtime required.
Code block highlighting powered by react-shiki.
Keyframes and theme variables for the streaming shimmer animation.
Citations collapsed into a pill, expanding into scannable source cards.
Display URL sources with favicon, title, and an external link.
Who is talking, once a thread holds more than a user and one model.
The most common structured answer after a table: one object, labeled.
You pressed stop. The half-written answer stays, and continuing is one tap away.
Tokens arrive softly: the newest words land in blue and settle into ink.
Parallel workers with their own progress, models, and completions.
Code block highlighting powered by react-syntax-highlighter.
Tabs for organizing content into switchable panels.
A delegated task with its state, timing, result, and transcript in one card.
A runtime task card with nested transcripts, timing, results, and grouped task lanes.