Skip to content
kitn AI/UI

Drop-in chat

One element gives you the whole experience: a message thread with markdown, reasoning, and tool calls, a composer, and optional header chrome. You bring the model; <kai-chat> renders the UI. The thread starts empty with a few starter suggestions — click one to begin, and the reply streams in.

Wire it up

Section titled “Wire it up”

Start with an empty thread and a few suggestions. The chips render in the empty state and disappear once the conversation has a message, so they read as conversation starters. Assign both in JavaScript (arrays can’t be HTML attributes), then handle kai-submit: append the user’s turn, add an empty assistant message, and grow its content as tokens arrive.

<kai-chat id="chat" chat-title="Assistant" placeholder="Ask anything…"></kai-chat>


<script type="module">
  import '@kitn.ai/ui/elements';
  const chat = document.getElementById('chat');
  chat.messages = [];
  chat.suggestions = [
    'What makes kitn-chat different from rolling my own?',
    'Show me a streaming example',
  ];


  chat.addEventListener('kai-submit', async (e) => {
    const prompt = e.detail.value;
    const aId = crypto.randomUUID();
    // append the user turn + an empty assistant message
    chat.messages = [
      ...chat.messages,
      { id: crypto.randomUUID(), role: 'user', content: prompt },
      { id: aId, role: 'assistant', content: '' },
    ];
    chat.loading = true;


    // stream tokens from your backend into the assistant message
    for await (const token of streamFromYourModel(prompt)) {
      chat.messages = chat.messages.map((m) =>
        m.id === aId ? { ...m, content: m.content + token } : m,
      );
    }
    chat.loading = false;
  });
</script>

The key move is reassigning chat.messages to a new array on each token — that’s what drives the re-render.

Next steps

Section titled “Next steps”