@chai-ui/vue

@chai-ui/vue

Chai UI for Vue: drop-in components for the input and output surfaces of generative-media apps: a prompt composer, a media-to-prompt analyzer, a result card and an edit card, with the AI work built in and wired to Fal.ai, OpenRouter, Vercel AI Gateway or your own engine. Everything is controlled: you own the state with v-model, and the components render it.

These are the same components as @chai-ui/react, with the same behavior. Both packages are thin layers over @chai-ui/core and run one shared test suite, including the WCAG 2.2 AA checks (axe, keyboard and focus).

Building with it, or pointing a coding agent at it? Read DESIGN.md (shipped in the package) for the usage rules, and the docs site's llms.txt for the API. The code boxes on chai-ui.com/docs switch between React and Vue.

Install

pnpm add @chai-ui/vue @chai-ui/core @chai-ui/tokens

Vue 3.5 or later, Nuxt included. No Chai-specific build setup.

// main.ts: tokens first, then the component styles that read them
import "@chai-ui/tokens/css";
import "@chai-ui/vue/style.css";

Quick start

<!-- App.vue -->
<template>
  <ChaiProvider><Create /></ChaiProvider>
</template>

<!-- Create.vue -->
<script setup lang="ts">
import { computed, ref } from "vue";
import { Composer, ResultCard, createFalEngine, useComposer } from "@chai-ui/vue";

const engine = createFalEngine();
const value = ref("");
const { run, submit, cancel, enhance, enhancing } = useComposer({ engine });
const busy = computed(() => run.value?.results.some((r) => r.status === "queued" || r.status === "running") ?? false);

async function onEnhance(prompt: string) {
  value.value = await enhance.value!(prompt);
}
</script>

<template>
  <Composer v-model="value" :submitting="busy" :enhancing="enhancing" @submit="submit" @abort="cancel" @enhance="onEnhance" />
  <ResultCard v-if="run" :results="run.results" :prompt="run.request.prompt" @action="() => {}" />
</template>

To try it before setting up a provider, leave out engine and ChaiProvider: everything then runs on built-in mocks, with no network calls and no keys.

The Vue names

The props are React's, with Vue's conventions:

Some controls show only with a listener, as in React: the optimize button with @enhance, stop with @abort, the Model menu with @update:model-id.

Docked the bar? Set menu-placement="above" for a bar fixed to the bottom of the screen, or "below" for one fixed to the top, so every menu opens into the page. Left out, it's "auto": each menu goes where it fits.

What's in the package

• Composer: prompt, attachments, use cases, model and aspect-ratio menus with suggested models, prompt optimization, auto-select routing, stop and retry.
• MediaAnalyzer + useMediaAnalyzer: media in, a generation-ready prompt out.
• ResultCard: one prompt's results, paged, with votes, retry, download, share, details, streaming text and quality-check tags (from your backend, or the opt-in judge: useComposer({ judgeModel, judgeChecks })).
• Analytics: @event on Composer and ResultCard reports what the person does and sees, one ChaiEvent each, for any analytics tool: @event="(e) => posthog.capture(e.name, e.properties)".
• EditCard: marked regions with their own instructions, and each edit as a new version.
• useComposer, ChaiProvider, and the Toggle, SearchMenu and Pagination building blocks.
• The engines and types from @chai-ui/core, re-exported.

Your API keys stay on your server

The engines call one route on your server, which adds the key. In Nuxt:

// server/api/chai/[...path].ts
import { createChaiHandler } from "@chai-ui/core/server";

const chai = createChaiHandler();
export default defineEventHandler((event) => chai.handle(toWebRequest(event)));

Then set FAL_KEY, OPENROUTER_API_KEY or AI_GATEWAY_API_KEY on your server. Other servers, and authorize and allowedModels for production, are in the @chai-ui/core README.