File: README.md
json-render
The Generative UI framework.
Generate dynamic, personalized UIs from prompts without sacrificing reliability. Predefined components and actions for safe, predictable output.
# for React
npm install @json-render/core @json-render/react
# for React with pre-built shadcn/ui components
npm install @json-render/shadcn
# or for React Native
npm install @json-render/core @json-render/react-native
# or for video
npm install @json-render/core @json-render/remotion
# or for PDF documents
npm install @json-render/core @json-render/react-pdf
# or for HTML email
npm install @json-render/core @json-render/react-email @react-email/components @react-email/render
# or for Vue
npm install @json-render/core @json-render/vue
# or for Svelte
npm install @json-render/core @json-render/svelte
# or for SolidJS
npm install @json-render/core @json-render/solid
# or for terminal UIs
npm install @json-render/core @json-render/ink ink react
# or for full Next.js apps (routes, layouts, SSR, metadata)
npm install @json-render/core @json-render/react @json-render/next
# or for 3D scenes (and gaussian splatting via the GaussianSplat component)
npm install @json-render/core @json-render/react-three-fiber @react-three/fiber @react-three/drei threeWhy json-render?
json-render is a Generative UI framework: AI generates interfaces from natural language prompts, constrained to components you define. You set the guardrails, AI generates within them:
- Guardrailed - AI can only use components in your catalog
- Predictable - JSON output matches your schema, every time
- Fast - Stream and render progressively as the model responds
- Cross-Platform - React, Vue, Svelte, Solid (web), React Native (mobile) from the same catalog
- Batteries Included - 36 pre-built shadcn/ui components ready to use
Quick Start
1. Define Your Catalog
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react/schema";
import { z } from "zod";
const catalog = defineCatalog(schema, {
components: {
Card: {
props: z.object({ title: z.string() }),
},
Metric: {
props: z.object({
label: z.string(),
value: z.string(),
format: z.enum(["currency", "percent", "number"]).nullable(),
}),
},
Button: {
props: z.object({
label: z.string(),
action: z.string(),
}),
},
},
actions: {
export_report: { description: "Export dashboard to PDF" },
refresh_data: { description: "Refresh all metrics" },
},
});2. Define Your Components
import { defineRegistry, Renderer } from "@json-render/react";
const { registry } = defineRegistry(catalog, {
components: {
Card: ({ props, children }) => (
<h3>{props.title}</h3>
{children}
),
Metric: ({ props }) => (
{props.label}
{format(props.value, props.format)}
),
Button: ({ props, emit }) => (
emit("press")}>{props.label}
),
},
});3. Render AI-Generated Specs
function Dashboard({ spec }) {
return <Renderer spec={spec} registry={registry} />;
}That's it. AI generates JSON, you render it safely.
Packages
| Package | Description |
|---|---|
@json-render/core |
Schemas, catalogs, AI prompts, dynamic props, SpecStream utilities |
@json-render/react |
React renderer, contexts, hooks |
@json-render/vue |
Vue 3 renderer, composables, providers |
@json-render/svelte |
Svelte 5 renderer with runes-based reactivity |
@json-render/solid |
SolidJS renderer with fine-grained reactive contexts |
@json-render/shadcn |
36 pre-built shadcn/ui components (Radix UI + Tailwind CSS) |
@json-render/shadcn-svelte |
36 pre-built shadcn-svelte components (Svelte 5 + Tailwind CSS) |
@json-render/react-three-fiber |
React Three Fiber renderer for 3D scenes (20 built-in components, including GaussianSplat) |
@json-render/react-native |
React Native renderer with standard mobile components |
@json-render/next |
Next.js renderer — JSON becomes full apps with routes, layouts, SSR |
@json-render/remotion |
Remotion video renderer, timeline schema |
@json-render/react-pdf |
React PDF renderer for generating PDF documents from specs |
@json-render/react-email |
React Email renderer for HTML/plain-text emails from specs |
@json-render/ink |
Ink terminal renderer with built-in components for interactive TUIs. |
@json-render/image |
Image renderer for SVG/PNG output (OG images, social cards) via Satori |
@json-render/directives |
Pre-built custom directives — $format, $math, $concat, $count, $truncate, $pluralize, $join, $t (i18n) |
@json-render/codegen |
Utilities for generating code from json-render UI trees |
@json-render/devtools |
Framework-agnostic devtools core — panel UI, event store, picker, stream taps |
@json-render/devtools-react |
React adapter for @json-render/devtools (drop-in <JsonRenderDevtools />) |
@json-render/devtools-vue |
Vue adapter for @json-render/devtools |
@json-render/devtools-svelte |
Svelte adapter for @json-render/devtools |
@json-render/devtools-solid |
SolidJS adapter for @json-render/devtools |
@json-render/redux |
Redux / Redux Toolkit adapter for StateStore |
@json-render/zustand |
Zustand adapter for StateStore |
@json-render/jotai |
Jotai adapter for StateStore |
@json-render/xstate |
XState Store (atom) adapter for StateStore |
@json-render/mcp |
MCP Apps integration for Claude, ChatGPT, Cursor, VS Code |
@json-render/yaml |
YAML wire format with streaming parser, edit modes, AI SDK transform |
Renderers
React (UI)
import { defineRegistry, Renderer } from "@json-render/react";
import { schema } from "@json-render/react/schema";
// Flat spec format (root key + elements map)
const spec = {
root: "card-1",
elements: {
"card-1": {
type: "Card",
props: { title: "Hello" },
children: ["button-1"],
},
"button-1": {
type: "Button",
props: { label: "Click me" },
children: [],
},
},
};
// defineRegistry creates a type-safe component registry
const { registry } = defineRegistry(catalog, { components });
<Renderer spec={spec} registry={registry} />;Vue (UI)
import { h } from "vue";
import { defineRegistry, Renderer } from "@json-render/vue";
import { schema } from "@json-render/vue/schema";
const { registry } = defineRegistry(catalog, {
components: {
Card: ({ props, children }) =>
h("div", { class: "card" }, [h("h3", null, props.title), children]),
Button: ({ props, emit }) =>
h("button", { onClick: () => emit("press") }, props.label),
},
});
// In your Vue component template:
// <Renderer :spec="spec" :registry="registry" />Svelte (UI)
import { defineRegistry, Renderer } from "@json-render/svelte";
import { schema } from "@json-render/svelte/schema";
const { registry } = defineRegistry(catalog, {
components: {
Card: ({ props, children }) => /* Svelte 5 snippet */,
Button: ({ props, emit }) => /* Svelte 5 snippet */,
},
});
// In your Svelte component:
// <Renderer spec={spec} registry={registry} />Solid (UI)
import { defineRegistry, Renderer } from "@json-render/solid";
import { schema } from "@json-render/solid/schema";
const { registry } = defineRegistry(catalog, {
components: {
Card: (renderProps) => {renderProps.children},
Button: (renderProps) => (
renderProps.emit("press")}>
{renderProps.element.props.label as string}
),
},
});
<Renderer spec={spec} registry={registry} />;shadcn/ui (Web)
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react/schema";
import { defineRegistry, Renderer } from "@json-render/react";
import { shadcnComponentDefinitions } from "@json-render/shadcn/catalog";
import { shadcnComponents } from "@json-render/shadcn";
// Pick components from the 36 standard definitions
const catalog = defineCatalog(schema, {
components: {
Card: shadcnComponentDefinitions.Card,
Stack: shadcnComponentDefinitions.Stack,
Heading: shadcnComponentDefinitions.Heading,
Button: shadcnComponentDefinitions.Button,
},
actions: {},
});
// Use matching implementations
const { registry } = defineRegistry(catalog, {
components: {
Card: shadcnComponents.Card,
Stack: shadcnComponents.Stack,
Heading: shadcnComponents.Heading,
Button: shadcnComponents.Button,
},
});
<Renderer spec={spec} registry={registry} />;React Native (Mobile)
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react-native/schema";
import {
standardComponentDefinitions,
standardActionDefinitions,
} from "@json-render/react-native/catalog";
import { defineRegistry, Renderer } from "@json-render/react-native";
// 25+ standard components included
const catalog = defineCatalog(schema, {
components: { ...standardComponentDefinitions },
actions: standardActionDefinitions,
});
const { registry } = defineRegistry(catalog, { components: {} });
<Renderer spec={spec} registry={registry} />;Remotion (Video)
import { Player } from "@remotion/player";
import {
Renderer,
schema,
standardComponentDefinitions,
} from "@json-render/remotion";
// Timeline spec format
const spec = {
composition: {
id: "video",
fps: 30,
width: 1920,
height: 1080,
durationInFrames: 300,
},
tracks: [{ id: "main", name: "Main", type: "video", enabled: true }],
clips: [
{
id: "clip-1",
trackId: "main",
component: "TitleCard",
props: { title: "Hello" },
from: 0,
durationInFrames: 90,
},
],
audio: { tracks: [] },
};
;React PDF (Documents)
import { renderToBuffer } from "@json-render/react-pdf";
const spec = {
root: "doc",
elements: {
doc: {
type: "Document",
props: { title: "Invoice" },
children: ["page-1"],
},
"page-1": {
type: "Page",
props: { size: "A4" },
children: ["heading-1", "table-1"],
},
"heading-1": {
type: "Heading",
props: { text: "Invoice #1234", level: "h1" },
children: [],
},
"table-1": {
type: "Table",
props: {
columns: [
{ header: "Item", width: "60%" },
{ header: "Price", width: "40%", align: "right" },
],
rows: [
["Widget A", "$10.00"],
["Widget B", "$25.00"],
],
},
children: [],
},
},
};
// Render to buffer, stream, or file
const buffer = await renderToBuffer(spec);React Email (Email)
import { renderToHtml } from "@json-render/react-email";
import { schema, standardComponentDefinitions } from "@json-render/react-email";
import { defineCatalog } from "@json-render/core";
const catalog = defineCatalog(schema, {
components: standardComponentDefinitions,
});
const spec = {
root: "html-1",
elements: {
"html-1": {
type: "Html",
props: { lang: "en", dir: "ltr" },
children: ["head-1", "body-1"],
},
"head-1": { type: "Head", props: {}, children: [] },
"body-1": {
type: "Body",
props: { style: { backgroundColor: "#f6f9fc" } },
children: ["container-1"],
},
"container-1": {
type: "Container",
props: {
style: { maxWidth: "600px", margin: "0 auto", padding: "20px" },
},
children: ["heading-1", "text-1"],
},
"heading-1": { type: "Heading", props: { text: "Welcome" }, children: [] },
"text-1": {
type: "Text",
props: { text: "Thanks for signing up." },
children: [],
},
},
};
const html = await renderToHtml(spec);Image (SVG/PNG)
import { renderToPng } from "@json-render/image/render";
const spec = {
root: "frame",
elements: {
frame: {
type: "Frame",
props: { width: 1200, height: 630, backgroundColor: "#1a1a2e" },
children: ["heading"],
},
heading: {
type: "Heading",
props: { text: "Hello World", level: "h1", color: "#ffffff" },
children: [],
},
},
};
// Render to PNG (requires @resvg/resvg-js)
const png = await renderToPng(spec, { fonts });
// Or render to SVG string
import { renderToSvg } from "@json-render/image/render";
const svg = await renderToSvg(spec, { fonts });Three.js (3D)
import { defineCatalog } from "@json-render/core";
import { schema, defineRegistry } from "@json-render/react";
import {
threeComponentDefinitions,
threeComponents,
ThreeCanvas,
} from "@json-render/react-three-fiber";
const catalog = defineCatalog(schema, {
components: {
Box: threeComponentDefinitions.Box,
Sphere: threeComponentDefinitions.Sphere,
AmbientLight: threeComponentDefinitions.AmbientLight,
DirectionalLight: threeComponentDefinitions.DirectionalLight,
GaussianSplat: threeComponentDefinitions.GaussianSplat,
OrbitControls: threeComponentDefinitions.OrbitControls,
},
actions: {},
});
const { registry } = defineRegistry(catalog, {
components: {
Box: threeComponents.Box,
Sphere: threeComponents.Sphere,
AmbientLight: threeComponents.AmbientLight,
DirectionalLight: threeComponents.DirectionalLight,
GaussianSplat: threeComponents.GaussianSplat,
OrbitControls: threeComponents.OrbitControls,
},
});
;Next.js (Full Apps)
import type { NextAppSpec } from "@json-render/next";
import { createNextApp } from "@json-render/next/server";
import { NextAppProvider } from "@json-render/next";
const spec: NextAppSpec = {
metadata: { title: { default: "My App", template: "%s | My App" } },
layouts: {
main: {
root: "shell",
elements: {
shell: { type: "Container", props: {}, children: ["nav", "slot"] },
nav: { type: "NavBar", props: {}, children: [] },
slot: { type: "Slot", props: {}, children: [] },
},
},
},
routes: {
"/": {
layout: "main",
metadata: { title: "Home" },
page: {
root: "hero",
elements: {
hero: { type: "Card", props: { title: "Welcome" }, children: [] },
},
},
},
},
};
// Server: creates Page, generateMetadata, generateStaticParams
const app = createNextApp({ spec });
// Client: wrap your layout with NextAppProvider
// <NextAppProvider registry={registry} handlers={handlers}>
// {children}
// </NextAppProvider>shadcn-svelte (Svelte)
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/svelte/schema";
import { defineRegistry, Renderer } from "@json-render/svelte";
import { shadcnComponentDefinitions } from "@json-render/shadcn-svelte/catalog";
import { shadcnComponents } from "@json-render/shadcn-svelte";
const catalog = defineCatalog(schema, {
components: {
Card: shadcnComponentDefinitions.Card,
Stack: shadcnComponentDefinitions.Stack,
Heading: shadcnComponentDefinitions.Heading,
Button: shadcnComponentDefinitions.Button,
},
actions: {},
});
const { registry } = defineRegistry(catalog, {
components: {
Card: shadcnComponents.Card,
Stack: shadcnComponents.Stack,
Heading: shadcnComponents.Heading,
Button: shadcnComponents.Button,
},
});
// In your Svelte component:
// <Renderer spec={spec} registry={registry} />Devtools
Drop-in inspector panel for any json-render app. Spec tree, state editor, action log, stream log, catalog browser, DOM picker.
// React
import { JsonRenderDevtools } from "@json-render/devtools-react";
<JSONUIProvider registry={registry} handlers={handlers}>
<Renderer spec={spec} registry={registry} />
<JsonRenderDevtools spec={spec} catalog={catalog} messages={messages} />
</JSONUIProvider>;Floating toggle appears bottom-right. Hotkey: Ctrl/Cmd + Shift + J. Tree-shakes to null in production.
Available for React, Vue, Svelte, and Solid — swap @json-render/devtools-react for the adapter that matches your renderer.
Ink (Terminal)
import { defineCatalog } from "@json-render/core";
import {
schema,
standardComponentDefinitions,
standardActionDefinitions,
defineRegistry,
Renderer,
JSONUIProvider,
} from "@json-render/ink";
const catalog = defineCatalog(schema, {
components: { ...standardComponentDefinitions },
actions: standardActionDefinitions,
});
const { registry } = defineRegistry(catalog, { components: {} });
const spec = {
root: "card-1",
elements: {
"card-1": {
type: "Card",
props: { title: "Status" },
children: ["status-1"],
},
"status-1": {
type: "StatusLine",
props: { label: "Build", status: "success" },
children: [],
},
},
};
<JSONUIProvider initialState={{}}>
<Renderer spec={spec} registry={registry} />
</JSONUIProvider>;Features
Streaming (SpecStream)
Stream AI responses progressively:
import { createSpecStreamCompiler } from "@json-render/core";
const compiler = createSpecStreamCompiler<MySpec>();
// Process chunks as they arrive
const { result, newPatches } = compiler.push(chunk);
setSpec(result); // Update UI with partial result
// Get final result
const finalSpec = compiler.getResult();AI Prompt Generation
Generate system prompts from your catalog:
const systemPrompt = catalog.prompt();
// Includes component descriptions, props schemas, available actionsConditional Visibility
{
"type": "Alert",
"props": { "message": "Error occurred" },
"visible": [
{ "$state": "/form/hasError" },
{ "$state": "/form/errorDismissed", "not": true }
]
}Dynamic Props
Any prop value can be data-driven using expressions:
{
"type": "Icon",
"props": {
"name": {
"$cond": { "$state": "/activeTab", "eq": "home" },
"$then": "home",
"$else": "home-outline"
},
"color": {
"$cond": { "$state": "/activeTab", "eq": "home" },
"$then": "#007AFF",
"$else": "#8E8E93"
}
}
}Expression forms:
{ "$state": "/state/key" }- reads a value from the state model{ "$cond": <condition>, "$then": <value>, "$else": <value> }- evaluates a condition and picks a branch{ "$template": "Hello, ${/user/name}!" }- interpolates state values into strings{ "$computed": "fn", "args": { ... } }- calls a registered function with resolved args
Actions
Components can trigger actions, including the built-in setState action:
{
"type": "Pressable",
"props": {
"action": "setState",
"actionParams": { "statePath": "/activeTab", "value": "home" }
},
"children": ["home-icon"]
}The setState action updates the state model directly, which re-evaluates visibility conditions and dynamic prop expressions.
State Watchers
React to state changes by triggering actions:
{
"type": "Select",
"props": {
"value": { "$bindState": "/form/country" },
"options": ["US", "Canada", "UK"]
},
"watch": {
"/form/country": {
"action": "loadCities",
"params": { "country": { "$state": "/form/country" } }
}
}
}watch is a top-level field on elements (sibling of type/props/children). Watchers fire when the watched value changes, not on initial render.
Demo
git clone https://github.com/vercel-labs/json-render
cd json-render
pnpm install
pnpm dev- http://json-render.localhost:1355 - Docs & Playground
- http://dashboard-demo.json-render.localhost:1355 - Example Dashboard
- http://react-email-demo.json-render.localhost:1355 - React Email Example
- http://remotion-demo.json-render.localhost:1355 - Remotion Video Example
- Chat Example: run
pnpm devinexamples/chat - Svelte Example: run
pnpm devinexamples/svelteorexamples/svelte-chat - Vue Example: run
pnpm devinexamples/vue - Vite Renderers (React + Vue + Svelte + Solid): run
pnpm devinexamples/vite-renderers - React Native example: run
npx expo startinexamples/react-native - Gaussian Splatting (R3F): run
pnpm devinexamples/react-three-fiber-gsplat - Gaussian Splatting (experimental standalone gsplat.js demo): run
pnpm devinexamples/gsplat
How It Works
flowchart LR
A[User Prompt] --> B[AI + Catalog]
B --> C[JSON Spec]
C --> D[Renderer]
B -.- E([guardrailed])
C -.- F([predictable])
D -.- G([streamed])- Define the guardrails - what components, actions, and data bindings AI can use
- Prompt - describe what you want in natural language
- AI generates JSON - output is always predictable, constrained to your catalog
- Render fast - stream and render progressively as the model responds
License
Apache-2.0
File: apps/web/app/(main)/docs/installation/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/installation")
Installation
Install the core package plus your renderer of choice.
For React UI
Peer dependencies: react ^19.0.0 and zod ^4.0.0.
For Vue
Peer dependencies: vue ^3.5.0 and zod ^4.0.0.
For Svelte
Peer dependencies: svelte ^5.0.0 and zod ^4.0.0.
For React UI with shadcn/ui
Pre-built components for fast prototyping and production use:
Requires Tailwind CSS in your project. See the @json-render/shadcn API reference for usage.
For React Native
For Remotion Video
For React Email
For External State Management (Optional)
If you want to wire json-render to an existing state management library instead of the built-in store, install the adapter for your library:
See the Data Binding guide for usage.
For AI Integration
To use json-render with AI models, you'll also need the Vercel AI SDK:
File: apps/web/app/(main)/docs/generation-modes/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/generation-modes")
Generation Modes
json-render supports two modes for AI-generated UI: Standalone mode for standalone UI and Inline mode for inline UI within a conversation.
The mode controls how the AI formats its output and how your app processes the stream. The underlying JSONL patch format is the same in both modes.
Standalone Mode
In standalone mode, the AI outputs only JSONL patches — no prose, no markdown. The entire response is a UI spec.
This is the default mode and is ideal for:
- Playground and builder tools
- Form generators
- Dashboard builders
- Any UI where the generated interface is the whole response
Setup
import { streamText } from "ai";
// Standalone mode is the default (no mode option needed)
const systemPrompt = catalog.prompt({
customRules: [
"Use Card as root for forms and small UIs.",
"Use Grid for multi-column layouts.",
],
});
const result = streamText({
model: "anthropic/claude-haiku-4.5",
system: systemPrompt,
prompt: userPrompt,
});Client
On the client, use useUIStream from @json-render/react or the lower-level createSpecStreamCompiler from @json-render/core to compile the JSONL stream into a spec:
import { useUIStream } from "@json-render/react";
function Playground() {
const { spec, isStreaming, send } = useUIStream({
api: "/api/generate",
});
return (
<Renderer
spec={spec}
registry={registry}
loading={isStreaming}
/>
);
}Example output
The AI outputs only JSONL — one patch per line, no surrounding text:
{"op":"add","path":"/root","value":"card-1"}
{"op":"add","path":"/elements/card-1","value":{"type":"Card","props":{"title":"Sign In"},"children":["email","password","submit"]}}
{"op":"add","path":"/elements/email","value":{"type":"Input","props":{"label":"Email","name":"email","type":"email"}}}
{"op":"add","path":"/elements/password","value":{"type":"Input","props":{"label":"Password","name":"password","type":"password"}}}
{"op":"add","path":"/elements/submit","value":{"type":"Button","props":{"label":"Sign In"}}}Inline Mode
In inline mode, the AI responds conversationally first, then outputs JSONL patches on their own lines. Text-only replies are allowed when no UI is needed (e.g. greetings, clarifying questions).
This is ideal for:
- AI chatbots with rich UI responses
- Copilot experiences
- Educational assistants
- Any conversational interface where generated UI is embedded in chat messages
Setup
import { streamText } from "ai";
import { pipeJsonRender } from "@json-render/core";
import { createUIMessageStream, createUIMessageStreamResponse } from "ai";
// Enable inline mode
const systemPrompt = catalog.prompt({ mode: "inline" });
const result = streamText({
model: yourModel,
system: systemPrompt,
messages,
});
// In your API route, pipe the stream through pipeJsonRender
// to separate text from JSONL patches
const stream = createUIMessageStream({
execute: async ({ writer }) => {
writer.merge(pipeJsonRender(result.toUIMessageStream()));
},
});
return createUIMessageStreamResponse({ stream });pipeJsonRender inspects each line of the AI's response. Lines that parse as JSONL patches are emitted as data-spec parts (which the renderer picks up). Everything else is passed through as text.
Client
On the client, use useJsonRenderMessage from @json-render/react to extract the spec from a chat message's parts:
import { useChat } from "@ai-sdk/react";
import { useJsonRenderMessage } from "@json-render/react";
function Chat() {
const { messages, input, handleInputChange, handleSubmit } = useChat();
return (
{messages.map((msg) => (
<ChatMessage key={msg.id} message={msg} />
))}
{/* input form */}
);
}
function ChatMessage({ message }) {
const { spec } = useJsonRenderMessage(message.parts);
return (
{/* Render text parts */}
{message.parts
.filter((p) => p.type === "text")
.map((p, i) => {p.text})}
{/* Render the generated UI inline */}
{spec && (
<Renderer
spec={spec}
registry={registry}
/>
)}
);
}Example output
The AI writes a brief explanation, then JSONL patches on their own lines:
Here's a dashboard showing the latest crypto prices:
{"op":"add","path":"/root","value":"dashboard"}
{"op":"add","path":"/state/prices","value":[{"name":"Bitcoin","price":98450},{"name":"Ethereum","price":3120}]}
{"op":"add","path":"/elements/dashboard","value":{"type":"Grid","props":{"columns":"2"},"children":["btc","eth"]}}
{"op":"add","path":"/elements/btc","value":{"type":"Metric","props":{"label":"Bitcoin","value":{"$state":"/prices/0/price"}}}}
{"op":"add","path":"/elements/eth","value":{"type":"Metric","props":{"label":"Ethereum","value":{"$state":"/prices/1/price"}}}}If the user asks a simple question ("what does BTC stand for?"), the AI replies with text only — no JSONL.
Quick Comparison
| Standalone | Inline | |
|---|---|---|
| Output format | JSONL only | Text + JSONL |
| Text-only replies | No | Yes |
| System prompt | {"catalog.prompt()"} |
{'catalog.prompt({ mode: "inline" })'} |
| Stream utility | {"useUIStream"} |
{"pipeJsonRender"}{" + "}{"useJsonRenderMessage"} |
| Typical use case | Playground, builders | Chatbots, copilots |
Both modes use the same JSONL patch format (RFC 6902) and the same catalog/registry system. The only difference is whether the AI is allowed to include prose alongside the patches.
Next
- Learn about the JSONL streaming format
- See the AI SDK integration for setup with the Vercel AI SDK
File: apps/web/app/(main)/docs/directives/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/directives")
Directives
Extend the spec language with custom $-prefixed dynamic values. Directives let you add formatting, math, string manipulation, i18n, and any other transformation without modifying core.
Overview
A directive is a user-defined dynamic value expression, like $state or $computed, but defined in userland. Each directive has a $-prefixed name, a Zod schema for validation, and a resolver function.
{
"type": "Text",
"props": {
"text": {
"$format": "currency",
"value": { "$state": "/cart/total" },
"currency": "USD"
}
},
"children": []
}Defining a Directive
Use defineDirective from @json-render/core:
import { defineDirective, resolvePropValue } from '@json-render/core';
import { z } from 'zod';
const doubleDirective = defineDirective({
name: '$double',
schema: z.object({
$double: z.unknown(),
}),
resolve(value, ctx) {
const resolved = resolvePropValue(value.$double, ctx);
return (resolved as number) * 2;
},
});The description field is optional. When generating prompts, the directive's schema fields are auto-described from the Zod schema; the description adds short behavioral context the schema can't express.
Wiring Directives
Pass directives to both the renderer (for runtime resolution) and the catalog prompt (for AI generation).
Runtime
import { JSONUIProvider, Renderer } from '@json-render/react';
import { standardDirectives } from '@json-render/directives';
<JSONUIProvider registry={registry} directives={standardDirectives}>
<Renderer spec={spec} registry={registry} />
</JSONUIProvider>Or with createRenderer:
const MyRenderer = createRenderer(catalog, components);
<MyRenderer spec={spec} directives={directives} />All four renderers (React, Vue, Svelte, Solid) accept the directives prop on their provider and createRenderer output.
Prompt Generation
const prompt = catalog.prompt({ directives });Each directive's schema is auto-described in the "CUSTOM DYNAMIC VALUES" section of the system prompt. The optional description field adds behavioral context inline.
Pre-built Directives
The @json-render/directives package ships ready-to-use directives:
import { standardDirectives, createI18nDirective } from '@json-render/directives';standardDirectives includes $format, $math, $concat, $count, $truncate, $pluralize, and $join. Add factory directives by spreading:
const directives = [...standardDirectives, createI18nDirective(config)];See the API reference for details on each directive.
Composition
Directives compose naturally. Each resolver calls resolvePropValue on its inputs, so directives can wrap other directives or built-in expressions like $state:
{
"$format": "currency",
"value": {
"$math": "multiply",
"a": { "$state": "/price" },
"b": { "$state": "/qty" }
},
"currency": "USD"
}This resolves inside-out: $state reads from state, $math multiplies the values, and $format formats the result as currency.
Built-in Precedence
Built-in expressions ($state, $computed, $cond, $template, etc.) always take precedence over custom directives. defineDirective throws if you try to register a name that conflicts with a built-in key.
Next
- API Reference — full directive reference
- Computed Values —
$computedand$templateexpressions - Data Binding —
$state,$item, and binding expressions
File: apps/web/app/(main)/docs/devtools/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/devtools")
Devtools
A drop-in inspector panel for any json-render app. See the spec tree, edit state inline, watch dispatched actions, follow stream patches live, browse your catalog, and pick DOM elements to map them back to spec keys.
Production-safe: the component tree-shakes to a null render when NODE_ENV === "production".
Install
Pick the adapter that matches your renderer.
React
npm install @json-render/devtools @json-render/devtools-reactVue
npm install @json-render/devtools @json-render/devtools-vueSvelte
npm install @json-render/devtools @json-render/devtools-svelteSolid
npm install @json-render/devtools @json-render/devtools-solidQuick Start
Drop <JsonRenderDevtools /> anywhere inside your existing <JSONUIProvider> (or the equivalent provider tree).
// React
import { JsonRenderDevtools } from "@json-render/devtools-react";
<JSONUIProvider registry={registry} handlers={handlers}>
<Renderer spec={spec} registry={registry} />
<JsonRenderDevtools spec={spec} catalog={catalog} />
</JSONUIProvider>That's it. A floating toggle appears in the bottom-right corner. Click it, or press Ctrl/Cmd + Shift + J, to open the drawer.
Chat apps (AI SDK)
When you're using @ai-sdk/react's useChat, pass the messages prop so the Stream tab captures spec patches as they arrive:
<JsonRenderDevtools
spec={spec}
catalog={catalog}
messages={messages}
/>Panels
| Tab | What it shows |
|---|---|
| Spec | Element tree rooted at spec.root. Expand to walk children. Selecting an element fills a detail pane with its full props, visibility condition, event bindings, watchers, and any issues reported by validateSpec. |
| State | Every leaf path in the state model listed via flattenToPointers. Click a value to edit inline — writes go through store.set, so conditional elements and computed props re-evaluate immediately. |
| Actions | Timeline of dispatched actions: name, params, result or error, duration. Newest first. Expand a row for the full JSON payload. |
| Stream | Patches, text chunks, token usage, and lifecycle markers from the AI generation stream. Grouped by generation. |
| Catalog | Components and actions declared in your catalog with prop chips and type hints. |
| Pick | Click any element in the page to surface its entry in the Spec tab. Works because the renderer transparently tags each element with data-jr-key while devtools is mounted. |
Props
| Prop | Type | Default | Description |
|---|---|---|---|
spec |
Spec null |
null |
The spec currently being rendered. |
catalog |
Catalog null |
null |
Catalog definition — required for the Catalog panel. |
messages |
UIMessage[] |
undefined |
AI SDK useChat messages. Scanned for spec data parts and streamed into the Stream panel. |
initialOpen |
boolean |
false |
Start the drawer open. |
position |
"bottom-right" "bottom-left" "right" |
"bottom-right" |
Floating toggle button position. |
hotkey |
string false |
"mod+shift+j" |
Keyboard shortcut. Use mod for Cmd on macOS / Ctrl elsewhere. Pass false to disable. |
bufferSize |
number |
500 |
Max events retained in the ring buffer. |
onEvent |
(evt: DevtoolsEvent) => void |
undefined |
Optional tap — fires for every event as it is recorded. Useful for forwarding to analytics. |
Production Safety
The component renders null when process.env.NODE_ENV === "production". Bundlers fold the constant check so the panel's code tree-shakes out of production builds.
If you want extra certainty, gate the import behind an env check:
import dynamic from "next/dynamic";
const JsonRenderDevtools = dynamic(
() =>
import("@json-render/devtools-react").then((m) => ({
default: m.JsonRenderDevtools,
})),
{ ssr: false, loading: () => null },
);Advanced
Imperative controls
Use useJsonRenderDevtools() (React adapter only) to open / close the panel or record custom events from anywhere in the app:
import { useJsonRenderDevtools } from "@json-render/devtools-react";
function DebugButton() {
const devtools = useJsonRenderDevtools();
return (
devtools?.toggle()}>Toggle devtools
);
}Server-side stream tap
Capture stream events before they reach the client. Useful for server logs:
import { tapJsonRenderStream } from "@json-render/devtools";
const tapped = tapJsonRenderStream(
result.toUIMessageStream(),
serverEventStore,
);
writer.merge(pipeJsonRender(tapped));The @json-render/devtools core package exports tapJsonRenderStream and tapYamlStream for this pattern.
File: apps/web/app/(main)/docs/data-binding/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/data-binding")
Data Binding
Connect UI elements to dynamic data using expressions in your JSON specs.
State Model
Every spec can include a state object that holds the data your UI reads from:
{
"root": "greeting",
"elements": {
"greeting": {
"type": "Text",
"props": { "content": { "$state": "/user/name" } },
"children": []
}
},
"state": {
"user": { "name": "Alice" }
}
}State can also be provided programmatically at runtime. In @json-render/react, this is done via StateProvider and hooks like useStateStore. See the React API reference for details.
JSON Pointer Paths
All paths in json-render follow JSON Pointer (RFC 6901). A path is a string of /-separated tokens starting from the root:
Given this state:
{
"user": { "name": "Alice", "email": "[email protected]" },
"todos": [
{ "title": "Buy milk", "done": false },
{ "title": "Walk dog", "done": true }
]
}
"/user/name" -> "Alice"
"/user/email" -> "[email protected]"
"/todos/0/title" -> "Buy milk"
"/todos/1/done" -> trueExpressions
Expressions are special objects you place in props to read dynamic values instead of hardcoding them. There are six expression types.
`$state` — Read from state
Use { "$state": "/path" } in any prop to read a value from the state model:
{
"type": "Card",
"props": {
"title": { "$state": "/user/name" },
"subtitle": { "$state": "/user/email" }
},
"children": []
}If state contains { "user": { "name": "Alice", "email": "[email protected]" } }, the Card renders with title "Alice" and subtitle "[email protected]".
`$item` — Read from the current repeat item
Use { "$item": "field" } inside a repeat to read a field from the current array item:
{
"type": "Text",
"props": { "content": { "$item": "title" } },
"children": []
}Use { "$item": "" } to get the entire item object.
`$index` — Current repeat index
Use { "$index": true } inside a repeat to get the current array index (zero-based number):
{
"type": "Text",
"props": { "content": { "$index": true } },
"children": []
}Repeat
The repeat field on an element renders its children once per item in a state array. It is a top-level field on the element, sibling of type, props, and children — not inside props.
{
"root": "todo-list",
"elements": {
"todo-list": {
"type": "Column",
"props": { "gap": 8 },
"repeat": { "statePath": "/todos", "key": "id" },
"children": ["todo-item"]
},
"todo-item": {
"type": "Card",
"props": {
"title": { "$item": "title" },
"subtitle": { "$item": "description" }
},
"children": []
}
},
"state": {
"todos": [
{ "id": "1", "title": "Buy milk", "description": "2% or whole" },
{ "id": "2", "title": "Walk dog", "description": "Around the park" }
]
}
}repeat.statePath: root JSON Pointer to the state array, or{ "$item": "field" }for an array on the enclosing repeat itemrepeat.key— field name on each item to use as a stable key for rendering
Inside todo-item, { "$item": "title" } reads the title field from whichever array item is currently being rendered. { "$index": true } would return 0 for the first item, 1 for the second, and so on.
Two-Way Binding with `$bindState`
Form components use { "$bindState": "/path" } on their natural value prop for two-way binding. The component reads from and writes to the state path.
Value prop (text inputs)
{
"type": "TextInput",
"props": {
"value": { "$bindState": "/form/email" },
"placeholder": "Enter your email"
},
"children": []
}Checked prop (switches, checkboxes)
{
"type": "Switch",
"props": {
"label": "Enable notifications",
"checked": { "$bindState": "/settings/notifications" }
},
"children": []
}Pressed prop (toggle buttons)
{
"type": "ToggleButton",
"props": {
"label": "Bold",
"pressed": { "$bindState": "/editor/bold" }
},
"children": []
}Two-Way Binding with `$bindItem`
Inside a repeat scope, use { "$bindItem": "field" } to bind to a field on the current item:
{
"type": "Switch",
"props": {
"label": "Done",
"checked": { "$bindItem": "completed" }
},
"children": []
}Use { "$bindItem": "" } to bind to the entire item.
statePath is not used for component binding. It remains for repeat.statePath (array iteration path) and action params like setState.statePath (target path for mutations).
Conditional Props
Use $cond / $then / $else to pick a prop value based on a condition:
{
"type": "Badge",
"props": {
"label": {
"$cond": { "$state": "/user/isAdmin" },
"$then": "Admin",
"$else": "Member"
}
},
"children": []
}The condition uses the same visibility expression format.
Template Strings
Use { "$template": "..." } to interpolate state values into a string using ${/path} syntax:
{
"type": "Text",
"props": {
"text": { "$template": "Welcome back, ${/user/name}!" }
},
"children": []
}See Computed Values for details on $template and $computed expressions.
Quick Reference
| Expression | Syntax | Context |
|---|---|---|
{"$state"} |
{'{ "$state": "/path" }'} |
Anywhere |
{"$item"} |
{'{ "$item": "field" }'} |
Inside repeat only |
{"$index"} |
{'{ "$index": true }'} |
Inside repeat only |
{"$cond"} |
{'{ "$cond": ..., "$then": ..., "$else": ... }'} |
Anywhere |
{"$bindState"} |
{'{ "$bindState": "/path" }'} |
Form components (value, checked, pressed) |
{"$bindItem"} |
{'{ "$bindItem": "field" }'} |
Form components inside repeat |
{"$template"} |
{'{ "$template": "Hello, ${/name}!" }'} |
Anywhere (string props) |
{"$computed"} |
{'{ "$computed": "fn", "args": { ... } }'} |
Anywhere (requires registered function) |
External Store (Controlled Mode)
For advanced use cases, you can pass a StateStore to StateProvider to use your own state management (Redux, Zustand, XState, etc.) instead of the built-in internal store:
import { createStateStore, type StateStore } from "@json-render/react";
const store = createStateStore({ user: { name: "Alice" } });
<StateProvider store={store}>
{children}
</StateProvider>
// Mutate from anywhere — React re-renders automatically:
store.set("/user/name", "Bob");When store is provided, initialState and onStateChange are ignored. The store is the single source of truth. See the React API reference for the full StateStore interface.
Next
- Visibility — conditionally show or hide elements
- Action handlers — respond to user interactions
- React API reference — React-specific hooks for programmatic state access
File: apps/web/app/(main)/docs/custom-schema/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/custom-schema")
Custom Schema & Renderer
Build your own schema and renderer with @json-render/core.
Overview
@json-render/core is schema-agnostic. While @json-render/react provides a ready-to-use schema and renderer, you can create your own to match any JSON structure - whether it's a domain-specific format, an existing protocol, or something entirely custom.
1. Define Your Schema
Start by defining the JSON structure your system will use. Here's an example of a simple dashboard schema:
{
"layout": "grid",
"columns": 2,
"widgets": [
{
"type": "metric",
"title": "Revenue",
"value": "$12,345",
"trend": "up"
},
{
"type": "chart",
"title": "Sales",
"chartType": "line",
"dataKey": "salesData"
},
{
"type": "table",
"title": "Recent Orders",
"columns": ["id", "customer", "amount"],
"dataKey": "orders"
}
]
}2. Create the Catalog
Define a catalog that describes your components and validates props using defineCatalog — see Catalog.
import { defineCatalog } from '@json-render/core';
import { z } from 'zod';
export const dashboardCatalog = defineCatalog(mySchema, {
components: {
metric: {
props: z.object({
title: z.string(),
value: z.string(),
trend: z.enum(['up', 'down', 'flat']).optional(),
change: z.string().optional(),
}),
},
chart: {
props: z.object({
title: z.string(),
chartType: z.enum(['line', 'bar', 'pie', 'area']),
dataKey: z.string(),
height: z.number().optional(),
}),
},
table: {
props: z.object({
title: z.string(),
columns: z.array(z.string()),
dataKey: z.string(),
pageSize: z.number().optional(),
}),
},
text: {
props: z.object({
content: z.string(),
variant: z.enum(['heading', 'body', 'caption']).optional(),
}),
},
},
});3. Define the Root Schema
Create a schema for the overall document structure:
import { z } from 'zod';
const WidgetSchema = z.object({
type: z.string(),
title: z.string().optional(),
// Additional props validated by catalog
}).passthrough();
export const DashboardSchema = z.object({
layout: z.enum(['grid', 'stack', 'tabs']),
columns: z.number().optional(),
widgets: z.array(WidgetSchema),
});
export type Dashboard = z.infer<typeof DashboardSchema>;
export type Widget = z.infer<typeof WidgetSchema>;4. Build the Renderer
Create a renderer that maps your schema to React components:
import React from 'react';
import { dashboardCatalog } from './catalog';
import type { Dashboard, Widget } from './schema';
// Widget component registry
const widgetComponents: Record<string, React.FC<any>> = {
metric: ({ title, value, trend, change }) => (
{title}
{value}
{trend && (
{trend === 'up' ? '+' : '-'}{change}
)}
),
chart: ({ title, chartType, data }) => (
{title}
{/* Your chart library here */}
{chartType} chart
),
table: ({ title, columns, data }) => (
{title}
| {col} |
| --- |
| {row[col]} |
),
text: ({ content, variant = 'body' }) => {
const className = {
heading: 'text-xl font-bold',
body: 'text-base',
caption: 'text-sm text-muted-foreground',
}[variant];
return {content};
},
};
// Main renderer
export function DashboardRenderer({
spec,
data = {},
}: {
spec: Dashboard;
data?: Record<string, any>;
}) {
const layoutClass = {
grid: `grid gap-4 ${spec.columns ? `grid-cols-${spec.columns}` : 'grid-cols-2'}`,
stack: 'flex flex-col gap-4',
tabs: 'space-y-4',
}[spec.layout];
return (
{spec.widgets.map((widget, index) => {
const Component = widgetComponents[widget.type];
if (!Component) {
console.warn(`Unknown widget type: ${widget.type}`);
return null;
}
// Resolve data references
const widgetData = widget.dataKey ? data[widget.dataKey] : undefined;
return (
<Component
key={index}
{...widget}
data={widgetData}
/>
);
})}
);
}5. Generate LLM Prompts
Use the catalog to generate system prompts for AI:
const systemPrompt = dashboardCatalog.prompt({
customRules: [
'Use metric widgets for single KPI values',
'Use chart widgets for time-series data',
'Use table widgets for lists of records',
'Limit dashboards to 6 widgets maximum',
],
});
// Use with any LLM
const response = await generateText({
model: 'gpt-4',
system: systemPrompt,
prompt: 'Create a sales dashboard with revenue, orders, and a chart',
});6. Validate Specs
Validate incoming specs against your schema. Use catalog.validate() to check AI output against the catalog's Zod schema:
function validateDashboard(spec: unknown) {
// Validate root structure
const rootResult = DashboardSchema.safeParse(spec);
if (!rootResult.success) {
return { valid: false, errors: rootResult.error.errors };
}
// Validate each widget's props against the catalog
const result = dashboardCatalog.validate(spec);
if (!result.success) {
return { valid: false, errors: result.error.errors };
}
return { valid: true, errors: [] };
}Usage Example
'use client';
import { useState } from 'react';
import { DashboardRenderer } from './renderer';
import type { Dashboard } from './schema';
const initialSpec: Dashboard = {
layout: 'grid',
columns: 2,
widgets: [
{ type: 'metric', title: 'Revenue', value: '$12,345', trend: 'up' },
{ type: 'metric', title: 'Orders', value: '156', trend: 'up' },
{ type: 'chart', title: 'Sales Trend', chartType: 'line', dataKey: 'sales' },
{ type: 'table', title: 'Recent Orders', columns: ['id', 'customer', 'amount'], dataKey: 'orders' },
],
};
const data = {
sales: [/* chart data */],
orders: [
{ id: '001', customer: 'Acme Inc', amount: '$500' },
{ id: '002', customer: 'Globex', amount: '$750' },
],
};
export function MyDashboard() {
const [spec, setSpec] = useState(initialSpec);
return <DashboardRenderer spec={spec} data={data} />;
}Next
See how to integrate with A2UI or Adaptive Cards protocols.
File: apps/web/app/(main)/docs/computed-values/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/computed-values")
Computed Values
Derive dynamic prop values using registered functions or string templates.
`$template` — String Interpolation
Use { "$template": "..." } to embed state values into a string. References use ${/path} syntax where the path is a JSON Pointer:
{
"type": "Text",
"props": {
"text": { "$template": "Hello, ${/user/name}! You have ${/inbox/count} messages." }
},
"children": []
}If state is { "user": { "name": "Alice" }, "inbox": { "count": 3 } }, the text renders as "Hello, Alice! You have 3 messages."
Missing paths resolve to an empty string.
`$computed` — Registered Functions
Use { "$computed": "<name>", "args": { ... } } to call a named function registered in your catalog. Each arg can be a literal value or any prop expression ($state, $item, $cond, etc.):
{
"type": "Text",
"props": {
"text": {
"$computed": "fullName",
"args": {
"first": { "$state": "/form/firstName" },
"last": { "$state": "/form/lastName" }
}
}
},
"children": []
}Registering Functions
Functions are registered in the catalog and provided at runtime.
Catalog definition (for AI prompt generation):
import { defineCatalog } from '@json-render/core';
import { schema } from '@json-render/react/schema';
const catalog = defineCatalog(schema, {
components: { /* ... */ },
functions: {
fullName: {
},
formatCurrency: {
},
},
});Runtime implementation:
import { JSONUIProvider } from '@json-render/react';
const functions = {
fullName: (args) => `${args.first ?? ''} ${args.last ?? ''}`.trim(),
formatCurrency: (args) => {
const value = Number(args.value ?? 0);
return new Intl.NumberFormat('en-US', {
style: 'currency',
currency: (args.currency as string) ?? 'USD',
}).format(value);
},
};
<JSONUIProvider registry={registry} functions={functions}>
<Renderer spec={spec} registry={registry} />
</JSONUIProvider>Using with `createRenderer`
const MyRenderer = createRenderer(catalog, components);
<MyRenderer
spec={spec}
functions={functions}
/>Combining Expressions
$computed args can use any expression type. This example computes a total from repeat item fields:
{
"$computed": "lineTotal",
"args": {
"price": { "$item": "price" },
"quantity": { "$item": "quantity" }
}
}Next
- Watchers — react to state changes with cascading actions
- Data Binding — all expression types
- Validation — validate form inputs
File: apps/web/app/(main)/docs/code-export/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/code-export")
Code Export
Export generated UI as standalone code for your framework.
Overview
While json-render is designed for dynamic rendering, you can export generated UI as static code. The code generation is intentionally project-specific so you have full control over:
- Component templates (standalone, no json-render dependencies)
- Package.json and project structure
- Framework-specific patterns (Next.js, Remix, etc.)
- How data is passed to components
Architecture
Code export is split into two parts:
1. @json-render/codegen (utilities)
Framework-agnostic utilities for building code generators:
import {
traverseSpec, // Walk the UI spec
collectUsedComponents, // Get all component types used
collectStatePaths, // Get all data binding paths
collectActions, // Get all action names
serializeProps, // Convert props to JSX string
} from '@json-render/codegen';2. Your Project (generator)
Custom code generator specific to your project and framework:
// lib/codegen/generator.ts
import { collectUsedComponents, serializeProps } from '@json-render/codegen';
export function generateNextJSProject(spec: Spec): GeneratedFile[] {
const components = collectUsedComponents(spec);
return [
{ path: 'package.json', content: '...' },
{ path: 'app/page.tsx', content: '...' },
// ... component files
];
}Example: Next.js Export
See the dashboard example for a complete implementation that exports:
package.json- Dependencies and scriptstsconfig.json- TypeScript confignext.config.js- Next.js configapp/layout.tsx- Root layoutapp/globals.css- Global stylesapp/page.tsx- Generated page with datacomponents/ui/*.tsx- Standalone components
Standalone Components
The exported components are standalone with no json-render dependencies. They receive data as props instead of using hooks:
// Generated component (standalone)
interface MetricProps {
label: string;
statePath: string;
data?: Record<string, unknown>;
}
export function Metric({ label, statePath, data }: MetricProps) {
const value = data ? getByPath(data, statePath) : undefined;
return (
{label}
{formatValue(value)}
);
}Using the Utilities
traverseSpec
import { traverseSpec } from '@json-render/codegen';
traverseSpec(spec, (element, key, depth, parent) => {
console.log(' '.repeat(depth * 2) + `${key}: ${element.type}`);
});collectUsedComponents
import { collectUsedComponents } from '@json-render/codegen';
const components = collectUsedComponents(spec);
// Set { 'Card', 'Metric', 'Chart', 'Table' }
// Generate only the needed component files
for (const component of components) {
files.push({
path: `components/ui/${component.toLowerCase()}.tsx`,
content: componentTemplates[component],
});
}serializeProps
import { serializeProps } from '@json-render/codegen';
const propsStr = serializeProps({
title: 'Dashboard',
columns: 3,
disabled: true,
});
// 'title="Dashboard" columns={3} disabled'Try It
Run the dashboard example and click "Export Project" to see code generation in action:
cd examples/dashboard
pnpm dev
# Open http://dashboard-demo.json-render.localhost:1355
# Generate a widget, then click "Export Project"File: apps/web/app/(main)/docs/changelog/page.mdx
import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/changelog")
Changelog
Notable changes and updates to json-render.
v0.20.0
August 15, 2026
New: Named Slots for React
React components can now declare named slots such as header and footer, while children remains the default slot. Named slots are preserved through validation, streaming, nested conversion, code export, playground views, and Devtools navigation.
This work builds on the original named slots contribution by @wotnak.
const catalog = defineCatalog(schema, {
components: {
Layout: {
props: z.object({}),
slots: ["default", "header", "footer"],
},
},
});
const { registry } = defineRegistry(catalog, {
components: {
Layout: ({ children, slots }) => (
<section>
<header>{slots?.header}</header>
<main>{children}</main>
<footer>{slots?.footer}</footer>
</section>
),
},
});New: Nested Repeats
repeat.statePath now accepts item-relative paths such as { "$item": "employees" }. Nested data can be rendered across React, React Native, React Email, React PDF, Image, Ink, Solid, Svelte, and Vue.
This work builds on the original nested repeats contribution by @tmchow.
{
"type": "Table",
"repeat": { "statePath": { "$item": "employees" }, "key": "id" },
"children": ["employee-row"],
"props": {}
}New: Harness Chat Example
Added a complete Next.js example using the AI SDK 7 harness adapter, agent delegation, sandbox transport, and json-render components.
Fixed: Chained Action Params
Named onSuccess and onError actions now receive their configured params across core and all renderer bridges. Custom renderer bridges must now accept an ActionBinding in the core executeAction callback instead of a bare action name.
Improved: Spec Validation and Compatibility
Element visibility remains optional across Zod 4 versions. Autofix now prunes dangling child references, validation reports malformed visibility conditions, and repeated items can be filtered safely.
Improved: Release Toolchain
The workspace now requires Node.js 24 and pnpm 11, enforces package engine checks, and applies a minimum package release age.
v0.19.0
May 6, 2026
New: Custom Directives API
@json-render/core now supports custom directives via defineDirective, letting you declare new JSON shapes (like $format, $math) that resolve to computed values at render time. Directives compose naturally -- nest $format over $math over $state and they resolve inside-out. All four renderers (React, Vue, Svelte, Solid) have built-in directive resolution.
New: `@json-render/directives`
New package shipping seven ready-made directives: $format (date, currency, number, percent via Intl), $math (arithmetic and rounding), $concat, $count, $truncate, $pluralize, and $join. Also exports createI18nDirective for $t translation keys with interpolation, and standardDirectives for one-line registration.
npm install @json-render/directivesimport { standardDirectives } from "@json-render/directives";
const catalog = createCatalog({
directives: standardDirectives,
// ...
});See the Directives guide and the API reference for details.
v0.18.0
April 17, 2026
New: Devtools
Five new packages for inspecting json-render apps in the browser:
@json-render/devtools-- framework-agnostic core@json-render/devtools-react@json-render/devtools-vue@json-render/devtools-svelte@json-render/devtools-solid
Drop <JsonRenderDevtools /> into your app to get a shadow-DOM-isolated panel with six tabs (Spec, State, Actions, Stream, Catalog, Pick), a DOM picker that maps clicked elements back to spec keys via data-jr-key, and a capped event store. Toggle with the floating button or Cmd/Ctrl + Shift + J. Tree-shakes to null in production.
npm install @json-render/devtools-reactimport { JsonRenderDevtools } from "@json-render/devtools-react";
<JsonRenderDevtools />See the Devtools guide and the API reference for details.
New: Devtools Example
New examples/devtools Next.js demo showing the full devtools panel wired up to an AI chat endpoint and a component catalog.
New: Action Observer and Devtools Flag in Core
@json-render/core now exposes an action observer and a devtools enablement flag that framework adapters use to mirror actions and stream events into the panel.
Fixed: Zod 4 Schema Formatting
formatZodType now correctly handles z.record(), z.default(), and z.literal() types from Zod 4, which previously produced incorrect or empty output in generated prompts and schemas.
v0.17.0
April 10, 2026
New: Gaussian Splatting
Added GaussianSplat component to @json-render/react-three-fiber, bringing the component count to 20. Composable with all existing R3F components (lights, controls, post-processing) via drei's Splat loader.
New: R3F Gaussian Splatting Example
Demo app with five scenes: splat showroom, splat with primitives, multi-splat, post-processing effects, and animated floating splat.
New: Standalone gsplat Example
Experimental demo app showcasing Gaussian Splatting with gsplat.js (no Three.js dependency), featuring scene selector, live JSON spec viewer, and progress indicator.
Improved: AI Output Quality
Improved prompt output and schema generation for more reliable AI-generated specs.
v0.16.0
March 27, 2026
New: `@json-render/next`
Next.js renderer that turns JSON specs into full Next.js applications with routes, layouts, SSR, metadata, data loaders, and static generation. Client and server entry points at @json-render/next and @json-render/next/server. Includes built-in Link, Slot, error boundary, loading, and not-found components.
npm install @json-render/nextNew: `@json-render/shadcn-svelte`
Pre-built shadcn-svelte components for json-render Svelte apps. 36 components built on Svelte 5 + Tailwind CSS with state binding, validation, and action support. Server-safe catalog at @json-render/shadcn-svelte/catalog.
npm install @json-render/shadcn-svelteImproved: Release Process
Switched from Changesets to a manual single-PR release workflow with changelog markers and automatic npm publish on version bump.
v0.15.0
March 23, 2026
New: `@json-render/ink`
Terminal renderer for json-render. JSON becomes terminal UIs powered by Ink. Stream AI-generated specs directly to the terminal with useUIStream.
npm install @json-render/inkImproved: YAML Format Support in `buildUserPrompt`
buildUserPrompt now accepts format and serializer options, enabling YAML as a wire format alongside JSON.
v0.14.0
March 13, 2026
New: `@json-render/yaml`
YAML wire format for json-render. Includes streaming YAML parser, yamlPrompt() for system prompts, and AI SDK transform (pipeYamlRender) as a drop-in alternative to JSONL streaming. Supports four fence types: yaml-spec, yaml-edit, yaml-patch, and diff.
npm install @json-render/yamlNew: Universal Edit Modes
Three strategies for multi-turn spec refinement in @json-render/core:
- Patch -- RFC 6902 JSON Patch
- Merge -- RFC 7396 Merge Patch
- Diff -- Unified diff
New editModes option on buildUserPrompt() and PromptOptions. New helpers: deepMergeSpec(), diffToPatches(), buildEditUserPrompt(), buildEditInstructions(), isNonEmptySpec().
Improved: Playground
Format toggle (JSONL / YAML), edit mode picker (patch / merge / diff), and token usage display with prompt caching stats.
Improved: Prompt Caching
Generate API uses Anthropic ephemeral cache control for system prompts.
v0.13.0
March 12, 2026
New: `@json-render/solid`
SolidJS renderer for json-render. JSON becomes Solid components with reactive rendering, schema export, and full catalog support.
npm install @json-render/core @json-render/solidNew: `@json-render/react-three-fiber`
React Three Fiber renderer for json-render. JSON becomes 3D scenes with 19 built-in components for meshes, lights, models, environments, text, cameras, and controls.
npm install @json-render/react-three-fiberImproved: Strict JSON Schema Mode
catalog.jsonSchema({ strict: true }) produces a JSON Schema subset compatible with LLM structured output APIs (OpenAI, Google Gemini, Anthropic). Ensures additionalProperties: false on every object and all properties listed in required.
v0.12.1
March 11, 2026
Changed: Generation Mode Renames
Renamed generation modes from "generate" / "chat" to "standalone" / "inline". The old names still work but emit a deprecation warning.
Fixed: MCP React Duplicate Module Error
Resolved React duplicate module error (useRef returning null) in @json-render/mcp by adding resolve.dedupe Vite configuration. Added ./build-app-html export entry point.
v0.12.0
March 6, 2026
New: `@json-render/svelte`
Svelte 5 renderer with runes-based reactivity. Full support for data binding, visibility, actions, validation, watchers, streaming, and repeat scopes. Includes defineRegistry, Renderer, schema, composables, and context providers.
npm install @json-render/core @json-render/svelteNew: `@json-render/react-email`
React Email renderer for generating HTML and plain-text emails from JSON specs. 17 standard components (Html, Head, Body, Container, Section, Row, Column, Heading, Text, Link, Button, Image, Hr, Preview, Markdown). Server-side renderToHtml / renderToPlainText APIs.
npm install @json-render/react-emailNew: `@json-render/mcp`
MCP Apps integration that serves json-render UIs as interactive apps inside Claude, ChatGPT, Cursor, VS Code, and other MCP-capable clients. createMcpApp server factory, useJsonRenderApp React hook for iframes, and buildAppHtml utility.
npm install @json-render/mcpv0.11.0
February 27, 2026
New: `@json-render/image`
Server-side image renderer powered by Satori. Turns the same { root, elements } spec format into SVG or PNG output for OG images, social cards, and banners.
npm install @json-render/imageimport { renderToSvg, renderToPng } from "@json-render/image/render";
import { standardComponentDefinitions } from "@json-render/image/catalog";
const svg = await renderToSvg(spec, { width: 1200, height: 630 });
const png = await renderToPng(spec, { width: 1200, height: 630 });9 standard components: Frame, Box, Row, Column, Heading, Text, Image, Divider, Spacer. Server-safe import path at @json-render/image/server.
v0.10.0
February 25, 2026
New: `@json-render/vue`
Vue 3 renderer for json-render with full feature parity with @json-render/react. Data binding, visibility conditions, actions, validation, repeat scopes, streaming, and external store support.
npm install @json-render/core @json-render/vueimport { h } from "vue";
import { defineRegistry, Renderer } from "@json-render/vue";
import { schema } from "@json-render/vue/schema";
const { registry } = defineRegistry(catalog, {
components: {
Card: ({ props, children }) =>
h("div", { class: "card" }, [h("h3", null, props.title), children]),
Button: ({ props, emit }) =>
h("button", { onClick: () => emit("press") }, props.label),
},
});Providers: StateProvider, ActionProvider, VisibilityProvider, ValidationProvider. Composables: useStateStore, useStateValue, useActions, useAction, useIsVisible, useFieldValidation, useBoundProp, useUIStream, useChatUI.
See the Vue API reference for details.
New: `@json-render/xstate`
XState Store (atom) adapter for json-render's StateStore interface. Wire an @xstate/store atom as the state backend for any renderer.
npm install @json-render/xstate @xstate/storeimport { createAtom } from "@xstate/store";
import { xstateStoreStateStore } from "@json-render/xstate";
const atom = createAtom({ count: 0 });
const store = xstateStoreStateStore({ atom });Requires @xstate/store v3+.
New: `$computed` and `$template` Expressions
Two new prop expression types for dynamic values:
$template-- interpolate state values into strings:{ "$template": "Hello, ${/user/name}!" }$computed-- call registered functions:{ "$computed": "fullName", "args": { "first": { "$state": "/form/firstName" } } }
Register functions via the functions prop on JSONUIProvider or createRenderer. See Computed Values for details.
New: State Watchers
Elements can declare a watch field to trigger actions when state values change. Useful for cascading dependencies like country/city selects.
{
"type": "Select",
"props": { "value": { "$bindState": "/form/country" }, "options": ["US", "Canada"] },
"watch": {
"/form/country": { "action": "loadCities", "params": { "country": { "$state": "/form/country" } } }
}
}watch is a top-level field on elements (sibling of type/props/children), not inside props. Watchers only fire on value changes, not on initial render. See Watchers for details.
New: Cross-Field Validation
New built-in validation functions for cross-field comparisons:
equalTo-- alias formatcheswith clearer semanticslessThan-- value must be less than another fieldgreaterThan-- value must be greater than another fieldrequiredIf-- required only when a condition field is truthy
Validation check args now resolve through resolvePropValue, so $state expressions work consistently.
New: `validateForm` Action
Built-in action (React) that validates all registered form fields at once and writes { valid, errors } to state:
{
"on": {
"press": [
{ "action": "validateForm", "params": { "statePath": "/formResult" } },
{ "action": "submitForm" }
]
}
}Improved: shadcn/ui Validation
All form components now support checks and validateOn props:
- Checkbox, Radio, Switch added validation support
validateOncontrols timing:"change"(default for Select, Checkbox, Radio, Switch),"blur"(default for Input, Textarea), or"submit"
New Examples
- Vue example -- standalone Vue 3 app with custom components
- Vite Renderers -- side-by-side React and Vue renderers with shared catalog
v0.9.1
February 24, 2026
Fixed: Install failure due to private dependency
@json-render/react, @json-render/react-pdf, and @json-render/react-native v0.9.0 failed to install because @internal/react-state (a private workspace package) was published as a dependency. The internal package is now bundled into each renderer at build time, so it no longer needs to be resolved from npm.
v0.9.0
February 24, 2026
New: External State Store
The StateStore interface lets you plug in your own state management (Redux, Zustand, Jotai, XState, etc.) instead of the built-in internal store. Pass a store prop to StateProvider, JSONUIProvider, or createRenderer for controlled mode.
- Added
StateStoreinterface andcreateStateStore()factory to@json-render/core StateProvider,JSONUIProvider, andcreateRenderernow accept an optionalstoreprop- When
storeis provided, it becomes the single source of truth (initialState/onStateChangeare ignored) - When
storeis omitted, everything works exactly as before (fully backward compatible) - Applied across all platform packages: react, react-native, react-pdf
- Store utilities (
createStoreAdapter,immutableSetByPath,flattenToPointers) available via@json-render/core/store-utilsfor building custom adapters
New adapter packages: @json-render/redux, @json-render/zustand, @json-render/jotai.
See the Data Binding guide for usage.
Changed: `onStateChange` signature updated (breaking)
The onStateChange callback now receives a single array of changed entries instead of being called once per path. This makes batch updates via update() easier to handle:
// Before
onStateChange?: (path: string, value: unknown) => void
// After
onStateChange?: (changes: Array<{ path: string; value: unknown }>) => voidThe callback is only called when a set() or update() call actually changes the state. A set() call produces a single-element array; an update() call produces one array with all changed paths.
Fixed: Server-safe schema import
@json-render/react barrel-imports React contexts that call createContext, which crashes in Next.js App Router API routes (RSC runtime strips createContext). All docs, examples, and skills now import schema from @json-render/react/schema instead of @json-render/react.
For combined imports, split into separate schema (subpath) and client API (main entry) lines:
import { schema } from "@json-render/react/schema";
import { defineRegistry, Renderer } from "@json-render/react";Fixed: Chaining actions
Fixed an issue where chaining multiple actions on the same event (e.g. setState followed by a custom action) did not execute all actions. Affected @json-render/react, @json-render/react-native, and @json-render/react-pdf.
Fixed: Zod array inner type resolution
Fixed safely resolving the inner type for Zod arrays in schema introspection, preventing errors when catalog component props use z.array().
v0.8.0
February 20, 2026
New: `@json-render/react-pdf`
PDF renderer for json-render, powered by @react-pdf/renderer. Define catalogs and registries the same way as @json-render/react, but output PDF documents instead of web UI.
npm install @json-render/core @json-render/react-pdfimport { renderToBuffer } from "@json-render/react-pdf";
import type { Spec } from "@json-render/core";
const spec: Spec = {
root: "doc",
elements: {
doc: { type: "Document", props: { title: "Invoice" }, children: ["page"] },
page: {
type: "Page",
props: { size: "A4" },
children: ["heading", "table"],
},
heading: {
type: "Heading",
props: { text: "Invoice #1234", level: "h1" },
children: [],
},
table: {
type: "Table",
props: {
columns: [
{ header: "Item", width: "60%" },
{ header: "Price", width: "40%", align: "right" },
],
rows: [
["Widget A", "$10.00"],
["Widget B", "$25.00"],
],
},
children: [],
},
},
};
const buffer = await renderToBuffer(spec);Server-side rendering APIs:
renderToBuffer(spec)-- render to an in-memory PDF bufferrenderToStream(spec)-- render to a readable stream (pipe to HTTP response)renderToFile(spec, path)-- render directly to a file
15 standard components covering document structure (Document, Page), layout (View, Row, Column), content (Heading, Text, Image, Link), data (Table, List), decorative (Divider, Spacer), and page-level (PageNumber).
Supports custom catalogs with defineRegistry, server-safe imports via @json-render/react-pdf/server, and full context support (state, visibility, actions, validation, repeat scopes).
v0.7.0
February 17, 2026
New: `@json-render/shadcn`
Pre-built shadcn/ui component library for json-render. 36 components built on Radix UI + Tailwind CSS, ready to use with defineCatalog and defineRegistry.
npm install @json-render/shadcnimport { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react/schema";
import { shadcnComponentDefinitions } from "@json-render/shadcn/catalog";
import { defineRegistry } from "@json-render/react";
import { shadcnComponents } from "@json-render/shadcn";
const catalog = defineCatalog(schema, {
components: {
Card: shadcnComponentDefinitions.Card,
Button: shadcnComponentDefinitions.Button,
Input: shadcnComponentDefinitions.Input,
},
actions: {},
});
const { registry } = defineRegistry(catalog, {
components: {
Card: shadcnComponents.Card,
Button: shadcnComponents.Button,
Input: shadcnComponents.Input,
},
});Components include: layout (Card, Stack, Grid, Separator), navigation (Tabs, Accordion, Collapsible, Pagination), overlay (Dialog, Drawer, Tooltip, Popover, DropdownMenu), content (Heading, Text, Image, Avatar, Badge, Alert, Carousel, Table), feedback (Progress, Skeleton, Spinner), and input (Button, Link, Input, Textarea, Select, Checkbox, Radio, Switch, Slider, Toggle, ToggleGroup, ButtonGroup).
See the API reference for full details.
New: Event Handles (`on()`)
Components now receive an on(event) function in addition to emit(event). The on() function returns an EventHandle with metadata:
emit()-- fire the eventshouldPreventDefault-- whether any action binding requestedpreventDefaultbound-- whether any handler is bound to this event
Link: ({ props, on }) => {
const click = on("click");
return (
<a href={props.href} onClick={(e) => {
if (click.shouldPreventDefault) e.preventDefault();
click.emit();
}}>{props.label}</a>
);
},New: `BaseComponentProps`
Catalog-agnostic base type for component render functions. Use when building reusable component libraries (like @json-render/shadcn) that are not tied to a specific catalog.
import type { BaseComponentProps } from "@json-render/react";
const Card = ({ props, children }: BaseComponentProps<{ title?: string }>) => (
{props.title}{children}
);New: Built-in Actions in Schema
Schemas can now declare builtInActions -- actions that are always available at runtime and automatically injected into prompts. The React schema declares setState, pushState, and removeState as built-in, so they appear in prompts without needing to be listed in catalog actions.
New: `preventDefault` on `ActionBinding`
Action bindings now support a preventDefault boolean field, allowing the LLM to request that default browser behavior (e.g. navigation on links) be prevented.
Improved: Stream Transform Text Block Splitting
createJsonRenderTransform() now properly splits text blocks around spec data by emitting text-end/text-start pairs. This ensures the AI SDK creates separate text parts, preserving correct interleaving of prose and UI in message.parts.
Improved: `defineRegistry` Actions Requirement
defineRegistry now conditionally requires the actions field only when the catalog declares actions. Catalogs with no actions (e.g. actions: {}) no longer need to pass an empty actions object.
v0.6.0
February 13, 2026
New: Chat Mode (Inline GenUI)
Note: These modes were renamed in v0.12.1 — "Generate" is now "Standalone" and "Chat" is now "Inline". The old names are accepted as deprecated aliases.
json-render now supports two generation modes: Generate (JSONL-only, the default) and Chat (text + JSONL inline). Chat mode lets the AI respond conversationally with embedded UI specs, ideal for chatbots and copilot experiences.
// Generate mode (default) — AI outputs only JSONL
const prompt = catalog.prompt();
// Chat mode — AI outputs text + JSONL inline
const chatPrompt = catalog.prompt({ mode: "chat" });On the server, pipeJsonRender() separates text from JSONL patches in a mixed stream:
import { pipeJsonRender } from "@json-render/core";
import { createUIMessageStream, createUIMessageStreamResponse } from "ai";
const stream = createUIMessageStream({
execute: async ({ writer }) => {
writer.merge(pipeJsonRender(result.toUIMessageStream()));
},
});
return createUIMessageStreamResponse({ stream });On the client, useJsonRenderMessage extracts the spec and text from message parts:
import { useJsonRenderMessage } from "@json-render/react";
function ChatMessage({ message }) {
const { spec, text, hasSpec } = useJsonRenderMessage(message.parts);
return (
{text && <Markdown>{text}</Markdown>}
{hasSpec && <Renderer spec={spec} registry={registry} />}
);
}New: AI SDK Integration
First-class Vercel AI SDK support with typed data parts and stream utilities.
SpecDataParttype fordata-specstream parts (patch, flat, nested payloads)SPEC_DATA_PART/SPEC_DATA_PART_TYPEconstants for type-safe part filteringcreateJsonRenderTransform()low-level TransformStream for custom pipelinescreateMixedStreamParser()for parsing mixed text + JSONL streams
New: Two-Way Binding
Props can now use $bindState and $bindItem expressions for two-way data binding. The renderer resolves bindings and passes a bindings map to components, enabling write-back to state without custom valuePath props.
{
"type": "Input",
"props": { "label": "Email", "value": { "$bindState": "/form/email" } }
}import { useBoundProp } from "@json-render/react";
Input: ({ props, bindings }) => {
const [value, setValue] = useBoundProp<string>(props.value, bindings?.value);
return setValue(e.target.value)} />;
}New: Expression-Based Props and Visibility
All dynamic expressions now use structured $state, $item, and $index objects instead of string token rewriting. This is simpler, more explicit, and works for both props and visibility conditions.
Props:
{ "title": { "$state": "/user/name" } }
{ "label": { "$item": "title" } }
{ "position": { "$index": true } }Visibility:
{ "$state": "/isAdmin" }
{ "$state": "/role", "eq": "admin" }
[{ "$state": "/isAdmin" }, { "$state": "/feature" }]
{ "$or": [{ "$state": "/roleA" }, { "$state": "/roleB" }] }
{ "$item": "isActive" }
{ "$index": true, "gt": 0 }Comparison operators: eq, neq, gt, gte, lt, lte, not.
New: React Chat Hooks
useChatUI()— full chat hook with message history, streaming, and spec extractionuseJsonRenderMessage()— extract spec + text from a message's parts arraybuildSpecFromParts()/getTextFromParts()— utilities for working with AI SDK message partsuseBoundProp()— two-way binding hook for$bindState/$bindItem
New: Chat Example
Full-featured chat example (examples/chat) with AI agent, tool calls (crypto, GitHub, Hacker News, weather, search), theme toggle, and streaming inline UI generation.
Improved: Renderer Performance
ElementRendereris nowReact.memo'd for better performance with repeat listsemitis always defined (neverundefined)- Repeat scope passes the actual item object, eliminating string token rewriting
Improved: Utilities
applySpecPatch()— typed wrapper for applying a single patch to a SpecnestedToFlat()— convert nested tree specs to flat formatresolveBindings()/resolveActionParam()— resolve binding paths and action params
Breaking Changes
{ $path }and{ path }replaced by{ $state },{ $item },{ $index }in props- Visibility:
{ path }->{ $state },{ and/or/not }->{ $and/$or }withnotas operator flag DynamicValue:{ path: string }->{ $state: string }repeat.path->repeat.statePath- Action params:
path->statePathin setState action actionHandlers->handlersonJSONUIProvider/ActionProviderAuthStateand{ auth }visibility conditions removed (model auth as regular state)- Legacy catalog API removed:
createCatalog,generateCatalogPrompt,generateSystemPrompt - React exports removed:
createRendererFromCatalog,rewriteRepeatTokens - Codegen:
traverseTree->traverseSpec
See the Migration Guide for detailed upgrade instructions.
v0.5.0
February 9, 2026
New: @json-render/react-native
Full React Native renderer with 25+ standard components, data binding, visibility, actions, and dynamic props. Build AI-generated native mobile UIs with the same catalog-driven approach as web.
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react-native/schema";
import {
standardComponentDefinitions,
standardActionDefinitions,
} from "@json-render/react-native/catalog";
import { defineRegistry, Renderer } from "@json-render/react-native";
const catalog = defineCatalog(schema, {
components: { ...standardComponentDefinitions },
actions: standardActionDefinitions,
});
const { registry } = defineRegistry(catalog, { components: {} });
<Renderer spec={spec} registry={registry} />Includes standard components for layout (Container, Row, Column, ScrollContainer, SafeArea, Pressable, Spacer, Divider), content (Heading, Paragraph, Label, Image, Avatar, Badge, Chip), input (Button, TextInput, Switch, Checkbox, Slider, SearchBar), feedback (Spinner, ProgressBar), and composite (Card, ListItem, Modal).
New: Event System
Components now use emit to fire named events instead of directly dispatching actions. The element's on field maps events to action bindings, decoupling component logic from action handling.
// Component emits a named event
Button: ({ props, emit }) => (
emit("press")}>{props.label}
),
// Element spec maps events to actions
{
"type": "Button",
"props": { "label": "Submit" },
"on": { "press": { "action": "submit", "params": { "formId": "main" } } }
}New: Repeat/List Rendering
Elements can now iterate over state arrays using the repeat field. Child elements use { "$item": "field" } to read from the current item and { "$index": true } for the current array index.
{
"type": "Column",
"repeat": { "statePath": "/posts", "key": "id" },
"children": ["post-card"]
}{
"type": "Card",
"props": { "title": { "$item": "title" } }
}New: User Prompt Builder
Build structured user prompts with optional spec refinement and state context:
import { buildUserPrompt } from "@json-render/core";
// Fresh generation
buildUserPrompt({ prompt: "create a todo app" });
// Refinement (patch-only mode)
buildUserPrompt({ prompt: "add a toggle", currentSpec: spec });
// With runtime state
buildUserPrompt({ prompt: "show data", state: { todos: [] } });New: Spec Validation
Validate spec structure and auto-fix common issues:
import { validateSpec, autoFixSpec } from "@json-render/core";
const { valid, issues } = validateSpec(spec);
const fixed = autoFixSpec(spec);Improved: State Management
DataProvider has been renamed to StateProvider with a clearer API. State is now a first-class part of specs. Elements can bind to state via $state expressions, and the built-in setState action updates state directly.
Improved: AI Prompts
Schema prompts now include streaming best practices, repeat/list examples, and state patching guidance. Schemas can also define defaultRules that are always included in generated prompts.
Improved: Documentation
- All documentation pages migrated to MDX
- AI-powered documentation chat
- Dynamic Open Graph images for all docs pages
- Improved playground
Breaking Changes
DataProviderrenamed toStateProvideruseDatarenamed touseStateStore,useDataValuetouseStateValue,useDataBindingtouseStateBindingonActionrenamed toemitin component contextDataModeltype renamed toStateModelActiontype renamed toActionBinding(old name still available but deprecated)
v0.4.0
February 5, 2026
New: Custom Schema System
Create custom output formats with defineSchema. Each renderer now defines its own schema, enabling completely different spec formats for different use cases.
import { defineSchema } from "@json-render/core";
const mySchema = defineSchema((s) => ({
spec: s.object({
pages: s.array(s.object({
title: s.string(),
blocks: s.array(s.ref("catalog.blocks")),
})),
}),
catalog: s.object({
blocks: s.map({ props: s.zod(), description: s.string() }),
}),
}), {
promptTemplate: myPromptTemplate,
});New: Component Slots
Components can now define which slots they accept. Use ["default"] for regular children, or named slots like ["header", "footer"] for more complex layouts.
const catalog = defineCatalog(schema, {
components: {
Card: {
props: z.object({ title: z.string() }),
slots: ["default"], // accepts children
},
Layout: {
props: z.object({}),
slots: ["header", "content", "footer"], // named slots
},
},
});New: AI Prompt Generation
Catalogs now generate AI system prompts automatically with catalog.prompt(). The prompt includes all component definitions, props schemas, and action descriptions - ensuring the AI only generates valid specs.
import { defineCatalog } from "@json-render/core";
import { schema } from "@json-render/react/schema";
const catalog = defineCatalog(schema, {
components: { /* ... */ },
actions: { /* ... */ },
});
// Generate system prompt for AI
const systemPrompt = catalog.prompt();
// Use with any AI SDK
const result = await streamText({
model: "claude-haiku-4.5",
system: systemPrompt,
prompt: userMessage,
});New: @json-render/remotion
Generate AI-powered videos with Remotion. Define video catalogs, stream timeline specs, and render with the Remotion Player.
import { Player } from "@remotion/player";
import { Renderer, schema, standardComponentDefinitions } from "@json-render/remotion";
const catalog = defineCatalog(schema, {
components: standardComponentDefinitions,
transitions: standardTransitionDefinitions,
});Includes 10 standard video components (TitleCard, TypingText, SplitScreen, etc.), 7 transition types, and the ClipWrapper utility for custom components.
New: SpecStream
SpecStream is json-render's streaming format for progressively building specs from JSONL patches. The new compiler API makes it easy to process streaming AI responses.
import { createSpecStreamCompiler } from "@json-render/core";
const compiler = createSpecStreamCompiler<MySpec>();
// Process streaming chunks
const { result, newPatches } = compiler.push(chunk);
setSpec(result); // Update UI with partial resultImproved: Dashboard Example
The dashboard example is now a full-featured accounting dashboard with:
- Persistent SQLite database with Drizzle ORM
- RESTful API for customers, invoices, expenses, accounts
- Draggable widget reordering
- AI-powered widget generation with streaming
- Real data binding to database records
Improved: Documentation
- Interactive playground for testing specs
- New guides: Custom Schema, Streaming, Code Export
- Full API reference for all packages
- Integration guides: A2UI, AG-UI, Adaptive Cards, OpenAPI
Breaking Changes
UITreetype renamed toSpec- Schema is now imported from renderer packages (
@json-render/react) not core defineCatalognow requires a schema as first argument
v0.3.0
January 20, 2026
Internal release with codegen foundations.
- Added
@json-render/codegenpackage (spec traversal and JSX serialization) - Configurable AI model via environment variables
- Documentation improvements and bug fixes
Note: Only @json-render/core was published to npm for this release.
v0.2.0
January 14, 2026
Initial public release.
- Core catalog and spec types
- React renderer with contexts for data, actions, visibility
- AI prompt generation from catalogs
- Basic streaming support
- Dashboard example application