DoclessAI SDK

Add a context-aware AI chat assistant to your application. Define your app's features — users ask questions, the AI answers based on what you've defined.

What problem does DoclessAI solve?

Apps often have features users don't understand — and they don't want to search through documentation to find answers. DoclessAI embeds an AI assistant directly inside your application. The assistant understands your app because you tell it what your app does — through feature definitions you write in the dashboard.

When a user asks "How do I export my data?", the assistant doesn't guess. It answers based on the export feature you defined, including the exact route, element ID to highlight.

Two integration paths

Quick Start

Get the assistant running in your app in a few steps.

 npm install @doclessai/sdk

 import { ChatWidget } from '@doclessai/sdk'

 export default function App()	{

 		return (

 			<div>

 				{/* AI assistant is now live */}

 				<ChatWidget appKey="your-app-key" name="your-assistant-name" />

 			</div>

 		)

 }

Dashboard Setup

The dashboard is where you manage your apps and the features the AI knows about.

Creating an app

Go to the dashboard and click Create App. Fill in:

• → App name — displayed to users in the chat widget
• → Contact email — associated with this app
• → App description — tells the AI what your app is about overall
• → Gemini API Key — your own Gemini key, used to power the AI responses for this app
• → App Features (JSON) — the feature definitions the AI learns from

After creation, you receive your App Key — this is what you pass to the SDK.

Defining Features

Features are how you give the AI knowledge about your app. Write them as a JSON array in the dashboard. Each feature object describes one capability of your app.

Feature structure

 [

 	{

 		"name": "Export Data"

 		"description": "Users can export their data as CSV or JSON. Go to Settings → Data → Export, choose a format and date range. The export file is sent to the account email."

 		"route": "/settings/export"

 		"elementId": "#export-button"

 	},

 	{

 		"name": "Team Roles"

 		"description": "Manage user roles and permissions within your team. Go to Settings → Team → Roles to define and assign roles."

 		"route": "/settings/team/roles"

 	},

 ]

Feature fields

ChatWidget (React)

The simplest way to add the AI assistant. Works in React and Next.js projects.

 import { ChatWidget } from '@doclessai/sdk'

 <ChatWidget
		appKey="your-app-key"
		name="your-assistant-name"
	/>

Props

What's included

• ✓ Floating chat button that opens a full chat panel
• ✓ Markdown rendering in responses (bold, code, lists)
• ✓ Image upload — users send images, AI analyzes them
• ✓ Toast notifications for errors
• ✓ Fully responsive on mobile
• ✓ Feature navigation (usesroute from response)

HTML Script Tag

Not using React? Add the assistant to any HTML page with a single script tag — no build step needed.

 <!-- Before closing </body> -->

 <script

 		src="https://cdn.jsdelivr.net/npm/@doclessai/sdk@0.3.5/dist/loader.standalone.js"

 		data-app-key="your-app-key-here"

 		data-name="YOUR_ASSISTANT_NAME"

 ></script>

DoclessClient (Headless)

Use the client directly if you want to build your own chat UI. You handle the interface — we handle the AI.

 import {  DoclessClient  } from '@doclessai/sdk'

 const ai = new DoclessClient({

 	appKey: 'your-app-key-here'

 })

 // Text query

 const res = await ai.ask(userQuery)

 // Text + image

 const res = await ai.ask(userQuery, imageFile)

 // Use the response

 console.log(res.res)       // AI answer text

 console.log(res.route)     // route to navigate to

 console.log(res.elementId) // DOM element to highlight

Next.js Setup

ChatWidget is a client component. Use"use client" and pass your key via environment variable.

.env.local

 NEXT_PUBLIC_DOCLESSAI_KEY=your-app-key-here

app/layout.tsx (App Router)

 'use client'

 import {  ChatWidget  } from '@doclessai/sdk'

 export default function RootLayout({ children }: { children: React.ReactNode }) {

 	return (

 		<html>

 			<body>

 				{children}

 				<ChatWidget

 					appKey={process.env.NEXT_PUBLIC_DOCLESSAI_KEY}

 					name="Aria"

 				/>

 			</body>

 		</html>

 	)

 }

Response Format

Both ChatWidget andDoclessClient.ask() return the same response shape:

 {

 		res: string           // AI's text answer (markdown supported)

 		route: string | null   // App route to navigate to

 		elementId: string | null // DOM element to highlight

 }