Installation

Install the SDK from npm and bundle it with your web application.

Install

npm install @invox-medical-npm/invox-dictation

The package provides ESM and CommonJS entry points, TypeScript declarations, the browser bundle, wake-word assets, and sample integrations. Use a bundler such as Vite, Webpack, or a framework build pipeline to resolve the package import.

Import the SDK

import INVOX, {
  AutoFillTemplatesDialog,
  BarComponent,
  DictionaryDialog,
  IntelligentFormattingDialog,
  QualityReviewDialog,
  TemplatesDialog,
  TransformationsDialog,
} from "@invox-medical-npm/invox-dictation";

Component CSS is injected by the SDK. Do not add a separate component stylesheet.

Deploy wake-word assets

If your integration uses wake-word detection, publish the package wake-word-assets directory with the application. By default, the SDK looks for it next to the SDK bundle; if you publish it in another location, set its public URL with wakeWordAssetBaseUrl during initialization. The files must be served from the same origin as the application, and the server must return the correct MIME types for .wasm and .onnx assets.

INVOX.init({
  organizationId: "provided-organization-id",
  appId: "provided_app_id",
  apiKey: "provided_api_key",
  wakeWordAssetBaseUrl: "/assets/wake-word-assets/",
});

The SDK selects its Direct Dictation endpoint from the installed build. Do not configure an ASR endpoint with transcriptionServiceUrl; that property is not part of the 2.9.1 initialization contract.

Browser requirements

Serve the application over HTTPS in a browser that supports getUserMedia, WebAssembly, Workers, and AudioWorklet. Opening an HTML file directly from the file system is not supported. Validate microphone permissions and the configured wake-word assets in every browser you intend to support.