Integrate the SDK › Implementation

Overview

Integrate Speechineer into your software so that your users fill your forms by speaking — or by pasting a text — while your own form keeps its state. It is four steps. What can go wrong at any step, and how to handle it in code, is described in Errors.

1. Prerequisites and technical requirements

What must exist in the workspace before you write code — a published form or the intent to define one in code, an API key in the right mode, a signing key pair for the authenticated way, a stable key per end user — and what your frontend, and for the authenticated way your backend, needs technically.

2. Create the client

One client, created once at the root of your app. Unauthenticated, it takes the Unsigned API key and the end user's account. Authenticated, it takes a function that fetches a token from your backend, so your backend needs an endpoint that signs one. Several clients can coexist in one app.

3. Build the form object

The form object every session takes: either the key, version and language of a form configured in the workspace, or, for a form configured in code, that same identity plus its fields and, optionally, its prompts and models.

4. Wire up the voice session

The session itself: start, stop and end it, put the recognized values into your own form state, render from the session state, react to callbacks, stream the transcript, and continue from values your form already holds.