Configure GPTfy: Kensho and Google Chirp Integration for GPTfy Voice

Configuring Kensho and Google Chirp for GPTfy Voice ensures seamless integration of advanced voice capabilities into your Salesforce environment. This guide walks you through the essential steps, from setting up connected apps and authentication to activating AI models, enabling efficient voice processing and async functionality.

Kensho (Async) Configuration for GPTfy Voice

Here are the high-level steps to configure Kensho with GPTfy Voice:

Create a salesforce-connected app
Salesforce Auth. Provider
Named credentials
Create 2 remote site settings
The org’s ‘My Domain’
Three AI Models
- Kensho (Async Processing)
- Kensho (Public URL Generator)
- Kensho (Polling)
A file needs to be uploaded whose id acts as a private key.
Need to create a public user group and share the private key with the public group
Drag and drop the Audio listener component on the record page.
Check all the users have given licenses in installed packages → manage licenses.

Create a Salesforce Connected app

Go to Setup → App Manager → Click on New connected App → Create a Connected App → Continue.

alt text

For the call back URL, we will get that from Auth. Provider.

alt text

Add the scope

alt text

Once done, click save. In the next window, hit Continue and wait for 10 minutes for the app to be created.

alt text

Auth. Provider

Create a new Auth. provider.

alt text

The values in the consumer key and secret must be changed once the connected app is created.
So first, input some dummy values (you can use the value in the table as dummy values) in these two, and once saved, we will get the Callback URL. Copy that URL and use it to connect app creation.

Once the connected app is created, click on the ‘Manage Consumer Details’ button.

alt text

It will take you to the login page, enter the credentials, and sometimes it may ask for a verification code, too.

Once done, it will open a screen with the Consumer key and secret. And the key will be visible there.

alt text

Copy those and update them in the Auth. Provider Created.

Named Credential

Create a Legacy type Named credential as per the table

alt text

URL should be from setup → My Domain → copy and paste with ‘https://’ at the front of the URL
Don’t forget to add ‘/’ at end of the URL. eg – https://sgpt-qa-dev-ed.develop.my.salesforce.com/
Select the Auth. Provider we created
Click Save, and it will get authenticated as the user you logged in

Create 2 Remote Site Settings

Create two remote site settings as per the screen shot.

Kensho_Okta

Kensho_Scribe

alt text

Create 3 AI Models

These models will have different sets of named cred and auth providers.
The model’s card details as shown in the table below

Model’s Connection details

alt text

Kensho Async processing will have additional parameters

{
  "privateKeyFileId": " ",
  "clientId": " ",
  "authUrl": "https://kensho.okta.com",
  "tokenUrl": "https://kensho.okta.com/oauth2/default/v1/token"
}

Model’s Named Credentials - legacy type

alt text

Upload private files by searching files in the app launcher (9 dots).
We are adding this file whose id acts as a private key.
Private.pem

alt text

Need to create public user group and share the private key to the public group.

alt text

Create a public group by searching groups in the app launcher (9 dots)
Click on new to create a new group and add a name to the group
Make the access type public and save

alt text

Go to the private key and share the file to this group.

alt text

Copy the private key file’s record id and paste it inside the additional parameters of the model – Kensho (Async Processing) as below.

{
  "privateKeyFileId": "069dN00001ThfjQAS",
  "clientId": " ",
  "authUrl": "https://kensho.okta.com",
  "tokenUrl": "https://kensho.okta.com/oauth2/default/v1/token"
}

alt text

Drag and Drop Components

Drag and drop the audio recorder and audio listener component on the record page where the voice recorder wants to be configured.

alt text

Check all the users have been given licenses in installed packages. Go to Manage Licenses.
Check in setup → installed packages → secure gpt → click on manage licenses.

We need to be sure that all users using the voice record (be an admin, be a std user) have to be given a license here.