🔧 Custom Functions – Extend Your Chatbot's Capabilities

Overview

Custom functions allows you to extend your chatbot by connecting it to external APIs — enabling it to do things like send emails, create documents, or schedule calendar events, all through natural language.

What We’ll Build

We’ll build a custom function that sends targeted emails to users who requested a specific feature — using insights from past call transcripts. For example, after releasing dark mode, the chatbot can automatically identify users who asked for it and send them a thank-you email.

This closes the feedback loop and turns feature launches into personal, high-impact moments — all triggered by a single natural language command.

Step 1: Requirements

To make this workflow possible, your call transcripts need to be stored in a database that BlazeSQL can query — such as BigQuery, PostgreSQL, or Snowflake.

If you’re using a tool like Fathom, you can automate this by connecting it to Zapier and writing each transcript into your preferred database. Once transcripts are stored, BlazeSQL can search them to identify users who requested specific features and trigger targeted emails in response.

Step 2: Create the custom function

Open the new function dialog

Click on the functions tab on the left side
Click on the new function button

Set Up the API Connection

API Endpoint: Enter the full URL to your email-sending API, for example use Zapier to send emails.

Example: https://hooks.zapier.com/hooks/catch/18139573/2jsqojd
Request Type: Choose POST.
Authentication: Toggle it On.
Type: Select Bearer token.
Token: Paste your token here (this is used in the Authorization header).

Click Next to continue.

Define Parameters

Set the name and description

Now define the core settings for your function.

Fill in the following:

Name: Send emails
Description: Sends an email to one or more recipients using the provided subject and HTML body.*

🔐 Ask for Approval Before Running

If you want to have control before the assistant executes the function, toggle on "Ask for approval in chat before running the function".

Here’s what it looks like in chat:

Set receivers (Array) parameter

Name: receivers
Type: Array
Description: List of email addresses to receive the email.

Set the subject (String) parameter

Name: subject
Type: String
Description: Subject line of the email.

Set the body (String) parameter

Name: body
Type: String
Description: Main content of the email in HTML.

Click Next to continue.

Test the Function

Now test the function by entering values for:

receivers: e.g., example1@mail.com, example2@mail.com
subject: e.g., Meeting Reminder
body: e.g., Don't forget about our 10 AM meeting tomorrow.

Click Test to verify the integration.

Save the Function

If the test is successful, you will see a confirmation message.
A Save button appears at the bottom.
Click Save to finalize and store your custom function.

Step 3: Use the function

Ask BlazeSQL to Send Emails

Once the function is set up, BlazeSQL can automatically use it in conversations — triggered by natural language.
For example, you can simply say:

Send an email to all users who mentioned ‘dark mode’ in their call transcripts. Let them know the feature is now available, thank them for their feedback, and make it look fancy.

Edit the Suggested Message (Optional)

BlazeSQL will generate a draft email based on your prompt. You can review and customize the content before sending it out.

Approve or Reject the Function Call

Before any email is sent, you’ll see a preview and can approve or reject the action — giving you full control over what gets delivered.

Updated on: 03/06/2025