Loading...

Build a low code conversational SMS bot with Azure Communication Services and Azure OpenAI

Build a low code conversational SMS bot with Azure Communication Services and Azure OpenAI

In our last blog, we walked through the process of building an SMS conversational bot. Today we want to showcase a similar bot, but this time built completely on top of low code connectors for Power Automate. It is incredibly easy and exciting to leverage Azure Communication Services connectors in conjunction with Large Language Models (LLM) like ChatGPT to build a conversation bot. For this blog, we will be building Anton, a personal trainer bot for Contoso Gym. See the preview:

 

ddematheu_0-1677025513942.png

 

To follow along you will need:

 

This application will leverage Azure Event Grid to listen for incoming text messages to Azure Communication Services number, and an Azure Function to process the event and respond with an Azure OpenAI generated response.

 

We will start by configuring an Event Grid trigger for your Power Automate flow. This trigger will automatically connect to your existing Azure subscription and to the Azure Communication Services resource to configure your event subscription.

 

ddematheu_1-1677025657239.png

 

To keep the data organized, you will add a Parse JSON connector to help us parse the event. The event schema follows this structure:

 

Schema (click to see) { "properties": { "data": { "properties": { "From": { "type": "string" }, "Message": { "type": "string" }, "MessageId": { "type": "string" }, "ReceivedTimestamp": { "type": "string" }, "To": { "type": "string" } }, "type": "object" }, "dataVersion": { "type": "string" }, "eventTime": { "type": "string" }, "eventType": { "type": "string" }, "id": { "type": "string" }, "metadataVersion": { "type": "string" }, "subject": { "type": "string" }, "topic": { "type": "string" } }, "type": "object" }

 

Add that schema to the Parse JSON connector and add the event body as the content.

ddematheu_2-1677025761655.png

 

Next, we must add a call to Azure OpenAI to ask the model to generate a response. We will use REST APIs to POST a request with our prompt. For the prompt, we will use a combination of the message sent by the user and a pre-designed text. In this example, we want the GPT-3 model to act like Anton, a personal trainer at Contoso Gym. We added some sample quotes from which the model can draw inspiration. These quotes help guide the model’s response and provide a more intuitive and conversational flow for the users. See the completed prompt below:

 

 

You're Anton Stanescu and you are a personal trainer at Contoso Gym. You are having a conversation with one of your students. You are trying to motivate them to work out. Examples of the types of things that Anton says: - The only bad workout is the one that didn't happen. - Strive for progress, not perfection. - You won't get the muscle without the dedication and the drive - Train hard...so you can stop a train if you need to. - Push up, push down, push up, push down - Exercise should be regarded as tribute to the heart. Anton: Welcome to the Contoso Gym! Big things await for you. Student:

 

 

To keep the flow organized, we will add a compose connector for our prompt so we can easily reference it later.

ddematheu_3-1677025812597.png

 

To communicate with the Azure OpenAI service, we will use POST request inside of an HTTP connector. In return, we will receive a text generated by the Azure OpenAI model. To configure the HTTP connector, you will need the URI to your Azure OpenAI resource. You will also need the `api-key` for your Azure OpenAI resource.

 

 

<base_url> + /openai/deployments/ + <deployment_name> + /completions?api-version=2022-12-01

 

 

ddematheu_4-1677025870598.png

 

Using the compose output for our prompt, we will concatenate together the prompt, the message the user sent through SMS, and a stop trigger for the Open AI model to stop generating text:

 

 

concat(outputs('Compose'), body('Parse_JSON')?['data']?['Message'], '\nAnton:')

 

 

Once we have the HTTP connector d, we will use a Parse JSON connector to process the response from Azure Open AI. The expected schema from Azure Open AI takes the following format:

 

Schema (click to see) { "properties": { "choices": { "items": { "properties": { "finish_reason": { "type": "string" }, "index": { "type": "integer" }, "logprobs": {}, "text": { "type": "string" } }, "required": [ "text", "index", "logprobs", "finish_reason" ], "type": "object" }, "type": "array" }, "created": { "type": "integer" }, "id": { "type": "string" }, "model": { "type": "string" }, "object": { "type": "string" } }, "type": "object" }

 

We will configure it to use the body of the http request as the content the schema above.

ddematheu_5-1677025928198.png

 

Finally, we will configure our SMS connector for Azure Communication Services to respond with the new response generated by Azure OpenAI. You will need a connection string for your Azure Communication Services resource to initialize the SMS connector.

 

ddematheu_7-1677025978170.png

 

Now that everything is connected, test the flow by sending an SMS from your phone to the phone number you acquired through the Azure Communication Services resource.

Congratulations! using Power Automate. Azure Communication Services provides out of the box connectors for SMS, Email and Chat. Whether it is having Anton, the personal trainer bot talk to your customers or having a real person interact with them, Azure Communication Services can help. To learn more about other resources that Azure Communication Services offers see the links below:

Published on:

Learn more
Azure Communication Services Blog
Azure Communication Services Blog

Azure Communication Services Blog articles

Share post:

Related posts

How Azure AI Search powers RAG in ChatGPT and global scale apps

Millions of people use Azure AI Search every day without knowing it. You can enable your apps with the same search that enables retrieval-augm...

21 hours ago

Episode 388 – Getting Started with Azure Bicep: Infrastructure as Code with a Domain Specific Language

Welcome to Episode 388 of the Microsoft Cloud IT Pro Podcast. In this episode, we dive into Azure Bicep, Microsoft’s streamlined language for ...

23 hours ago

RAG with SQL Vector Store: A Low-Code/No-Code Approach using Azure Logic Apps

Data is at the heart of every AI application, and efficient data ingestion is critical for success. With over 1,400 enterprise connectors, Log...

1 day ago

Exciting Announcement: Public Preview of Native Vector Support in Azure SQL Database!

Public Preview of Native Vector Support in Azure SQL Database We are excited to share that the dedicated vector data type in Azure SQL Databas...

1 day ago

Scaling Azure Container Apps for Peak Performance

Unlock the full potential of Azure Container Apps with smart scaling! In this final part of our series, explore how to keep your apps responsi...

2 days ago
Stay up to date with latest Microsoft Dynamics 365 and Power Platform news!
* Yes, I agree to the privacy policy