AI Ticket Summary Documentation

Owned by Tony Thai Do
Apr 12, 2024

AI Ticket Summary simplifies the management of IT support tickets. When dealing with multiple support requests, it's easy for crucial information to get buried under numerous responses and updates. This solution adeptly analyzes the content of ITSM tickets, extracting and summarizing the key details from extensive conversations.

You may use multiple languages.

By delivering these summaries, it allows support teams to quickly understand the main issues without wading through all the interactions. As a result, Ticket Summary aids in maintaining focus and improving the efficiency of problem resolution, providing a straightforward and effective tool for IT support workflows.

Architecture Overview

You have full control on the provisioned resources.
Azure Managed Applications provide a Infrastructure as a service (IaaS) solution for delivering fully managed applications within Azure, offering numerous benefits for you seeking a reliable, scalable, and secure cloud environment.
This service allows you to simplify deployment and manage applications without worrying about the underlying infrastructure, enabling you to focus on your core business functions.

Key Reasons to Choose Azure Managed Applications:

  • Subscription Independence: You can deploy these applications in their own Azure subscriptions, ensuring they maintain control over their resources while benefiting from a managed service.

  • Policy Safety: Azure Managed Applications adhere to strict governance and compliance standards, ensuring that applications are secure and conform to organizational policies.

  • Advantages: Simplified management, reduced operational overhead, and the assurance that their applications are running in a secure, best-practice environment.

The architecture of this specific solution includes Azure Function App, its associated storage account, Cognitive Services, and the deployment of the GPT-3.5-instruct model. This configuration offers a robust, scalable platform for leveraging advanced AI capabilities within your applications.

Technical details of AI Ticket Summary solution

Before Deployment

Prerequisites

Make sure to have enough permission to provision resources for Function App, its associated Storage Account, Cognitive Services, and its model deployment

If you have deployed any Azure AI service before you may skip following steps.

1. Register the Cognitive Services Resource Provider

Azure Resource Providers are services that provide Azure resources. Before you can create a Cognitive Services resource (which is a service we do use in our solution), you need to ensure that the Cognitive Services Resource Provider is registered.

Check, Register Through Azure Portal
Note: To register a resource provider, you must have the Microsoft.Authorization/register/action permission. This permission is included in the built-in roles for Owner, User Access Administrator. The Contributor role does not have this permission.

  1. Sign in to the Azure portal.

  2. In the left-hand menu, click on "Resource groups".

  3. Select your resource group.

  4. Click on "Settings" > "Resource providers".

  5. In the list of resource providers, look for "Microsoft.CognitiveServices".

  6. If the status of "Microsoft.CognitiveServices" is "NotRegistered", click on "Register".

2. Deploy a Test AI Service to Accept the Responsible AI Notice

When you create a Cognitive Services resource for the first time (or actually any AI service), you need to accept the Responsible AI notice. This notice outlines Microsoft's guidelines and expectations for using AI responsibly.

  1. In the Azure portal, click on "Create a resource".

  2. Search for "Cognitive Services" and click on "Create".

  3. Fill in the required information and click on "Review + create".

  4. You will see the Responsible AI terms notice. Read the notice and click on "Accept" to accept the terms.

Note: After you have accepted the RAI notice, you can delete the test AI service. The acceptance of the RAI notice is tied to your account, not to the specific Cognitive Services resource.

After you have completed these steps, you can proceed with deploying the AI Ticket Summary application.

After deployment

Find your newly deployed resources of your managed application located in the managed resource group. Its model deployment properties will be needed for the configuration of AI Ticket Summary Solution.

Find your AI cognitive resource deployment name

  • Model deployments → Manage Deployments

find_name.png

Retrieve the Endpoint and API Token:

  • After deploying the Managed Application, go to GPT-3.5-instruct model, access the Azure portal to find the endpoint URL and API token. These are crucial for enabling your AI Ticket Summary to communicate with the model.

image-20240410-074623.png

Configure AI Ticket Summary Solution:

  1. Find in deployed resources your Azure Function App settings within the Azure portal.

  2. Navigate to the "Configuration" section where you can set application settings and connection strings.

  3. Add application settings for the GPT-3.5-instruct model endpoint, model deployment and API token.

  4. Use your deployed resources variables that you retreived in last step. You will need models api key, model deployments name, and its endpoint. These function app configuration settings are AZURE_OPENAI_KEY, AZURE_OPENAI_GPT_DEPLOYMENT and AZURE_OPENAI_ENDPOINT and can look like c701a96263d9465baf714137e43b831b, east-us-gpt-35-turbo-instruct, https://test-east-us-azure-openai.openai.azure.com/

  5. Ensure these settings are saved and properly secured within your Azure Function App environment.

By correctly configuring these settings, your AI Ticket Summary solution will be equipped to interact with the AI model, leveraging its AI capabilities to enhance your application's functionality. Remember to review and test the integration thoroughly to ensure optimal performance and reliability.

After Configuration

Use Case: Sending Requests to AI Ticket Summary

Congratulations! You're now prepared to utilize the AI Ticket Summary service to analyze and summarize tickets effectively. To ensure secure and authorized access to the service, it's crucial to include the appropriate headers in your requests.

Authorization and Security:

When interacting with the AI Ticket Summary service, it's essential to prioritize security by employing proper authorization mechanisms.

To achieve this, your requests should include an authorization header with an access token obtained from the Azure Function App. This access token serves as a form of authentication, verifying that the requester has the necessary permissions to access the service.

Including Headers:

In your requests, make sure to include the following header x-functions-key

  • Authorization: The authorization header should contain the access token retrieved from the Function App's App Keys. This token serves as proof of authentication and grants access to the AI Ticket Summary service.

Instructions:

  1. Navigate to the Azure Function App's management portal within your newly deployed resources.

  2. Locate the "App Keys" section within the Function App settings.

  3. Copy the default function app key provided in the "App Keys" section.

  4. When making requests to the AI Ticket Summary service, include the following header:

    x-functions-key: [your_function_app_key]

Example request

curl -X POST <URL>/api/ticket_summary \
-H "Content-Type: application/json" \
-H "x-functions-key: <your-function-key>" \
-d '{"prompt":"Description:As of this morning, I'm unable to access our company's website. The browser displays a \\\"server not found\\\"error", "language":"English"}'


Note:
You can find the URL of the application within your newly deployed resources like this:

  • Find the Deployed resource of Function App type and check its overview