Example: Ask AI (Chat Completions API) with Service Principal authentication and save the response to a Local Path

In this example, the user configures a Universal Task to invoke the Chat Completions API using Azure Service Principal authentication and the gpt-4o model. The task is configured to capture the AI-generated response in the EXTENSION output rather than printing it to STDOUT. Finally, the user enables the display of token usage metadata, providing visibility into the resources consumed during response generation.

Example: Ask AI (Chat Completions API) with API Key authentication and fetch the prompts from UAC Scripts

In this example, the user configures a Universal Task to invoke the Chat Completions API using API Key authentication. Both the system prompt and the user prompt (conversation thread) are sourced from UAC Scripts. The task is also configured to display the entire conversation in STDOUT as well as in the Extension Output.

Example: Ask AI (Chat Completions API) with API Key using Azure Foundry endpoint and a 3rd-party Model

In this example, the user configures a Universal Task to invoke the Chat Completions API using API Key authentication against an Azure Foundry endpoint. Instead of an OpenAI model, the task is configured to use the Llama-3.3-70B-Instruct third-party model. The task is configured to display the latest response in STDOUT as well as in the EXTENSION Output.

{
    "exit_code": 0,
    "status_description": "Task executed successfully",
    "invocation": {
        "extension": "ue-azure-openai",
        "version": "1.0.0",
        "fields": {
            "action": "Ask AI (Chat Completions API)",
            "azure_credentials": { ... },
            "authentication_method": "API Key",
            "azure_endpoint": "https://resource-name.openai.azure.com/openai/v1/",
            "model_dynamic": "",
            "model_manual": "gpt-4o",
            "system_prompt_source": "Text Field",
            "system_prompt_text": "You are a helpful assistant.",
            "system_prompt_script": null,
            "conversation_thread_source": "Text Field",
            "conversation_thread_text": "What is 2 + 2?",
            "conversation_thread_script": null,
            "stdout_output": ["Show Latest Response Only"],
            "extension_output": ["Show Latest Response Only", 
                "Show Entire Conversation", 
                "Token Usage and Metadata"
            ],
            "azure_tenant_id": "",
            "azure_subscription_id": "",
            "resource_group": "",
            "advanced_options": false,
            "files": null,
            "response_format_type": "Text",
            "response_format_json_schema": null,
            "save_options": "-- None --",
            "save_source": "",
            "save_local_path": "",
            "dry_run": false,
            "temperature": 1.0,
            "max_tokens": 1000,
            "top_p": 1.0,
            "frequency_penalty": 0.0,
            "presence_penalty": 0.0,
            "allow_conversation_retention": false
        }
    },
    "result": {
        "response": "2 + 2 = 4",
        "metadata": {
            "prompt_tokens": 25,
            "completion_tokens": 8,
            "total_tokens": 33,
            "finish_reason": "stop"
        },
        "complete_conversation": [
            {
                "role": "system",
                "content": "You are a helpful assistant."
            },
            {
                "role": "user",
                "content": "What is 2 + 2?"
            },
            {
                "role": "assistant",
                "content": "2 + 2 = 4"
            }
        ]
    }
}

{
    "exit_code": 1,
    "status_description": "API Request Failed: Chat completion failed: Error code: 400 - 
        {'error': {'code': 'unknown_model','message': 'Unknown model: gpt-9', 
        'details': 'Unknown model: gpt-9'}}",
    "invocation": {
        "extension": "ue-azure-openai",
        "version": "1.0.0",
        "fields": {
            "save_options": "-- None --",
            "stdout_output": "Show Latest Response Only",
            "azure_endpoint": "https://resource-name.openai.azure.com/",
            "model_manual": "gpt-9",
            "presence_penalty": null,
            "azure_subscription_id": null,
            "model_dynamic": null,
            "response_format_type": "Text",
            "dry_run": false,
            "save_local_path": null,
            "extension_output": [
                "Show Latest Response Only",
                "Show Entire Conversation",
                "Token Usage and Metadata"
            ],
            "allow_conversation_retention": false,
            "resource_group": null,
            "max_tokens": 1000,
            "conversation_thread_text": "What is 2 + 2?",
            "save_source": null,
            "conversation_thread_source": "Text Field",
            "frequency_penalty": null,
            "action": "Ask AI (Chat Completions API)",
            "authentication_method": "API Key",
            "system_prompt_text": "You are a helpful assistant.",
            "top_p": 1.0,
            "azure_tenant_id": null,
            "advanced_options": false,
            "system_prompt_source": "Text Field",
            "temperature": 1.0,
            "azure_credentials": { ... }
        }
    },
    "result": {
        "errors": [ 
            "API Request Failed: Chat completion failed: Error code: 400 - 
            {'error': {'code': 'unknown_model', 'message': 'Unknown model: gpt-9', 
            'details': 'Unknown model: gpt-9'}}"
        ]
    }
}

Use Case 1: Email Classification, Output Transformation & Automated Incident Creation on ServiceNow

Description

This workflow uses Azure OpenAI to analyze incoming emails and send them to the right process. The Email Monitor integration watches a mailbox and creates a Universal Event for each new message, triggering a UAC Workflow. Azure OpenAI UAC Tasks then classify the email as a purchase request, an infrastructure IT issue, or an undefined case that needs human review.

If the email is an infrastructure IT request, another Azure OpenAI Task converts the email content into structured JSON, which the ServiceNow Incident integration uses to automatically create an Incident. This demonstrates end-to-end branching and data transformation in UAC powered by Azure OpenAI.​

The tasks configured demonstrate the following capabilities among others:

• Capability to trigger a task or a workflow based on received Emails.
• Propagation of useful information (like Email attributes) to downstream tasks.
• Utilize Azure OpenAI Tasks to classify information and transform it into structured formats.
• Handle business cases where the automatic creation of ServiceNow Incidents is necessary.

The components of the solution are described below:

1. Represents the receipt of an e-mail to a Mailbox.
2. "UC1: Azure OpenAI - Monitor Email from Microsoft Server (Event Producer)" - Subscribes to Microsoft Server, monitors emails, saves attachments and emails on the Agent's filesystem and sends Universal Event for each monitored email.
3. "UC1: Azure OpenAI - Universal Monitor: Wait for Email (Event Consumer)" - Monitors the Universal Events published from "UC1: Azure OpenAI - Monitor Email from Microsoft Server (Event Producer)"
4. "UC1: Azure OpenAI - Trigger Workflow" - Triggers the configured workflow "UC1: Azure OpenAI - Email Classification & Transformation" when the Monitor condition on (3) is true.
5. "UC1: Azure OpenAI - Email Classification Task" - Calls gpt-4 to classify the contents of the received e-mail with one of the following values "purchase", "infrastructure", and "undefined".
6. "UC1: Azure OpenAI - Email Transformation to Incident Task" - Transforms the contents of the received e-mail into a "ServiceNow Incident"-compatible JSON.
7. "UC1: Azure OpenAI - Create a ServiceNow Incident Task" - Creates a ServiceNow Incident using the JSON produced in the previous step.
8. "UC1: Azure OpenAI - Forward Purchase Request To Finance" - Task representing actions related to forwarding a purchase request to a Finance department.
9. "UC1: Azure OpenAI - Represents Manual Actions" - Task representing actions related to introducing human intervention in cases that were unable to be classified by the "Azure OpenAI: Email Classification Task".

How to Run

Execution Steps:

• STEP 1: Enable Trigger "UC1: Azure OpenAI - Trigger Workflow". This automatically starts "UC1: Azure OpenAI - Universal Monitor: Wait for Email (Event Consumer)" and "UC1: Azure OpenAI - Monitor Email from Microsoft Server (Event Producer)"
• STEP 2: Send an e-mail to the address monitored by "UC1: Azure OpenAI - Universal Monitor: Wait for Email (Event Consumer)".
• STEP 3: Review the outputs of (7,8,9) depending on the classification produced by (5).

Expected Results:

• Mail is received, successfully monitored, and a workflow is launched.
• Task "UC1: Azure OpenAI - Email Classification Task" successfully produces a JSON-structured classification with one of the following values: "purchase", "infrastructure", "undefined".
• If the classification label is "infrastructure", Task "UC1: Azure OpenAI - Email Transformation to Incident Task" launches successfully.

Ready-to-use e-mail messages:

Dear Support Team,

I hope you are doing well.

I am writing to kindly ask for your assistance, as I am currently experiencing an issue accessing the VPN. 
Unfortunately, I am unable to connect as expected, and this is preventing me from accessing the systems I normally use for my work.
So far, I have tried the following steps:

- Restarting my computer
- Disconnecting and reconnecting the VPN
- Checking my internet connection

However, the issue still persists.
I would greatly appreciate it if you could please advise me on what steps I should take next, or let me know if you require any additional information from my side.

Thank you very much for your time and support.

Kind regards,
John Doe

Dear Support Team,

I hope you are doing well.

I am writing to kindly request your assistance regarding the purchase of a laptop for a new team member who will be joining us shortly.
Please let me know if there is any information you require from my side.

Kind regards,
John Doe

Hello there,

Big news from CompanyName Cloud Services 👋

If you’re building products that rely on fast, reliable customer communication, we’ve got something you don’t want to miss.

Meet ConnectFlow™

Our all-in-one messaging and automation platform helps teams send SMS, email, and notifications at scale — without worrying about infrastructure, retries, or delivery issues.

Why teams are switching to ConnectFlow™:

⚡ Send messages globally in seconds

🔐 Enterprise-grade security and compliance

🔁 Built-in retries, monitoring, and analytics

🧩 Easy-to-use APIs and no-code options

📈 Designed to grow with you, from MVP to millions of users

Thousands of companies already trust ConnectFlow™ to power password resets, alerts, onboarding messages, and critical notifications — and now it’s your turn.

👉 Get started in minutes
👉 Free trial included
👉 No credit card required

Have questions? Our solutions team is ready to help you design the perfect setup for your use case.

Thanks for building with us,
The CompanyName Team

You’re receiving this email because you signed up for updates from CompanyName.
Prefer fewer emails? You can update your preferences or unsubscribe at any time.

© 2026 CompanyName, Inc. All rights reserved.

You are an assistant that replies in specific format as mentioned in the examples below

[User] 
Tokyo is experiencing a beautiful spring day with clear blue skies and gentle breezes flowing through the streets. The temperature has reached a comfortable 22 degrees, making it perfect for 
outdoor activities. 

[Assistant] 
{ "city": "Tokyo", "temperature": 22}
 

[User] 
New York is facing a harsh winter storm today with heavy snowfall blanketing the entire metropolitan area. Public transportation is experiencing major delays and cancellations. The temperature has plummeted to
a frigid 5 degrees below zero, making it one of the coldest days this season. 

[Assistant] 
{ "city": "New York", "temperature": -5}

[User] 
London is having a typical autumn day with overcast skies and light drizzle throughout the morning hours. The temperature is holding steady at 15 degrees, which is quite typical for this time of year. 

[Assistant] 
{ "city": "London", "temperature": 15}


[User] 
Sydney is experiencing an intense summer heatwave with scorching temperatures breaking records across the region.  Authorities have issued heat warnings advising people to stay hydrated and avoid prolonged sun exposure. The temperature
has soared to an extreme 35 degrees, making it one of the hottest days of the summer. 

{ 
 ...
 "result": {
   "response": {
      "city": "Sydney",
      "temperature": 35
   },
   "metadata": {
   "prompt_tokens": 276,
   "completion_tokens": 14,
   "total_tokens": 290,
   "finish_reason": "stop"
}

You are a helpful assistant providing information strictly in JSON format.

The JSON format should be:

{ "City": "<city_name>", "temperature": <number>}

[User]

Input text is : 
"New York is facing a harsh winter storm today with heavy snowfall blanketing the entire metropolitan area. The streets are covered in thick layers of snow, and visibility is significantly reduced due to the ongoing blizzard 
conditions. Strong winds are creating dangerous wind chills, and residents are advised to stay indoors unless absolutely necessary. Public transportation is experiencing major delays and cancellations. The temperature has plummeted to
a frigid 5 degrees below zero, making it one of the coldest days this season."

{ 
 ...
 "result": {
   "response": {
      "city": "New York",
      "temperature": -5
   },
   "metadata": {
   "prompt_tokens": 276,
   "completion_tokens": 14,
   "total_tokens": 290,
   "finish_reason": "stop"
}

Limitations

You are a helpful assistant replying with a predefined JSON structure

[User]

New York is facing a harsh winter storm today with heavy snowfall blanketing the entire metropolitan area. The streets are covered in thick layers of snow, and visibility is significantly reduced due to the ongoing blizzard
conditions. Strong winds are creating dangerous wind chills, and residents are advised to stay indoors unless absolutely necessary. Public transportation is experiencing major delays and cancellations. The temperature has plummeted to
a frigid 5 degrees below zero, making it one of the coldest days this season.

{
  "type": "object",
  "properties": {
    "city": {
      "type": "string",
      "description": "City Name"
    }, 
    "temperature": {
      "type": "number",
      "description": "Temperature of the City, Number"
    } 
  },
  "required": ["city", "temperature"],
  "additionalProperties": false
}

{ 
 ...
 "result": {
   "response": {
      "city": "New York",
      "temperature": -5
   },
   "metadata": {
   "prompt_tokens": 276,
   "completion_tokens": 14,
   "total_tokens": 290,
   "finish_reason": "stop"
}

Overview

The Azure OpenAI Integration provides configurable data retention controls through the Allow Conversation Data Retention parameter. This section explains how data retention works, what the integration controls, and what additional configuration may be required for your organization's compliance needs.

What the Integration Controls

When the Allow Conversation Data Retention parameter is disabled (default behavior), the integration sets the store parameter to false in all API calls to Azure OpenAI Service and Azure AI Foundry. This disables the following Azure-managed storage features:

• Stored Completions: Input-output pairs are not stored for evaluation or fine-tuning purposes
• Responses API State: Response data is not retained beyond the immediate request/response cycle
  

What the Integration does not control

Stonebranch Universal Automation Center and this integration do not control Azure's abuse monitoring and content filtering systems. By default, Microsoft Azure retains prompts and responses for up to 30 days for abuse monitoring purposes, regardless of the store parameter setting. This retention is managed entirely by Microsoft Azure and is outside the scope of this integration.

Achieving Zero Data Retention

For true zero data retention, your organization must:

• Use this integration with Allow Conversation Data Retention field disabled (sets store=false)
• Obtain Microsoft approval for Modified Abuse Monitoring on your Azure deployment

Modified Abuse Monitoring requires:

• Enterprise Agreement (EA) or Microsoft Customer Agreement (MCA)
• Explicit approval from Microsoft and Request submission via Microsoft's Abuse Monitoring Form

Only when both conditions are met (Allow Conversation Data Retention field disabled + Modified Abuse Monitoring) will your deployment achieve zero data retention.

Important Notes

• Integration Scope: This integration provides controls for Azure's feature-based data storage through the `store` parameter. Microsoft Azure's data retention policies and practices are managed by Microsoft and governed by your Azure service agreement.
• Customer Responsibilities: Customers should review Microsoft's data retention documentation to ensure their Azure configuration meets their compliance requirements. For questions about Azure's data handling or to request Modified Abuse Monitoring, contact Microsoft Azure support.

Official Microsoft Documentation 

• Azure OpenAI Data Privacy
• Azure OpenAI Responses API
• Zero Data Retention Requirements
• Data Retention Q&A (2025)
• Chat Completions and Stored Completions

How To

Import Universal Template

To use the Universal Template, you first must perform the following steps

1. This Universal Task requires the Resolvable Credentials feature. Check that the Resolvable Credentials Permitted system property has been set to true.

2. Import the Universal Template into your Controller:

   1. Extract the zip file, you downloaded from the Integration Hub.

   2. In the Controller UI, select Services > Import Integration Template option.

   3. Browse to the “export” folder under the extracted files for the ZIP file (Name of the file will be unv_tmplt_*.zip) and click Import.

   4. When the file is imported successfully, refresh the Universal Templates list; the Universal Template will appear on the list.

Modifications of this integration, applied by users or customers, before or after import, might affect the supportability of this integration. For more information refer to Integration Modifications.

Configure Universal Task

For a new Universal Task, create a new task, and enter the required input fields.

Integration Modifications

Modifications applied by users or customers, before or after import, might affect the supportability of this integration. The following modifications are discouraged to retain the support level as applied for this integration.

• Python code modifications should not be done.

• Template Modifications

  • General Section

    • "Name", "Extension", "Variable Prefix", and "Icon" should not be changed.

  • Universal Template Details Section

    • "Template Type", "Agent Type", "Send Extension Variables", and "Always Cancel on Force Finish" should not be changed.

  • Result Processing Defaults Section

    • Success and Failure Exit codes should not be changed.

    • Success and Failure Output processing should not be changed.

  • Fields Restriction Section
    The setup of the template does not impose any restrictions. However, concerning the "Exit Code Processing Fields" section.

    1. Success/Failure exit codes need to be respected.

    2. In principle, as STDERR and STDOUT outputs can change in follow-up releases of this integration, they should not be considered as a reliable source for determining the success or failure of a task.

Users and customers are encouraged to report defects, or feature requests at Stonebranch Support Desk.