Azure OpenAI

Using Engini’s Azure OpenAI activities, you can: create, get and update records to manage and define databases.

Getting Started with Azure OpenAI #

Prerequisites #

• A Azure OpenAI account
• Azure OpenAI Key

Connecting Engini to Azure OpenAI #

To connect your Azure services (like OpenAI or Cognitive Services) to Engini, you will need an API Key and an Endpoint. Here is how to find them:

1. Sign in to the Azure Portal – Log in to your account at portal.azure.com.
2. Sign In to Microsoft – Before you can access your resources, you must authenticate

3. Locate your Resource – Once you are logged into the dashboard:

• Use the top search bar to find your specific service (e.g., “Azure OpenAI” or “Cognitive Services”).
• Select your resource from the results list.

1. Retrieve Keys and Endpoint – In the left-hand sidebar, navigate to the Resource Management section.
   • Click on Keys and Endpoint.

• Copy your Key – You will see KEY 1 and KEY 2. Click the “Copy to clipboard” icon next to either one.
• Copy your Endpoint – Copy the URL listed under Endpoint (e.g., https://your-name.openai.azure.com/).

Add a connection to Azure OpenAI in Engini #

1. Enter your Engini account at https://app.engini.io.
2. Navigate to Connections page by clicking on the Connections on the left sidebar or by clicking here.
3. Click on the Add connection option located at the topbar.

4. Choose Azure OpenAI option from the available applications

5. Enter the following details in the “Add Connection” form and press Save:

1. Connection name – Enter a unique and descriptive name for this connection. This name will help you identify and manage the connection in your Engini account. “Azure OpenAI” by default.
2. Url – Enter your Azure OpenAI URL here.
3. Api Key – Enter the access token you obtained from Azure OpenAI.
4. Deployment Name – This field requires the unique name of your Azure OpenAI deployment. You can deploy different types of models, for example.
   • LLM models – GPT models for text generation and chat.
   • DALL·E 3 – for image generation
   • Whisper – for speech-to-text transcription

Make sure the deployment name matches exactly (case-sensitive) the one configured in Azure.

5. API Version – This field requires the Azure OpenAI API version used by your deployment. Enter the exact API version string supported by your Azure OpenAI resource (for example: 2024-02-15-preview).
6. Save Settings – Saves the connection configuration and applies the selected settings.

Note: The Azure OpenAI API URL is deployment-specific. Each connection is tied to a single deployment name. If you want to use multiple deployments (for example, one for text generation and another for image generation), you must create a separate Azure OpenAI connection for each deployment.

Actions #

Create Transcription #

This action provides a setup for defining which model deployment to use and the specific file data to be processed for transcription.

1. Add Field – A functional button that allows users to include additional parameters or custom fields to the configuration if needed.
2. Filename – A text field where you specify the name of the file being transcribed. This helps in identifying and organizing the output in relation to the source file.
3. Content (Base64) – This field requires the actual file content encoded in Base64 format. This allows the audio or document data to be transmitted as a text string for processing.

Note: For transcription tasks, you must use the Whisper model, as it is the only supported model for converting audio to text in this action.

Create Translation #

This action allows users to submit an audio or document file to be translated into English through a specific model deployment.

1. Add Field – A functional option that allows the user to append additional configuration parameters to the action.
2. Filename – A text input where the user enters the name of the source file. This serves as a label to track and identify the specific content being processed.
3. Content (Base64) – This field is for the actual data of the file, converted into a Base64 string. This encoding ensures the file content is transmitted correctly as text for the AI to process.

Note: For translation tasks, the Whisper model is the exclusive requirement, as it is the only model supported for these specific audio translation operations.

Create Chat Completion #

This action is used to generate AI responses based on a provided conversational history or a set of instructions.

1. Messages[] – This field accepts an array of objects (initialized from an object array) representing the conversation history. Each object typically includes a “role” (system, user, or assistant) and the “content” of the message to provide context for the AI’s response.
2. Add Field – A button that allows users to add optional configuration parameters, such as “Temperature” or “Max Tokens,” to fine-tune the model’s output.

Note: For chat completion, you should use large language models (LLMs) such as GPT-4o, GPT-4 Turbo, or GPT-3.5 Turbo.

Generate Image #

This action enables the creation of original images from textual descriptions by utilizing the DALL-E 3 model.

1. Prompt – A text area where you provide a detailed description of the image you want the AI to create. The more specific the instructions, the more accurate the generated visual output will be.
2. Add Field – A button that allows users to include optional parameters, such as image resolution or quality settings, to further customize the output.

Note: To generate images, you must use the DALL-E 3 model, as these are the specialized models for text-to-image tasks.

Initialize Object Array #

The Initialize object array allows users to define and initialize a structured array of objects that can be used as input for subsequent steps, such as conversational history in chat completions.

1. Variable Name – This field is for entering a unique name for the array being created. This name acts as a reference variable that allows you to access this specific set of data later in your workflow.
2. Object Element Type – This dropdown menu allows you to select the specific schema or category for the objects within the array. It ensures that the data structured in the array follows the correct format required by the intended AI model.
3. Add Field – A functional button used to add more specific properties or initial values to the objects within the array during the setup process.

Append to Object Array #

append to object array allows users to add a new data entry to an existing array that was previously created in the workflow, such as an array of chat messages.

• Variable – This dropdown menu allows you to select a pre-existing array created by the Initialize Object Array, once selected, the interface automatically generates input fields based on the specific object type (such as “Messages”) to let you add new data to that array.

Send Api Request #

This interface allows users to perform a Send API Request action within the Azure OpenAI Studio. It is used to manually configure and execute custom HTTP requests to specific Azure OpenAI endpoints for advanced integration or specialized tasks.

1. Base Url – This field displays the primary endpoint address for your Azure OpenAI resource; the resource name is automatically fetched and populated from your connector settings, typically following the format https://{your-resource-name}.openai.azure.com.
2. Relative Url – This field is for entering the specific API path corresponding to the service you want to call, such as completions or embeddings.
3. Method – This dropdown specifies the HTTP verb for the request, with POST being the default for sending data payloads to the model.
4. Body Type – This indicates the format of the data being sent, pre-set to application/json to ensure compatibility with Azure OpenAI’s API requirements.
5. Body – A text area where you enter the raw JSON payload containing the parameters and instructions for the API request.
6. Add Headers – A functional button to include necessary metadata with the request, such as your api-key for authentication.
7. Add Queries – An option to append specific query parameters to the URL to further refine the API call.