Azure OpenAI Service

Azure OpenAI Service provides access to Open AI language model that includes GPT-35, GPT-4 and Embedding Model Series. 

With General Availability of GPT-4, the models can now be used but not limited to generate content, summarization, sematic search, and natural language to code translation (hello co-pilot). Accessibility can be achieved through REST API, Python and Azure OpenAI Studio.

In this article we will evaluate the service, how to get started and its usefulness to your environment. The service is subscription based through your Azure subscription and at the publishing of this article it’s billed by the language model you prefer. For the latest pricing, please visit Microsoft Cost Calculator

It’s also important to practice responsible AI, we all have obligation to make the world a safer place.

Sample Use Case: Users Chat Interaction with your document’s library

Saving time and raising productivity is the aim of any business. When your employees need information, being able to get it fast and accurately is important. OpenAI can have a conversation with the employee and answer questions based on what is in your document’s library.

Example, I need to find out what is summer dress code for our organization when I work as a part time employee.

OpenAI will evaluate the question posed by user and get all the relevant information to allow a near accurate answer. To configure the environment to read from your document’s library, we would follow the following steps.

Prerequisites

1. Have an Azure Subscription, azure offers a trial subscription if needed.
2. Fill out the access request form – at the moment the service is available on invite basis.

Configuration

Navigate to Azure Portal

Find Azure Open AI Service

• Create OpenAI Service
• Open the Resource and you should see OpenAI Service
• For our sample we will focus on Chat Playground
  • Getting Started

1. Configuring the Chat Playground
   1. Assistance Setup
      1. Allows you to configure pre-load System Messages and or your Data.
      1. In the System Message, you can configure the questions, the personality and even imaging for your user “fun” part.
   1. Chat Sessions
      1. Allows you to test the chats from Assistance Setup
   1. Configuration
      1. Deployment Process
      1. How long the past messages should show.

You may now go forth and Chatbot.