How to create your Copilot with Azure AI Studio in 30 minutes
Victor Karabedyants
MSDP in Software Engineering, CTO, MBA, Cloud Manager at Sitecore | AI Engineer | Azure Solutions Architect | Azure Administrator | Azure Security Engineer | Azure Developer | Azure Data Engineer and Devops| CKA
Who isn't using ChatGPT right now? Who has yet to hear of OpenAI? Admit it :) I'm sure many have tried using these technologies at least once. But we often lack fresh knowledge in these models, or specific knowledge about our topic, for which we want to communicate with the model, for example, our knowledge base. In this article, we will look at how you can create your model based on ChatGPT technology using the powerful tools of Azure AI Studio.
Azure AI Studio is a powerful integrated environment for developing and deploying artificial intelligence (AI) models and programs on the Microsoft Azure platform.
It provides a wide range of tools and services that simplify the process of creating, training, and deploying AI models, including features such as:
With Azure AI Studio, developers can easily build, train, and deploy various AI models and applications, from classic machine learning algorithms to deep neural networks and natural language processing applications.
In this article, we will take the following steps:
After completing all these steps, we will receive a fully functional application capable of processing requests based on PDF documents using a model and a web interface.
How to start working with AI studio, everything is simple register or log in with a Microsoft account to the site? - https://ai.azure.com/
After entering Azure AI studio, we can create a new project, of course, by clicking on the New project button, the interface is generally very convenient and intuitive.
To create a new project, we will need a new hub.
Hub is a collaboration environment for your team to share your project work, model endpoints, compute, connections, and security settings
Select a subscription, resource group, and project name. At this stage, we do not need Azure search yet
Click to create a project. And we go to the project interface.
For further testing, we need a model that will be used by us, we go to the catalog of models, and... your eyes may run out, there are a lot of them here
Let's start with gpt-4o
Go to the model and click deploy:
After the model is deployed, we can already interact with it by going to the chat, and for example, find out how it is doing:
领英推荐
What is the next step? Let's add our information to be processed by the model, for this, we need to add a data source:
After that, it is necessary to add how exactly we will receive our data. There are several options, these are ordinary files that can be downloaded and blobs, etc. I chose the easiest option - I uploaded my resume:
Now we need to create an Azure AI Search service, you can use the current one or create a new one, you can also choose the name of the index and the machine on which it will be run - run indexing jobs:
A sample diagram of the solution we are building:
We press further and we can watch how our data is indexed and enters the AI search
After Azure AI Studio adds our data, we can try to query the chat for the information we've already added.
Now we can safely deploy our model in the application, and this is done with one button - we click “deploy to a web app”:
We need to choose a subscription, group resource, and hosting plan for our program.
And you can go and enjoy the minimalistic interface:
Conclusions: In this article, we considered the process of creating and deploying a custom model based on OpenAI technology using Azure AI Studio. We started working with the model immediately after its deployment and then added new test documents to test the model's performance with a variety of data. After successful testing, we deployed our application to Azure Web App, which allowed us to make it available for use from a web browser.
Examples of use:
In general, creating and deploying your model in Azure AI Studio opens up wide opportunities for creating intelligent applications and automating processes based on processing text data.
We also have a channel on Azure in Telegram, join the community? - https://t.me/azureuacommunity
Financial Analyst
4 个月Insightful!