To gain all the benefits of our native integration with OpenAI, there are some foundational tasks that need to be done in your company’s Microsoft Azure environment. This will likely need someone who has the right permissions to operate in there—it might be you as bot builder, or someone within your organisation’s IT team or department.
This guide assumes that you have those permissions, or are being assisted by someone in your organisation who does.
Create a resource
Navigate to https://portal.azure.com, and under Azure services, click on ‘Create a resource’:
Search for OpenAI and pick Azure OpenAI by Microsoft:
Click create, select a subscription and select or create a resource group (Create recommended), then pick region, name and pricing tier.
Select ‘All Networks’—this is required so that our bots can use it
Add any tags (optional), hit Next, then Create, and wait for deployment to complete.
At this point, click on ‘Go to resource’, then go to ‘Azure OpenAI studio’ and click ‘Create new deployment’
Select model: `gpt-35-turbo`, version: `auto-update` and give it a name
Hit `Create`
Go back to Azure Portal, Resource Groups and select the Resource Group created/selected in step 4: https://portal.azure.com/#view/HubsExtension/BrowseResourceGroups
Click `Create` at the top
Search for `Azure AI Search` and select it in the results
Click ‘Create’
Select subscription and the Resource Group (protip: leave as same resource group as selected for OpenAI)
Give it a Service name and Location (protip: same location as OpenAI)
Hit ‘Review + Create’
Review costs and hit ‘Create’
Wait for the deployment to complete
Go back to Resource Groups and select the same Resource Group as last time: https://portal.azure.com/#view/HubsExtension/BrowseResourceGroups
Click ‘Create’
Search for ‘Storage account’ by Microsoft and select
Click ‘Create’
Select Subscription and Resource Group (protip: leaving as pre-selected resource group from earlier step)
Give it a Name, Region and Performance (protip: Standard performance and Geo-redundant storage (GRS)
Click ‘Review’
Click ‘Create’
Wait for the deployment to succeed
Back to Resource Groups and select the same Resource Group as last time: https://portal.azure.com/#view/HubsExtension/BrowseResourceGroups
Click your OpenAI resource
Click ‘Go to Azure OpenAI studio’
Click ‘Chat’ in the side-menu
Click ‘Add your data (preview)’
Click ‘Add a data source’
Select ‘Upload files’ as the data source
Select Azure storage service resource you created earlier
Click ‘Turn on CORS’
Select the Azure AI search resource you created earlier
Select an index name
Tick to acknowledge the cost implications
Hit ‘Next’
Upload your files, then hit ‘Next’
Select ‘Keyword’ as the search type
Hit ‘Save and Close’
Wait for ingestion to complete
Chat to your bot on the right hand panel “Chat Session”
Click View code to see how to integrate/get config options
Accessing your content and data in a future moment
Go to Resource Groups and select the resource group: https://portal.azure.com/#view/HubsExtension/BrowseResourceGroupsClick your OpenAI resource
Click ‘Go to Azure OpenAI studio’
Click ‘Chat’ in the side-menu
Click ‘Add your data (preview)’
Click ‘Add a data source’
Select ‘Azure AI search’ from the dropdown
Select your search service and index
Click ‘Acknowledge usage’ and ‘Next’
Hit ‘Next’
Select ‘Keyword’ as search type
Hit ‘Next’ then ‘Save and Close’
Chat on the right hand panel (their playground sometimes takes a couple of attempts, or a bit of time to hook up despite saying its ready, e.g.
Deleting assets and resources
Assuming everything deployed in the same resource group, simply delete the resource group. Otherwise delete the storage service, azure ai search and open ai resources created, then any resource groups created for this deployment