update deployment script, update docker file image, (#73)

* update deployment sku

* add support and chat endpoint to env and checks in notebook

* move chat questions up to the top

* add ai sdk eval notebook for contoso-chat

* add df as input

* Add deployment of pf to provision.sh

* update provision script to westus2 and gpt4t

* remove gpt4turbo

* updating gpt4 to 35 for testing

* revert changes

* fix gpt

* remove github env var

* add register model as promptflow

* update region

* change deployment name

* update deployment logic in provision script

* update endpoint name

* change to dynamic naming

* change names to random numbers

* update chat env to dockerfile

* revert chat deployment file

* split out deployment script

* move deployment script

* remove unused txt file

* add bicep changes for testing

* fix bicep to include metadata for connection

* add custom env create

* fix deployment script

deployment/deployment.sh

@@ -1,43 +1,86 @@
-# get config.json
-echo "Writing config.json file for PromptFlow usage..."
-subscriptionId=$(az account show --query id -o tsv)
-resourceGroupName=$(az group show --name $resourceGroupName --query name -o tsv)
-mlProjectName=$(az deployment group show --name contchat --resource-group $resourceGroupName --query properties.outputs.mlproject_name.value -o tsv)
+ #!/bin/sh
+
+echo "Loading azd .env file from current environment..."
+
+while IFS='=' read -r key value; do
+    value=$(echo "$value" | sed 's/^"//' | sed 's/"$//')
+    export "$key=$value"
+done <<EOF
+$(azd env get-values)
+EOF
 
 # create a random hash for the endpoint name all lowercase letters
 endpointName="contoso-chat-$RANDOM"
 # create a random hash for the deployment name
 deploymentName="contoso-chat-$RANDOM"
 
-echo "{\"subscription_id\": \"$subscriptionId\", \"resource_group\": \"$resourceGroupName\", \"workspace_name\": \"$mlProjectName\"}" > config.json
-$(cat principal.txt) --secret-permissions get list
+az extension add -n ml -y
+
+# Execute the command and parse versions directly, storing the highest version number
+max_version=0
+for version in $(az ml environment list -n $acrRepository --resource-group $AZURE_RESOURCE_GROUP -w $mlproject_name | jq -r '.[].version'); do
+    if [ "$version" -gt "$max_version" ]; then
+        max_version=$version
+    fi
+done
+
+# Increment the highest version number
+next_version=$((max_version + 1))
+
+# Store the next version number in a variable
+new_version=$next_version
+echo "The next version number is: $new_version"
+
+
+az acr login --name $AZURE_CONTAINER_REGISTRY_NAME
+az ml environment create --file deployment/docker/environment.yml --resource-group $AZURE_RESOURCE_GROUP --workspace-name $mlproject_name --version $new_version
+
+# Pause for 600 seconds to allow the environment to be created
+echo "Waiting for environment to be created..."
+sleep 600
+
+#get registry first repository name as variable
+acrRepository=$(az acr repository list --name $AZURE_CONTAINER_REGISTRY_NAME --output tsv --query "[0]")
+# get repository name
+# get envirnment image name from acr
+image_tag=$(az acr repository show-tags --name $AZURE_CONTAINER_REGISTRY_NAME --repository $acrRepository --query "[0]" -o tsv)
+#concatenate the image name
+image_name=$AZURE_CONTAINER_REGISTRY_NAME.azurecr.io/$acrRepository:$image_tag
 
 # register promptflow as model
 echo "Registering PromptFlow as a model in Azure ML..."
-az extension add -n ml -y
-az ml model create --file deployment/chat-model.yaml  -g $resourceGroupName -w $mlProjectName
+az ml model create --file deployment/chat-model.yaml  -g $AZURE_RESOURCE_GROUP -w $mlproject_name
 
 # Deploy prompt flow
 echo "Deploying PromptFlow to Azure ML..."
-az ml online-endpoint create --file deployment/chat-endpoint.yaml -n $endpointName -g $resourceGroupName -w $mlProjectName
+az ml online-endpoint create --file deployment/chat-endpoint.yaml -n $endpointName -g $AZURE_RESOURCE_GROUP -w $mlproject_name
+
+PRT_CONFIG_OVERRIDE=deployment.subscription_id=$AZURE_SUBSCRIPTION_ID,deployment.resource_group=$AZURE_RESOURCE_GROUP,deployment.workspace_name=$mlproject_name,deployment.endpoint_name=$endpointName,deployment.deployment_name=$deploymentName
+sed -i "s/PRT_CONFIG_OVERRIDE:.*/PRT_CONFIG_OVERRIDE: $PRT_CONFIG_OVERRIDE/g" deployment/chat-deployment.yaml
 
 # Setup deployment
 echo "Setting up deployment..."
-az ml online-deployment create --file deployment/chat-deployment.yaml --name $deploymentName --endpoint-name $endpointName --all-traffic -g $resourceGroupName -w $mlProjectName
-az ml online-endpoint show -n $endpointName -g $resourceGroupName -w $mlProjectName
-az ml online-deployment get-logs --name $deploymentName --endpoint-name $endpointName -g $resourceGroupName -w $mlProjectName
+az ml online-deployment create --file deployment/chat-deployment.yaml --name $deploymentName --endpoint-name $endpointName --all-traffic -g $AZURE_RESOURCE_GROUP -w $mlproject_name --set environment.image=$image_name
+az ml online-endpoint show -n $endpointName -g $AZURE_RESOURCE_GROUP -w $mlproject_name
+az ml online-deployment get-logs --name $deploymentName --endpoint-name $endpointName -g $AZURE_RESOURCE_GROUP -w $mlproject_name
 
 # Read endpoint principal
+
 echo "Reading endpoint principal..."
-az ml online-endpoint show -n $endpointName -g $resourceGroupName -w $mlProjectName > endpoint.json
-jq -r '.identity.principal_id' endpoint.json > principal.txt
-echo "Principal is: $(cat principal.txt)"
-
-#Assign Permission to Endpoint Principal
-echo "Assigning permissions to Principal..."
-az role assignment create --assignee $(cat principal.txt) --role "AzureML Data Scientist" --scope "/subscriptions/$subscription_id/resourcegroups/$resourceGroupName/providers/Microsoft.MachineLearningServices/workspaces/$mlProjectName"
-az role assignment create --assignee $(cat principal.txt) --role "Azure Machine Learning Workspace Connection Secrets Reader" --scope "/subscriptions/$subscription_id/resourcegroups/$resourceGroupName/providers/Microsoft.MachineLearningServices/workspaces/$mlProjectName/onlineEndpoints/$endpointName"
-# Get keyValueName from Azure ML
-keyValueName=$(az ml online-endpoint show -n $endpointName -g $resourceGroupName -w $mlProjectName --query "identity.principal_id" -o tsv)          
-echo "assigning permissions to Principal to Key vault.."
-az keyvault set-policy --name $keyValueName --resource-group $resourceGroupName --object-id 
+principal_id=$(az ml online-endpoint show -n $endpointName -g $AZURE_RESOURCE_GROUP -w $mlproject_name --query "identity.principal_id" -o tsv)
+echo "Principal is: $principal_id"
+
+# Assign Permission to Endpoint Principal
+
+echo "Assigning Data Scientist permissions to Principal..."
+az role assignment create --assignee $principal_id --role "AzureML Data Scientist" --scope "subscriptions/$AZURE_SUBSCRIPTION_ID/resourcegroups/$AZURE_RESOURCE_GROUP/providers/Microsoft.MachineLearningServices/workspaces/$mlproject_name"
+echo "Assigning permissions to Principal to Endpoint..."
+az role assignment create --assignee $principal_id --role "Azure Machine Learning Workspace Connection Secrets Reader" --scope "subscriptions/$AZURE_SUBSCRIPTION_ID/resourcegroups/$AZURE_RESOURCE_GROUP/providers/Microsoft.MachineLearningServices/workspaces/$mlproject_name/onlineEndpoints/$endpointName"
+
+# Get Key Vault name
+
+echo "Getting keyValueName from Azure ML..."
+keyValueName=$(az keyvault list --resource-group $AZURE_RESOURCE_GROUP --query "[0].name" -o tsv)
+echo "Key Vault Name is: $keyValueName"
+echo "Assigning permissions to Principal to Key Vault..."
+az keyvault set-policy --name $keyValueName --resource-group $AZURE_RESOURCE_GROUP --object-id $principal_id --key-permissions get list --secret-permissions get list