Skip to content

Latest commit

 

History

History
130 lines (88 loc) · 8.03 KB

knownissues.md

File metadata and controls

130 lines (88 loc) · 8.03 KB
Raw

Known Issues

Here are some commonly encountered issues when deploying the PS Info Assistant Accelerator.

This subscription cannot create AzureIAServices until you agree to Responsible AI terms for this resource

Error: This subscription cannot create CognitiveServices until you agree to Responsible AI terms for this resource. You can agree to Responsible AI terms by creating a resource through the Azure Portal then trying again. For more detail go to https://aka.ms/csrainotice"}]

Solution : Manually create a "Azure AI services" in your Azure Subscription and Accept "Responsible AI Notice"

  1. In the Azure portal, navigate to the “Create a resource” page and search for “Azure AI Services”
  2. Select “Azure AI services” from the list of results and click “Create”
  3. On the “Create” page, provide the required information such as Subscription, Resource group, Region, Name, and Pricing tier "Standard S0".
  4. Review and accept the terms "Responsible AI Notice".
  5. Create the Azure AI Service.

IMPORTANT: In some instances, an older subscription that has already had the "Responsible AI Notice" accepted for the old "Cognitive Services multi-service account" may require this process to be repeated for the new "Azure AI Service".

Error "Your administrator has configured the application infoasst_web_access_xxxxx to block users..."

By default Info Assistant deploys the webapp to require users to be a member of an Azure Active Directory Enterprise Application to access the website. If a user is not a member of the AAD EA they will receive this error:

image.png

Solution

Option 1

Add the user to the Azure Active Directory Enterprise Application.

  1. Log into the Azure Portal
  2. Navigate to the App Service object in your resource group, named infoasst-web-xxxxx.
  3. View the Authentication tab. Select the "Identity Provider" link. Image of identity provider link
  4. In the Overview tab, Select the link under the "Essentials" section labeled "Managed application in..." that should have a value like infoasst_web_access_xxxxx. Image of identity provider link
  5. Select the Users and Groups tab and use the Add user/group to add the user to the Azure Active Directory Enterprise Application.

Option 2

Turn off the option to require membership for the Azure Active Directory Enterprise Application.

  1. Log into the Azure Portal
  2. Navigate to the App Service object in your resource group, named infoasst-web-xxxxx.
  3. View the Authentication tab. Select the "Identity Provider" link named infoasst_web_access_xxxxx.
  4. In the Overview tab, Select the link under the "Essentials" section labeled "Managed application in..." that should have a value like infoasst_web_access_xxxxx.
  5. Select the Properties tab. Change the value for Assignment Required to No. Click Save.

Errors due to throttling or overloading Form Recognizer

Occasionally you will hit a 429 return code in the FileFormRecSubmissionPDF which indicates that you need to retry your submission later, or an internal error returned by AI Document Intelligence in the FileFormRecPollingPDF function, which indicates the service has hit internal capacity issues. Both of these situations will occur under heavy load, but the accelerator is designed to back off and retry at a later time, up to a maximum set of retries, which is configurable. These values surface as configuration settings in the Azure function and can be revised there, or they can be updated at deployment in function.bicep, or they can be updated in the file local.settings.json which is used when debugging a function in VS Code. These values are as follows...

@description('The maximum number of seconds  between uploading a file and submitting it to FR')
param maxSecondsHideOnUpload string

@description('The maximum number of times a file can be resubmitted to FR due to throttling or internal FR capacity limitations')
param maxSubmitRequeueCount string

@description('the number of seconds that a message sleeps before we try to poll for FR completion')
param pollQueueSubmitBackoff string

@description('The number of seconds a message sleeps before trying to resubmit due to throttling request from FR')
param pdfSubmitQueueBackoff string

@description('max times we will retry the submission due to throttling or internal errors in FR')
param maxPollingRequeueCount string

@description('number of seconds to delay before trying to resubmit a doc to FR when it reported an internal error')
param submitRequeueHideSeconds string

@description('The number of seconds we will hide a message before trying to repoll due to FR still processing a file. This is the default value that escalates exponentially')
param pollingBackoff string

@description('The maximum number of times we will retry to read a full processed document from FR. Failures in read may be due to network issues downloading the large response')
param maxReadAttempts string

Error : Error due to service unavailability

InvalidTemplateDeployment - The template deployment 'infoasst-myworkspace' is not valid according to the validation procedure. The tracking id is 'XXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXX'. See inner errors for details.
InsufficientQuota - The specified capacity '1' of account deployment is bigger than available capacity '0' for UsageName 'Tokens Per Minute (thousands) - GPT-35-Turbo'.

###Solution

This means that you have exceeded the quota assigned to your deployment for the GPT-35-Turbo model.The quota is the maximum number of tokens per minute (thousands) that you can use with this model. You can check your current quota and usage in the Azure portal. To increase the quota learn more

Error:'OpenAI' is either invalid or unavailable in given region

InvalidTemplateDeployment - The template deployment 'infoasst-myworkspace-xxxxx' is not valid according to the validation procedure. The tracking id is 'xxxxx-xxxx-xxxxx-xxxx-xxxxxxx'. See inner errors for details.
InvalidApiSetId - The account type 'OpenAI' is either invalid or unavailable in given region.

Solution:

Deploy Azure OpenAI Service only in the supported regions. Review the local.env file and update the location as per supported models and region availability

Error: jq parse error: Expected value before ','

If you see a jq parse error while doing deployments, it means one of the makefile scripts to extract environment variables is failing to find a value it expects to be there. The files related would be the main.parameters.json file which is the variables from bicep output from the infrastructure create. The other would be the env file used during build and deploy time

Solution:

To resolve carefully check your deployment .env file for any missing but required values. There are rare times when ARM has issues and output values are not written. In which case simply double check your configuration and rerun the make deploy and/or make extract-env command so that the bicep outputs can be written again

Error:

If you encounter an error similar to the one below that indicates your device must be managed.

ERROR: AADSTS530003: Your device is required to be managed to access this resource.
Trace ID: xxxxxxxx-xxxx-xxxx-xxxxxxxx
Correlation ID: xxxxxxxx-xxxx-xxxx-xxxxxxxx
Timestamp: 2023-10-05 19:54:05Z
Interactive authentication is needed. Please run:
az login --scope https://graph.microsoft.com//.default
make: *** [Makefile:18: infrastructure] Error 1

Solution

You will need to open your GitHub Codespaces in VSCode on your managed device. Please read more about opening your GitHub Codespaces using VSCode.