Open azakharovams opened 2 months ago
After conducting multiple tests, we've determined that the web app functions properly only when using the updated new interface (look), and your model, web app, and resource group are all located within the same region.
I'm having trouble as well. Switched to new interface. Deployment worked. Says web app is running.
Can you please look at this? I guess I could deploy a standard plan. :(
Logs in Azure shown below.
2024-08-30T21:49:47.832464576Z: [ERROR] WARNING:root:No datasource configuration found in the environment -- calls will be made to Azure OpenAI without grounding data. 2024-08-30T21:49:47.878981802Z: [ERROR] WARNING:root:[{'type': 'missing', 'loc': ('service',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}, {'type': 'missing', 'loc': ('index',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}] 2024-08-30T21:49:47.879023904Z: [ERROR] WARNING:root:[{'type': 'missing', 'loc': ('service',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}, {'type': 'missing', 'loc': ('index',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}] 2024-08-30T21:49:47.879032605Z: [ERROR] WARNING:root:No datasource configuration found in the environment -- calls will be made to Azure OpenAI without grounding data. 2024-08-30T21:49:47.879038405Z: [ERROR] WARNING:root:[{'type': 'missing', 'loc': ('service',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}, {'type': 'missing', 'loc': ('index',), 'msg': 'Field required', 'input': {'top_k': '5', 'strictness': '3', 'enable_in_domain': 'false', 'use_semantic_search': 'false'}, 'url': 'https://errors.pydantic.dev/2.8/v/missing'}] 2024-08-30T21:49:48.170732758Z: [ERROR] [2024-08-30 21:49:48 +0000] [73] [INFO] Started server process [73] 2024-08-30T21:49:48.170775561Z: [ERROR] [2024-08-30 21:49:48 +0000] [75] [INFO] Started server process [75] 2024-08-30T21:49:48.170781661Z: [ERROR] [2024-08-30 21:49:48 +0000] [75] [INFO] Waiting for application startup. 2024-08-30T21:49:48.170786161Z: [ERROR] [2024-08-30 21:49:48 +0000] [73] [INFO] Waiting for application startup. 2024-08-30T21:49:48.170791062Z: [ERROR] [2024-08-30 21:49:48 +0000] [75] [INFO] Application startup
Same problem here, only the resource group isn't created in the same region, and I don't think I have permission to change it.
Has anyone debugged the issue with deploying web app, i still have this problem "Web App deployment failed InvalidTemplate: Deployment template validation failed: 'The value for the template parameter 'azureOpenAIEmbeddingEndpoint' at line '1' and column '5094' is not provided. Please see https://aka.ms/arm-create-parameter-file for usage details.'."
I'm having the same issue, very frustrating.
Same issue, tried in both new and old interface and all objects are in the same region.
I am having the same issue with new and old interface.
I am running into the same issue. I tried with various locations, including EastUS or EastUS2, with combinations of RGs in different ones. Looks like something changed since last week as I was able to deploy. I also tried the old/new look with the same issue "azureOpenAIEmbeddingEndpoint is not provided" Can someone from Microsoft comment and get this fixed, please? Thank you
I could not take photos as this was on my work laptop but I was facing the same issue and was able to deploy by doing the following:
with the network tab open, click the deploy button (I was updating an existing web app)
copy the failed PUT request from the network tab
go to postman
add the body (copy the payload from there request in the network tab) update the azureOpenAIEmbeddingEndpoint value:
"azureOpenAIEmbeddingEndpoint" : { "value":"" }
grab your bearer token as well and add it in the auth tab (postman) and fire off the request
I got a 201
I was also able to verify from the environment variables of my Web App that the deployment was indeed updated (was upgrading from gpt3 to 4-mini)
I have all my services in East US except my data source (using Azure AI search) which is in West US
wow! The deployment worked using the postman trick with put/bearer token and modifying the "azureOpenAIEmbeddingEndpoint" : {} to "azureOpenAIEmbeddingEndpoint" : { "value":""} . I got a 201 as well. Thank you so so much.
I tried a similar workaround as follows:
However, I received the following error:
"error":{ "code":"InvalidTemplateDeployment", ... ,"details":[{"code":"ValidationForResourceFailed","message":"Validation failed for a resource. Check 'Error.Details[0]' for more information.","details":[{"code":"SubscriptionIsOverQuotaForSku","message":"This region has quota of 0 instances for your subscription. Try selecting different region or SKU." ...
I'm not sure which part of the settings is causing this issue. Does anyone have any ideas on how to fix or work around this?
Did anyone try today? Is it fixed?
I tried a similar workaround as follows:
- Copy the failed POST request using the "Copy > Copy as cURL (bash)" option in Chrome's developer tools.
- Modify the copied cURL command as described above.
- Open the Azure Portal, launch the Cloud Shell in bash, paste the modified command, and run it.
However, I received the following error:
"error":{ "code":"InvalidTemplateDeployment", ... ,"details":[{"code":"ValidationForResourceFailed","message":"Validation failed for a resource. Check 'Error.Details[0]' for more information.","details":[{"code":"SubscriptionIsOverQuotaForSku","message":"This region has quota of 0 instances for your subscription. Try selecting different region or SKU." ...
I'm not sure which part of the settings is causing this issue. Does anyone have any ideas on how to fix or work around this?
unfortunately I have not seen this before - I would suggest checking the request body is complete
Just checked again and failed with the same message "InvalidTemplate: Deployment template validation failed: 'The value for the template parameter 'azureOpenAIEmbeddingEndpoint' at line '1' and column '8426'"... Last and only time it worked for me was on Aug 30 16:40 UTC but I was not able to repeat deployment since then. Tried several combinations of RG, location, new and earlier Studio UI etc. The working deployment from Aug 30 has these locations: Web app: North Europe RG: North Europe Azure OpenAI: West Europe Azure AI Services: West Europe
Is this error likely to be resolved or is it worth just using the above workarounds going forward? Seems like an unnecessary amount of steps for a new process.
Is this error likely to be resolved or is it worth just using the above workarounds going forward? Seems like an unnecessary amount of steps for a new process.
There are other potential workarounds that bypass the AzureOAI Studio deployment template. I experimented with the RM template based deployment that is invoked with a click on the "Deploy to Azure" icon in the Readme. Unfortunately it has its own bugs: the temperature and top_p are typed int instead of float (mismatching the type in the Python code), and the UI customization is apparently out of scope. A documentation link is broken in the relevant section of the Readme.
Besides, not as straightforward as the template sniffing and rectification approach discussed above. I wonder if the RM template based deployment method is worth fixing or alternatives are recommended, e.g. CLI ?
Is this error likely to be resolved or is it worth just using the above workarounds going forward? Seems like an unnecessary amount of steps for a new process.
I can't imagine they won't fix it - probably they just don't know about it.. Not sure how to even report it lol
https://github.com/microsoft/sample-app-aoai-chatGPT/discussions/1085
If everyone can upvote this discussion it could help lol! need this fixed asap
Hi all, the template validation error in the playground deployment flow is a known issue with a fix underway. Apologies for the inconvenience. In the meantime deploying from this repo with the 'deploy to azure' button or the azure CLI are possible workarounds.
i used the postman trick and got this error ; this is the error for now : The template deployment failed with error: 'Authorization failed for template resource '22a25cd7-ed25-5eff-8929-38a526b663dc' of type 'Microsoft.Authorization/roleAssignments'. The client 'OfekProduction@caesarea.com' with object id '6ee952df-0c71-466c-a8f6-174d3b5a2287' does not have permission to perform action 'Microsoft.Authorization/roleAssignments/write' at scope '/subscriptions/85902c28-eff5-423a-b460-f53331149 .... what permissions i need for this ?
Appears to be fixed.
not fixed for me testing 5 minutes ago in the new Azure AI Studio Env. (All resources set to "West Europe")
Appears to be fixed.
No change on my side, unfortunately. I tried with a few combinations of location, RG etc. My successful deployments in the past 10 days were all independent of the Studio.
I also tried the az webapp CLI based deployment as workaround and it works. Not a direct replacement however, but offers more control.
@jschmHoBa @gaborvar the fix is making its way to all environments. For now, I believe it is fixed in Azure OpenAI studio, if you use the "old view".
Yes, it seems like the bug has been fixed, and the command modification workaround is no longer necessary.
However, I am now directly getting the following error when using the deploy button. This indicates that while the workaround mentioned in my previous post was correct, there are still some issues in my environment.
Any suggestions to fix this?
SubscriptionIsOverQuotaForSku: This region has quota of 0 instances for your subscription. Try selecting different region or SKU.
@jschmHoBa @gaborvar the fix is making its way to all environments. For now, I believe it is fixed in Azure OpenAI studio, if you use the "old view".
It works, thank you! (West Europe location, new OpenAI Studio)
This issue is stale because it has been open for 30 days with no activity.
this can be closed
Describe the bug Impossible to deploy to web app, multiple users confirmed that they have the following error message while trying to deploy: "InvalidTemplate: Deployment template validation failed: 'The value for the template parameter 'azureOpenAIEmbeddingEndpoint' at line '1' and column '4690' is not provided. Please see https://aka.ms/arm-create-parameter-file for usage details.'."
To Reproduce Steps to reproduce the behavior in old or new look:
Expected behavior It should deploy the app but in the old look, I have the error "InvalidTemplate: Deployment template validation failed: 'The value for the template parameter 'azureOpenAIEmbeddingEndpoint' at line '1' and column '4690' is not provided. Please see https://aka.ms/arm-create-parameter-file for usage details.'." In the new look, I also have an error but it's not specified.
Screenshots