Azure CLU works in Postman but Health Bot fails with clu_result is not defined

Hi Suma Nalluri

Thank you for bringing this to our attention. We are currently checking the details and working on resolving the issue. We’ll update you as soon as we have more information.

Hi Suma Nalluri

Good day. I am trying to replicate the issue.

Could you share contact details in teams to connect on teams call to assess the problem.

Thank you.

Hi Suma Nalluri,

I was trying to replicate the scenario with reference to this documentation.

https://learn.microsoft.com/en-us/azure/health-bot/language_model_conversational-language-understanding

But saw maintenance notice on Language studio dashboard while creating a CLU model for health care Agent (In Maintenance since 27th Feb)

Checking internally with Engineering Team on the situation now.

Thank you.

Hi Suma Nalluri,

Commenting here to un-idle the case on forum.

Attached case summary for reference

Issue -

Runtime issue while accessing finetune CLU model from foundry via Health care agent/ Language models

Fine tune CLU model from instead of Language model from Language studio resource

Current status of the issue-

Resource id and sample files fetched from customer via support ticket to replicate the issue

Changing API version did not help customer

Issue escalated to engineering ticket for addressing.

Additional on 8th March

Language studio resource seems to back from maintenance now.

Trying to test the model from language studio in same region of foundry now to isolate the issue now.

https://learn.microsoft.com/en-us/azure/health-bot/language_model_conversational-language-understanding

Feel free your share your observation with Language model from Language studio as

Thank you for your inputs.

Hi Manas,

Thank you for the update.

I wanted to share the current issue I am facing while integrating the CLU model from Azure Language Studio with the Healthcare Agent.

I am using a CLU model created and deployed from Azure Language Studio and connecting it through Healthcare Intelligence → Language → Models in the Healthcare Agent service. The model connects successfully and the intents are fetched correctly.

However, I am encountering issues with the conversation flow during runtime.

My chatbot architecture currently includes:

• A Basic Greeting scenario that collects the user's name and email.

Multiple intent-specific scenarios such as medicalInformation, productQualityControl, adverseEvent, etc.

The CLU model configured under Language Models with intent mapping.

The problem occurs when the language model runs before the greeting flow completes. For example, when the user enters their name (e.g., "suma"), the CLU model attempts to classify it as one of the intents with low confidence and routes the conversation to one of the mapped scenarios instead of continuing the greeting flow.

To work around this, I also tried invoking the CLU model manually inside the scenario using the healthbot step and a Switch node (cluResponse.result.prediction.topIntent). However, when the language model mapping is enabled, the behavior becomes inconsistent and sometimes causes runtime issues depending on the configuration.

My intended conversation flow is:

The Basic Greeting scenario collects the user name and email.

The bot asks "How may I assist you today?"

The user’s question is then processed by the CLU model.

Based on the detected intent, the appropriate scenario (medicalInformation, productQualityControl, adverseEvent, etc.) should be triggered.

Currently, the CLU model seems to process the input before the greeting flow finishes, which interrupts the conversation logic.

Could you please confirm the recommended integration pattern when using a Language Studio CLU model with Healthcare Agent scenarios? Specifically, whether the recommended approach is:

Using intent mapping directly to scenarios, or

Calling the CLU model inside the scenario and routing with a Switch step.

Please let me know if you need any additional logs, traces, or configuration details from my side.

Thank you for your help.

Best regards, Suma