Ai agent connected to fabric data agent and search service not answering fabric questions

[chilukurimadhu]

created ai agent that is connected to both fabric data agent and search service.

case 1 : It's working well in play ground and created web app locally.

case 2 : when deployed it to azure web app, its only answering questions of search service not fabric lakehouse tables.

but it worked first but when table connected to lakehouse,when table deleted and readded not working.

things tried :

redeployed web app multiple times and created all permissions. still no luck.
app.py

[amynic]

Hi @chilukurimadhu - thanks for your question - lots of integrating parts here so want to make sure we know what the path of development was

• you mention AI Agent - is this in Azure AI Foundry? Using the Azure AI Foundry Agent Service?
• If you are using the agent service is the Azure AI Search service being used or are you adding files to the agent
• Confirming the 'playground' is the Agent playground in Azure AI Foundry
• sounds like authentication would be a possible cause if other services are working in your hosted web app scenario - what authentication is used for the fabric connector?

[nicholasdbrady]

Hey @chilukurimadhu , since you haven't responded to @amynic in two weeks, have you found a solution or still need help?

[chilukurimadhu]

Hi Amy , Sorry for the late reply, issue not Et resolved.Yes it's a AI agent in Azure foundry.its connected to AI Search service not uploading files. Detailed problem statement: An AI agent was created in Azure AI Foundry, connected to both the Azure AI Search Service and a Fabric Data Agent, with Cosmos DB storing conversation history. It worked correctly in Foundry and in a locally hosted Python web app built in VS Code. However, after deploying the web app to Azure Web Apps, it initially responded correctly from both data sources, but stopped returning responses from the Fabric Data Agent after the Lakehouse table connected to it was recreated. Permissions and access settings have been verified and are not the cause, as the setup worked initially; the issue appears to be related to the Fabric Data Agent connection post–Lakehouse recreation.

…

On Thu, 25 Sept, 2025, 8:10 pm Amy Boyd, ***@***.***> wrote: Hi @chilukurimadhu <https://github.com/chilukurimadhu> - thanks for your question - lots of integrating parts here so want to make sure we know what the path of development was - you mention AI Agent - is this in Azure AI Foundry? Using the Azure AI Foundry Agent Service? - If you are using the agent service is the Azure AI Search service being used or are you adding files to the agent - Confirming the 'playground' is the Agent playground in Azure AI Foundry - sounds like authentication would be a possible cause if other services are working in your hosted web app scenario - what authentication is used for the fabric connector? — Reply to this email directly, view it on GitHub <#174 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJS6B27DRJNNPSHDGUBZOTT3UP5EVAVCNFSM6AAAAACHPW5V2KVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTINJRGIZDGNA> . You are receiving this because you were mentioned.Message ID: <azure-ai-foundry/discussions/repo-discussions/174/comments/14512234@ github.com>

[toolboc]

@chilukurimadhu,

Is the newly create Lakehouse properly connected with new access settings? I understand that this worked initially but did not work after recreating the Lakehouse which woul indicate to me that perhaps the settings are still being inferred for the initial Lakehouse.

[chilukurimadhu]

yes, Its not newly created lakehouse, its recreated table in existing lakehouse.All the setting were fine,it was working fine in the azure foundry portal and also in local python app, its not working when deployed to web app.if the setting's is the issue it shouldn't work in ai agent in azure portal and local app.

the issue is , it's not working when deployed to webapp.

[abhilashjaiswal0110]

Generally the integration issues with the connecting the Fabric Data Agent. Hope you are using the correct workspace and artifacts id of the data agent created. Data Agent can consume semantic model directly (but provide less customizations), however once connected to Lakehouse table by manually selecting it then you can customize the example queries and instructions.

Its working for me at CPS agent level by leveraging Fabric Data Agent, similarly connected the same data agent in multi-agents orchestration in Azure AI Foundry agent. Getting good accurate responses as well.