Azure AI Foundry Agent incomplete response from MS Fabric data agent

[Allessandro]

I've created a data agent in MS Foundry, published it. Created an agent in Azure AI Foundry, added sucessfully the corresponding data agent. I am able to play with it in the Playground.

Issue: the Foundry Agent gets the response from the data agent, but the Foundry Agent may not pick the "complete" message/response or rephrase the answer. I am talking about the Foundry Agent rephrasing the answer frequently, or it can also miss few additional rows at the end of the message, trailing part of the data agent run/answer, etc. If I compare the answer in Playgound with the actual run of data agent (via a link that is in the thread in Foundry), some (trailing) content can be missing at all in Playground and agent thread logs.

So far, I use a prompt like ""You run the data agent, and output the complete data/response from the data agent, without any changes and rephasing its content. Wait for the data agent to generate the final response, as it may come in batches asynchronously.
Knowledge Base: Your sole source of truth is the connected data agent: name_of_agent."".
Still experimenting with T=0.2-1, topp=0.9 - 1, and this typically gets the complete final answer from the data agent (tool_call), but still repharses it (data agent's output is being typically rephrased, what I want to avoid). Of course, if still frequently may miss "parts" of the complete answer, prematurely answering with the partial data.

So, the question is, is anyone able to have exact 1-2-1 complete response from data agent in Foundry playground consistently? At what conditions (model, temperature, topp)?

Surely, we can later use the source url, or invoke via API, and get the complete answer from the data agent, but that is not the point - can we do it via pure low-code via UI consistently?