Mia Assistant UI errors on large queries despite successful backend execution

[carrielevy]

What product would you like to leave feedback for?

Console

What Console version are you using?

v14.4.0

Description

We are seeing consistent failures in the Mia Assistant UI when submitting larger or more complex queries, even though the backend execution appears to complete successfully.

Summary of the issue

When users submit broader queries to the Mia Assistant, the UI returns an error message indicating the request failed. However, when inspecting backend logs, the request completes successfully and returns the expected data.

This appears to be a UI handling issue rather than a failure in the MCP server or Mia Assistant execution layer.

Example query

Please give me this information for everyone who's name starts with a letter in the range of A - D

Observed behavior

• The Mia Assistant UI displays an error for the response
• The UI shows placeholder tool call errors instead of rendered content
• No error logs are present for the assistant execution
• Grafana logs show successful responses with full payloads returned from the backend

Expected behavior

• The UI should successfully render the response returned by the assistant
• If payload size or response structure exceeds UI limits, the error should be explicit and actionable
• Backend success should not surface as a UI failure without context

Additional observations

• Smaller or batched queries sometimes work, for example requesting 25 users at a time
• Larger scoped queries fail more consistently
• Logs indicate success from the MCP server and assistant processing
• The failure only appears to occur at the UI rendering or response handling layer

Evidence

• Attached Grafana logs showing successful backend execution
• Attached UI screenshot showing Mia Assistant error and placeholder tool call output

This issue makes it difficult for users to trust the assistant for broader informational queries, even though the backend is clearly capable of handling them. Happy to provide additional logs or repro steps if helpful.

[ThisIsDemetrio]

Hi @carrielevy , thanks for reporting this issue.

We are trying to investigate and replicate the issue. In the meanwhile I'd like to ask you if the correct response is shown completely in the UI (despite the error text showing in the UI) and/or which is the status code revceived in the frontend application. Just to check if there's an actual error in the communication between the FE and the BE (such as an interruption of the stream) or some text that is not correctly recognized by the frontend which causes to show this error.

Thanks,

[carrielevy]

Hello Demetrio,

Unfortunately the full correct response is not shown completely in the UI. Example prompt: "list out all users in this company, what role they have, and when their last login was"

It got about 35 rows of info into the response and then it errored out.

[ThisIsDemetrio]

Hi @carrielevy . Thanks for the info, it looks like - for some reason - an error happens and the streaming is interrupted. I'm assuming you are using OpenAI models. We're going to perform some other internal tests to check if we can replicate this issue and propose a solution.

We will keep you updated in this post. Thanks again for bringing it up.
|D|