Skip to main content

remberg AI Copilot: No Answers are generated

Find suggested solutions here if the AI Copilot does not generate any answers

Written by Katharina Kos

Sometimes the AI Copilot will show an indeterminate load indicator after you ask a question. This guide will help you troubleshoot this problem:

1. Reload the page

Reload the page and try asking the question again. There may be a temporary problem with your network connection.

If the problem persists, continue with this guide.

2. Make sure no firewall is blocking the answer

The AI Copilot "streams" the answer live, which is sometimes blocked by firewalls or other security solutions. In this case, remberg may not know that there is a problem and will not display the final response.

There are possible checks for this problem:

  • Try another internet connection (e.g. public WiFi, smartphone hotspot) to ask a question.

  • Try asking a question from a different, non-company device (e.g., personal smartphone, tablet).

  • Some colleagues with different e.g. permissions or IT setup will get a response

  • You are not receiving live notifications within the remberg (top right).

In this case, security software is probably causing problems and you need to contact your IT department. The following details will help them troubleshoot:

Technical requirements

The remberg AI Copilot uses Server-Sent-Events (https://en.wikipedia.org/wiki/Server-sent_events) to stream the response back to the user. This is a commonly used standard technique and requires a continuous connection back to the client.

Some security solutions (e.g. doing deep package inspection) interrupt or modify this data stream, causing the problem of no response.

While the exact problem varies from vendor to vendor, we recommend an exception in the firewall for the remberg feature.

We do not support manual, non-SSE display of the response for end users.

Exception notes

The SSE are sent from the remberg API which lives at {workspace}.app.remberg.de/api/.

The exact endpoint is e.g. /api/v1/chat, but this may change with updates.

Note that you should exclude the hostname, not the IP, because the underlying services are dynamically allocated within the AWS network.

Did this answer your question?