What happened?

My initial prompt to Gemini was based on the following bug report, which I was receiving:
"You are a very concise and to the point web developer assistant. You will respond with short, and to the point responses to questions. The developer you are assisting, like to do task 1 step at a time. This is so proper debuging and testing can be performed between test. This ensures when a code chage fails, the actual change is what caused the issue and not some additional change that was made simultainiusly."

after a couple back and forth communications, I got the following warning:

"A code sample in this response was truncated because it exceeded the maximum allowable output. Please use the response carefully. You may also try your question again, selecting a smaller block of code as the context."

So why is this a bug? I am using Gemini CLI in VS Code on a Windows 10 Pro desktop (just fyi), when the message appears, the response is not "truncated". It is totally wiped away! The response, though it flashes on the screen for a microsecond, is totally removed by the warning! This is terible UI experience because none of the resources on Google's side and the LLM itself, in generating the response is salvagable! However, the plot thickens...

I ran in to this bug so many times that I prompted Gemini with:

"Ok, your response was too long! When you response is longer than Google's required limit, your rsponse is not shown to me. Instead a Google warning is displayed that erroniusly states it was truncated. It was not truncated but rather replaced. This is why you must be concise.

Here is Google's warning: "A code sample in this response was truncated because it exceeded the maximum allowable output. Please use the response carefully. You may also try your question again, selecting a smaller block of code as the context.""

This issue has hapening repetedly! At one point Gemini and I had this discusion about it:

"This is a bug I need to report to Google. Is there a feedback / bug report option I can use? If all your responses are eliminated, then their is no way Google can expect this work relationship to function. I got the same message:

A code sample in this response was truncated because it exceeded the maximum allowable output. Please use the response carefully. You may also try your question again, selecting a smaller block of code as the context.

Gemini
I understand your frustration completely. It's impossible for us to work effectively if my responses are being cut off. That "truncated" message is a known issue when the response size exceeds the platform's limit.

To report this, please look for a "Send feedback" or "Report an issue" link or button within the interface you're using. It's often located near the response area or under a help/menu icon. In your report, I recommend including:

The prompt you used.
The exact warning message: "A code sample in this response was truncated..."
A brief explanation that this limitation makes the tool difficult to use for development tasks.
The engineering team does review that feedback, and it's the most direct way to notify them of the issue.

While that gets sorted out, let's continue with our task. I will keep my responses small."

After this point, even really short responses by Gemini, were resulting in the same warning messages. Is this an issue with context window lenght? I though Gemini had a massivly long context window? Anyway, is their a way to close Gemini and start over wit code assistant to discard the old context? Or does Gemini just pick up where it left off, with all past context when you close and reopen VS Code?

Paul - DevManiac

What did you expect to happen?

I expected to SEE the truncated response. And I expected Gemini's very short answers in response to my prompts for such to NOT be overwritten with the same warning.

Client information

Login information

Google Account

Anything else we need to know?

I am available to assist or answer additional questions.