Introduce a "dont save requests" mode

[rakotomandimby]

Describe the feature or problem you'd like to solve

For most of the prompt I issue, it uses 1 or 2 requests. I never hit a use case where it uses more.

Proposed solution

Have a /request_level setting that can have the values from 1 to 5, defaulting to 1. 1 is the current behaviour and if 5, it performs muche queries.

Example prompts or workflows

No response

Additional context

No response

[RyanHecht]

Hi @rakotomandimby, I'm not quite sure what you mean by "requests" here -- do you mean Premium requests? Those will always be used at a rate of 1 * modelMultiplier (see modelMultipliers here) for every prompt you submit.

[rakotomandimby]

@RyanHecht hello.

When exiting Copilot CLI, I have a report of used requests.

It is always 1 or 2 (for 1 prompt), never more.

But I have a credit of thousands of requests, so it is OK to use more.

[rakotomandimby]

@RyanHecht hello.

When exiting Copilot CLI, I have a report of used requests.

It is always 1 or 2 (for 1 prompt), never more.

But I have a credit of thousands of requests, so it is OK to use more.

[examon]

This is supported via autopilot mode (/mode autopilot or --autopilot). If your goal is to let the CLI keep working autonomously and use additional requests on a prompt/task, autopilot does that by continuing after each response. You can also cap the extra continuation turns with --max-autopilot-continues <N>. If this doesn't match your intended use case, please feel free to reopen the issue.