-
Notifications
You must be signed in to change notification settings - Fork 1.5k
bug: AI asks questions instead of using available tools #3873
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
I see the problem and agree that it makes sense in implementation focused modes, but I think these instructions would conflict with what we want it to do in architect mode for instance. |
can you provide an example? I think the problem that I had was either in "ask" or "architect" mode, it was asking me to tell it which file to look at when it was fully capable of using the tools that are defined for that mode. maybe the solution is to be more generic in the statement and tell it to use tools that are available so if anything is not available (like execute command is unavailable in architect mode) then it will still use everything at it's disposal to get the answers |
what about this
|
Which version of the app are you using?
v3.16.5
Which API Provider are you using?
Google
Which Model are you using?
gemini-2.5
What happened?
When working with files that are not in obvious locations, the AI asks the user for file paths instead of using available tools to find the information itself. This creates unnecessary back-and-forth and puts the burden on the user to provide information that the AI could discover programmatically.
In a recent session, when asked to modify a component file, the AI immediately asked for the file path instead of using search_files or list_files to locate it.
Also tell us, what did you expect to happen?
The AI should automatically use search_files, list_files, or other appropriate tools to locate files or gather information before resorting to asking the user questions.
Steps to reproduce
Suggested System Instructions Fix
This issue can be fixed completely with additional system instructions for the ask_followup_question tool. The following addition would resolve the problem:
This simple addition to the system instructions would ensure the AI exhausts available tools before asking questions, significantly improving the user experience without requiring complex natural language processing of procedural languages.
The text was updated successfully, but these errors were encountered: