Is your feature request related to a problem? Please describe.
Stupid open source models often repeat themselves by executing the same tool again and again. This currently requires human intervention every time. Better solution would be to have a separate, lower limit for repeated tool execution which would cause immediate reporting back to the model to select a different strategy if the limit is hit. Human intervention would be required only if the model doesn't take advice and continues repeating tool execution.
Describe the solution you'd like
Separate, lower limit for repeated tool execution which would cause immediate reporting back to the model to select a different strategy
Describe alternatives you've considered
None.
Additional context
Testing with kimi-k2.7code yielded no human intervention required for several days already whereas before it needed hand-holding multiple times a day.
Is your feature request related to a problem? Please describe.
Stupid open source models often repeat themselves by executing the same tool again and again. This currently requires human intervention every time. Better solution would be to have a separate, lower limit for repeated tool execution which would cause immediate reporting back to the model to select a different strategy if the limit is hit. Human intervention would be required only if the model doesn't take advice and continues repeating tool execution.
Describe the solution you'd like
Separate, lower limit for repeated tool execution which would cause immediate reporting back to the model to select a different strategy
Describe alternatives you've considered
None.
Additional context
Testing with kimi-k2.7code yielded no human intervention required for several days already whereas before it needed hand-holding multiple times a day.