Anthropic Claude Code grep tool has a parameter called output_mode that allow the model choose the kind of output is more effective for the task. It support three types "files_with_matches", "content" and "count".
For now, DAIV grep tool only support one output_mode: "files_with_matches". This is ok for the most of the tasks DAIV deal, but some of them could be optimized and save some tokens. For instance: count output mode is useful when a task includes counting occurrences, the llm don't need to read the files to count.
Anthropic Claude Code
greptool has a parameter calledoutput_modethat allow the model choose the kind of output is more effective for the task. It support three types"files_with_matches","content"and"count".For now, DAIV
greptool only support one output_mode:"files_with_matches". This is ok for the most of the tasks DAIV deal, but some of them could be optimized and save some tokens. For instance:countoutput mode is useful when a task includes counting occurrences, the llm don't need to read the files to count.