Skip to content

CLI use

Documents - Find where in the document "a specific term" is mentioned

Benchmark commands (mm bench)

Each benchmark is a real CLI invocation measured end-to-end via subprocess.

Metadata scan (directory-level)

Name Command
mm find . mm find {dir} --format json
mm find . (table) mm find {dir} --format tsv
mm wc . mm wc {dir} --format json
mm sql 'GROUP BY kind' mm sql 'SELECT kind, COUNT(*) ...' --dir {dir} --format json
mm sql 'SUM(size) BY kind' mm sql 'SELECT kind, COUNT(*), SUM(size), AVG(size) ...' --dir {dir} --format json
mm sql 'TOP 10 largest' mm sql 'SELECT name, kind, size ... LIMIT 10' --dir {dir} --format json
mm sql 'GROUP BY ext' mm sql 'SELECT ext, COUNT(*), SUM(size) ...' --dir {dir} --format json
mm find --kind image mm find {dir} --kind image --format json
mm find --kind audio mm find {dir} --kind audio --format json (skipped if no audio)
mm find --kind document mm find {dir} --kind document --format json (skipped if no documents)

Fast mode — content extraction (file-level)

Use --no-cache to bypass caching.

Name Command Selection
mm cat <code> (x20) mm cat {files} --mode fast --format json first 20 code files
mm cat <image> mm cat {file} --mode fast --format json first image
mm cat <image> (x20) mm cat {files} --mode fast --format json first 20 images
mm cat <audio> mm cat {file} --mode fast --format json first audio
mm cat <video> mm cat {file} --mode fast --format json first video
mm cat <pdf> mm cat {file} --mode fast --format json first document
mm cat <pdf> (x10) mm cat {files} --mode fast --format json first 10 documents
mm grep /pattern/ mm grep 'import\|include\|require' {dir} --format json all text files

Accurate mode (LLM pipelines)

Use --no-cache to bypass caching.

Name Command Selection
mm cat <image> --mode fast mm cat {file} --mode fast --format json first image
mm cat <image> --mode accurate mm cat {file} --mode accurate --format json first image
mm cat <video> --mode fast mm cat {file} --mode fast --format json first video
mm cat <audio> --mode fast mm cat {file} --mode fast --format json smallest audio

Placeholders

  • {dir} — resolved absolute directory path
  • {file} — single file auto-picked by kind (first match, or smallest for accurate-mode audio)
  • {files} — space-separated list of files (batch commands, up to N)

Commands are skipped when the required file kind is absent from the target directory.