All work

myMend

Three passes for repairing OCR'd books, with an edit CLI for proofreader agents.

Source on GitLab Python · MPL-2.0 · v0.9.0
$ cat sample.md
# A small book

Words  with spaces , before a comma.

$ uv run --project ~/my/libs/mend clean --no-corpus --rule spacing sample.md
cleaned 1 file(s)
  spacing.double_space               1 edits
  spacing.space_comma                1 edits

$ cat sample.md
# A small book

Words with spaces, before a comma.

I built myMend to turn raw OCR into readable editions and give proofreader agents a small, precise editing tool. Its three commands divide the work: rinse structures an edition, clean repairs recurring errors with a regex pattern library, and proof handles the close reading and edits.

I built proof on inTact so edits have a journal and an undo path. It can compare an edition against its PDF's text layer, queue disagreements for a reader to check, and keep track of which passages the reader has seen.