Skip to content

Inspect and recover

The corpus is a git repository, so everything here is backed by a real commit. See The corpus is a git repository.

What is in there

kennis corpus list                        # every document, with identifiers
kennis corpus list --collection notes     # one collection
kennis corpus tree                        # the directories as they really are

tree is the one to use when you want to see grouping. list is the one to pipe.

What has changed

kennis corpus status

This compares the corpus on disk against what kennis last recorded, so it catches edits you made in your editor. The corpus is ordinary markdown and you are allowed to do that. The next change kennis makes records your edits first, as their own commit named record(corpus), so the history never puts them under the name of an add or a note.

It also reports whether each collection's index is in step with its documents. If it is not, reindex:

kennis corpus index

What happened

kennis corpus history
kennis corpus history --collection docs --limit 50

  48f6d723  2026-09-24 11:52  index(corpus): 48 documents, 334 chunks
  9016f9bf  2026-09-23 19:11  add(literature): 1 added
  0f390d48  2026-09-23 18:34  init(corpus): create the corpus

The short commit in the first column is what restore and return take. Every line is a point the corpus can be returned to.

Put a document back

kennis corpus restore <document-id>                      # from HEAD
kennis corpus restore <document-id> --commit 9016f9bf    # from there

The document returns to its state at that commit keeping its identifier, which is what the identifier being surrogate is for. Anything that referenced it still resolves.

Return the whole corpus to a point

kennis corpus return 9016f9bf
kennis corpus return 9016f9bf --yes    # do not ask

This makes every document as it was at that commit: a document changed since goes back to how it was, one added since is removed, one removed since is put back. It says which first, by title, and asks.

Nothing is lost by it:

  • What is not yet recorded is recorded first, as its own commit, so an edit you made in your editor is in the history before the return replaces it.
  • The return is a commit too. To undo it, return to the commit before it; corpus history lists both.

Two things it leaves alone:

  • The index. Search still reads the index built before the return, and corpus status says which documents it is behind on. Run kennis corpus index afterwards.
  • Installed packs. They are not in the corpus history, so the next kennis corpus sync writes a pack's documents again if the return took them away.

In the window, the Activity page lists the same history. Each line but the newest has Return to this point, which says what would change before it does anything, and a removal lists what it removed, each with Put back.

Rename or regroup

kennis corpus move <handle> --title "A better title"
kennis corpus move <handle> --group radio-astronomy
kennis corpus move <handle> --group ""          # back to the collection root

--title renames the file on disk to match. The identifier does not change, so handles you have written down keep working.

Remove

kennis corpus remove <handle>
kennis corpus remove <handle> --yes    # do not ask

It asks first unless you pass -y. Assets are removed with the document. It is a commit, so it is recoverable through restore at an earlier commit.

In the window, a document's menu in the tree has Remove, which asks first. The Activity page then lists the removal with Put back.

When another write is running

A write that arrives while another runs waits for it and says what it is waiting behind. A write that fails or is stopped part of the way through changes nothing, so it can simply be run again. See The kennis service.

The log

kennis config path      # the config directory

kennis logs every run. When something went wrong, the log already exists - its path is on the second hint: line of any error.