Spell checking

kvim checks spelling with a bundled English (United States) Hunspell dictionary. Markdown and plain-text buffers are checked as prose. In supported source files, including C and C++, tree-sitter comment and documentation-comment captures are checked. Identifier captures are split at snake_case, camelCase, PascalCase, and digit boundaries; acronyms and components shorter than three characters are ignored. String literals and executable code are excluded. At startup, Kvim uses DECRQSS to probe the live terminal for extended underline support, with Kterm's explicit capability advertisement as a fallback if the bounded probe times out. Misspelled words use an undercurl when the terminal confirms SGR 4:3, otherwise they fall back to a straight underline. The decoration inherits the text foreground, so spelling feedback does not dim or recolor the source text. Existing decorations follow edits while a fresh result is pending, so background checks do not make them flicker. Checks run on a bounded background worker and are skipped for buffers larger than 2 MiB.

Kvim home

The third service indicator in the status line shows spell-check state:

Identifier checking is enabled by default. Disable it without disabling prose and comment checking:

[editor]
spell_check_identifiers = false

Move the cursor onto an unknown word and run one of:

Use <space>ss or :spell-errors to review misspellings across every open buffer owned by the active workspace. The searchable panel groups case-insensitive variants of the same word, shows the number of occurrences, and previews the first location.

When no words are selected, Ctrl-g and Ctrl-w apply to the current row. Successful additions are applied as one batch and spelling results are refreshed once. A failed global dictionary write leaves the in-memory dictionary unchanged.

Global words are stored in kvim's platform state directory. Workspace words are stored in that managed workspace's profile. Personal dictionaries contain words only and never buffer contents.