Skip to content

ALL-WRIT1001 · comment_language

Keep source comments in one configured project language.

This is a contextual rule for all languages. Read its implementation.

Classify each substantial comment as target, other, or uncertain. The configured target is target_language, which defaults to American English. Judge only the natural-language text supplied by the comment provider. A directive or commented-out source is not prose. Do not guess from a short or ambiguous fragment, a proper name, an identifier, or a code term.

Each finding cites the exact comment record, its source range, model confidence, and model provenance. minimum_characters excludes fragments that language identification models cannot distinguish reliably.

Quoted user text, protocol tokens, names, and required foreign terms can remain when the surrounding explanation uses the project language. Configure another target for a repository whose readers use another language.

With the default target, # Retry after the peer closes the socket is target. A substantial Japanese explanation is other. # Tokio is excluded by the length floor.

  • Cites “GlotLID”, model guidance on confidence and short text. Open reference
  • Cites “Lingua”, language detection for short and mixed-language text. Open reference