Mixed-direction rendering fixtures for read-only Lexical consumers

[CodeinScrubs]

Hi — I maintain BidiLens, an MIT toolkit for mixed-direction text, and wanted to share it as a possible fixture resource for applications consuming Lexical content.

I read the discussion in #2610, including the work on manual direction and browser-native behavior. I am not proposing to replace that work or change Lexical's defaults. The remaining distinctions around explicit auto, inheritance and serialization need Lexical-specific decisions.

Our most relevant use case is a read-only preview or AI-generated prose displayed outside the editable surface: Persian paragraphs containing Latin technical names, English paragraphs quoting Persian, punctuation, and direction independent of physical alignment. First-strong is often appropriate; where intent is ambiguous, an explicit author choice is preferable to pretending a heuristic always knows.

BidiLens is not a Lexical editor plugin. Attaching a DOM observer inside contenteditable could interfere with reconciliation, selection, undo and IME, so that is not a recommended integration. Any editor-side change needs those tests plus serialization/collaboration compatibility; our renderer checks do not establish that coverage.

A recent concrete example is our native mixed-direction message documentation contribution, now merged into PersianLabs without a new dependency. It illustrates the kind of small, source-preserving contribution we aim for, not a claim of BidiLens adoption.

If a read-only fixture/example would be useful to Lexical consumers, I would be glad to scope one around your preferred boundary. Otherwise, the existing native direction work may already be the right solution for a given application.