[Docs Search Modal] IME / CJK input strings captured prematurely before character selection

[jasonatcb]

Component

Docs

Priority

P1 - High (Strongly needed)

SDK Language(s)

No response

Bug Description

When entering CJK (Chinese, Japanese, Korean) text into the documentation search modal, the search input directly captures uncommitted IME composition keystrokes (e.g., ㄊㄨˊㄆㄧㄢˋ) rather than waiting for the user to select and commit the finished candidate characters (e.g., 圖片).

Steps to Reproduce

1. Open the search modal on a Fern-generated documentation site.
2. Switch to a CJK Input Method Editor (such as Traditional Chinese Zhuyin/Bopomofo or Simplified Chinese Pinyin).
3. Type a query using phonetic keys (e.g., type ㄊㄨˊㄆㄧㄢˋ intending to select 圖片).
4. Observe the input field and search results.

Expected Behavior

The search component should ignore input/change events while active IME composition is in progress (or check event.nativeEvent.isComposing === false) and wait for compositionend before executing search queries.

Actual Behavior

The input captures raw phonetic composition characters as they are typed, populating the field with strings like ㄊㄨˊㄆㄧㄢˋ and returning "No results found" before the user can select candidate characters.

Environment

• Doc Platform: Fern Docs
• Browser: Chrome / Safari / Edge
• Input Method: CJK IME (e.g., Zhuyin / Bopomofo, Pinyin)

Versions

Fern CLI version: 5.59.0

Workaround

Copy and paste Chinese character string from other places into the search modal query input box.

Are you interested in contributing a fix?

No