Avoid activating Windows TSF during dictation start

[aimod-cc]

User description

This avoids activating the OpenLess TSF IME when a dictation session starts.

On Windows, starting dictation previously called prepare_session(), which can activate the OpenLess TSF profile and cause host processes such as Explorer to load OpenLessIme.dll. In some cases Explorer crashes with ntdll.dll / 0xc0000374 around dictation start.

The default Windows insertion path already supports paced Unicode SendInput, so this change skips TSF activation when non-TSF insertion is enabled. TSF is now prepared lazily only when non-TSF fallback is disabled and TSF insertion is explicitly required.

Verified with:

• npm run build
• cargo check
• git diff --check

---

PR Type

Bug fix

---

Description

• Skip TSF activation at dictation start

• Use non-TSF insertion immediately

• Lazily prepare TSF only when needed

• Adjust Windows IME logging and helpers

---

Diagram Walkthrough

flowchart LR
  A["Dictation session starts"] -- "no eager TSF prep" --> B["Non-TSF insertion path"]
  B -- "Unicode SendInput" --> C["Text inserted without TSF"]
  A -- "TSF required later" --> D["Lazy TSF session preparation"]
  D -- "submit through TSF" --> E["Windows IME insertion"]

Loading

File Walkthrough

Relevant files

Bug fix

coordinator.rs Defer Windows TSF preparation until needed                              openless-all/app/src-tauri/src/coordinator.rs Skip TSF activation when non-TSF fallback is allowed. Lazily call prepare_session() only for required TSF inserts. Update fallback logging to reflect actual insertion outcomes. Restrict prepared-session storage helper to tests. +14/-19  dictation.rs Stop preparing TSF on session begin                                            openless-all/app/src-tauri/src/coordinator/dictation.rs Remove eager Windows IME preparation from begin_session(). Prevent dictation start from activating the TSF profile. Keep session initialization focused on state setup. +0/-6

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ No major issues detected