feat: 支持模糊搜索指令

[gdm257]

Issues

Closes #495
Related #462 #352

TL;DR

• 指令搜索允许 tm → Task Manager、renwu → 任务管理器 这类首字母 / 词首命中，而旧实现只支持首字母。
• 尽量不破坏原有排序设计（偏好、系统应用、指令、频率 etc），可通过设置里的「分词搜索」开关，默认开启。
• 核心思想「顺序匹配」（原有设计）+「分词排序过滤」（参考了 fzf、IDE 语法补全工具）。为最核心的分词、切分、匹配、排序提供了 tests。
• 已在本地测试优化了几天，尽量减少已有文件的改动，并保持兼容性

Screenshots

权重排序

权重	直接分词	编码分词(拼音)	query → item 示例
1000	multi tokens 完整项 ✅	multi tokens 完整项 ✅	task manager → Task Manager / renwuguanliqi → 任务管理器
900	multi tokens 连续词首 ✅	multi tokens 连续词首 ✅	tm → Task Manager / rwgl → 任务管****理器
800	single token 全词 ✅	single token 全词 ✅	task → task manager / guan → 管理器
600	single token 词首子串 ✅	—	tas → task manager
400	multi tokens 非连续词首 ✅	multi tokens 非连续词首 ✅	vc → Visual + Code（跳过 Studio）/ rlq → 任务理****器（跳过）
—	❌ single token 非词首子串、子序列	❌ single token 词首子串、非词首子串、子序列	ana → manager / mgr → manager
—	❌ multi tokens 跨词混合、跨词全词中	❌ multi tokens 跨词混合、跨词全词中	askan → task+manager

• 系统应用略微加权：同档位内 direct+app 比指令略优先，但不会跨档压过更高档位（保留原有的设计）
• 同权重高频高优先级：同档位同类型按使用频率降序排，常用项靠前（保留原有的设计）

MECE 分类

分类	定义	直接分词	编码分词(拼音)	示例(query → item)
multi tokens 完整项	Q = 待匹配项原文(覆盖全部 token,每段完整)	✅	✅	task manager → Task Manager
multi tokens 连续词首	每段=各自 token 连续词首子串,token 相邻	✅	✅	tm → [task]+[manager];tasm → [tas]k+[m]anager;ren w g l → [任][务][管][理]器
single token 全词	Q = 某 token 完整(连续、词首、覆满)	✅	✅	task → [task, manager]
single token 词首子串	Q = 某 token 的连续词首子串(tokenStart=0)	✅	❌	man → [task, manager];m → manager
multi tokens 非连续词首	每段=各自 token 连续词首子串,token 有跳跃	✅	✅	vc → [visual]+[code](跳过 Studio)
single token 非词首子串	Q = 某 token 内连续子串(tokenStart>0)	❌	❌	ana → [task, manager]
single token 子序列	Q = 某 token 内非连续有序字符	❌	❌	mgr → [task, manager]
multi tokens 跨词混合	跨 ≥2 token,含词首段且含词中段	❌	❌	askan → t[ask]+m[an]ager
multi tokens 跨词全词中	跨 ≥2 token,所有段都在词中(无词首段)	❌	❌	skan → ta[sk]+m[an]ager

• 拼音约束更严：音节必须完整，如音节 gu 不会命中音节 guan
• 顺序匹配：query 的每个字符必须按其在 query 中的顺序出现在待匹配项中，不满足顺序则不命中。
• 分词过滤与排序：将命令按 ASCII 边界切成 token（CJK 每个字为单独 token），再依据 query 相对 token 的落点判定匹配模式，「过滤」信息量低的无用模式，并按模式分档「排序」

已知限制 / 预期设计

• query 不支持混合拼音与英文：要么全拼音，要么全英文（或原始 Unicode 字符）。混合编码会在搜索结果中引入其他语言的噪音，且算法复杂、影响性能。当前只支持英文与拼音，若未来加入双拼、五笔、罗马音等编码，混合输入的噪音问题会更严重。
• query 不支持符号：符号未被算作 token。影响很小，例如 7-Zip File Manager 无法用 7-Zip 搜索，用 7z 即可；只有极少数应用（如 Dism++）以符号命名。

[gemini-code-assist]

Code Review

This pull request introduces a new word token search feature (分词搜索) that supports cross-word abbreviation matching. It includes a new tokenizer, a token search engine, a setting toggle in the general settings UI, and updated highlighting and sorting logic. The review feedback highlights several critical improvements: ensuring Vue's reactivity tracks the search setting correctly, resolving a highlighting misalignment bug for mixed pinyin/number queries by including digits in character position mapping, preventing a potential infinite loop in pinyin segmentation when encountering empty syllables, and adding defensive checks in the character classification function to handle undefined inputs.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Issue: Access commandDataStore.wordTokenEnabled directly here to ensure Vue's reactivity system tracks the dependency correctly and re-evaluates the computed property when the setting changes.

Suggested change

	if (wordTokenEnabled) {
	if (commandDataStore.wordTokenEnabled) {

[gemini-code-assist]

🔴 Critical Highlighting Alignment Bug\n\nIssue:\nThe getChineseCharPositions function only extracts the indices of Chinese characters. However, buildPinyinFields in commandDataStore.ts includes digits 0-9 in pinyinTokens to support mixed pinyin/number queries (e.g., "360极速" -> ['3', '6', '0', 'ji', 'su']).\n\nBecause digits are omitted from getChineseCharPositions, the mapped indices for pinyin matches containing numbers will be completely misaligned, leading to incorrect characters being highlighted (e.g., searching "360" on "360极速" highlights "极速" instead of "360").\n\nResolution:\nUpdate getChineseCharPositions to also include digits 0-9 (ASCII range 0x30 to 0x39) so that it aligns perfectly with the characters processed in pinyinTokens.

Suggested change

	function getChineseCharPositions(text: string): number[] {
	const positions: number[] = []
	for (let i = 0; i < text.length; i++) {
	const code = text.charCodeAt(i)
	if (
	(code >= 0x4e00 && code <= 0x9fff) ||
	(code >= 0x3400 && code <= 0x4dbf) ||
	(code >= 0xf900 && code <= 0xfaff)
	) {
	positions.push(i)
	}
	}
	return positions
	}
	function getChineseCharPositions(text: string): number[] {
	const positions: number[] = []
	for (let i = 0; i < text.length; i++) {
	const code = text.charCodeAt(i)
	if (
	(code >= 0x4e00 && code <= 0x9fff) ||
	(code >= 0x3400 && code <= 0x4dbf) ||
	(code >= 0xf900 && code <= 0xfaff) ||
	(code >= 0x30 && code <= 0x39)
	) {
	positions.push(i)
	}
	}
	return positions
	}

[gemini-code-assist]

⚠️ Potential App-Freezing Infinite Loop\n\nIssue:\nIn segmentPinyin, if any element in syllables is empty or falsy, rest.startsWith(syl) will evaluate to true (since any string starts with ""), and qStart will be incremented by syl.length (which is 0). This results in an infinite loop that freezes the entire Electron application.\n\nResolution:\nAdd a defensive check if (!syl) continue at the beginning of the loop to skip empty syllables.

Suggested change

	for (let i = cursor; i < syllables.length; i++) {
	const syl = syllables[i]
	for (let i = cursor; i < syllables.length; i++) {
	const syl = syllables[i]
	if (!syl) continue

[gemini-code-assist]

Issue: If classifyChar is ever called with undefined or an empty string, ch.codePointAt(0) will throw a TypeError: Cannot read properties of undefined (reading 'codePointAt'). Adding a defensive check at the beginning of the function makes it robust against unexpected inputs.

Suggested change

	function classifyChar(ch: string): CharKind {
	if (ch === ' ' || ch === '\t' || ch === '\n' || ch === '\r' || ch === '_' || ch === '-')
	return 'separator'
	function classifyChar(ch: string | undefined): CharKind {
	if (!ch || ch === ' ' || ch === '\\t' || ch === '\\n' || ch === '\\r' || ch === '_' || ch === '-')
	return 'separator'