feat(cli): improve slash command discovery#3736
Merged
Merged
Conversation
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Collaborator
Author
E2E Test ReportScopeThis E2E pass focused on the user-visible slash command Phase 3 behavior:
Environment
Build used for E2Enpm run build && npm run typecheck && npm run bundleResult: passed. Known existing warnings appeared from Focused tests also runcd packages/cli && npx vitest run src/ui/components/Help.test.tsx src/ui/commands/helpCommand.test.ts
cd packages/cli && npx vitest run src/ui/hooks/useCommandCompletion.test.ts src/ui/utils/highlight.test.ts src/ui/utils/commandUtils.test.ts src/ui/hooks/slashCommandProcessor.test.ts src/ui/AppContainer.test.tsxObserved results: After resolving merge conflicts with latest cd packages/cli && npx vitest run src/ui/components/Help.test.tsx src/ui/commands/helpCommand.test.ts src/ui/hooks/useDialogClose.test.ts src/ui/AppContainer.test.tsxObserved result: TUI E2E scenarios1.
|
Bojun-Vvibe
added a commit
to Bojun-Vvibe/oss-contributions
that referenced
this pull request
Apr 29, 2026
- aaif-goose/goose#8884 blog post: goose with peekaboo (merge-after-nits) - QwenLM/qwen-code#3736 Phase 3 slash command discovery (merge-after-nits) - INDEX.md: drip-170 section with 8 PRs, 1 merge-as-is + 7 merge-after-nits
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Contributor
Code Coverage Summary
CLI Package - Full Text ReportCore Package - Full Text ReportFor detailed HTML reports, please see the 'coverage-reports-22.x-ubuntu-latest' artifact from the main CI run. |
wenshao
reviewed
Apr 30, 2026
|
|
||
| ```bash | ||
| tmux new-session -d -s qwen-slash-phase3 -x 200 -y 50 \ | ||
| "cd /tmp/qwen-slash-phase3 && /Users/mochi/code/qwen-code-test/dist/cli.js --approval-mode yolo" |
Collaborator
There was a problem hiding this comment.
[Suggestion] This E2E command hardcodes a developer-local CLI path, so the verification steps are not reproducible for other contributors. The preceding instructions already build the repo-local bundle and say to use node dist/cli.js; this example should use a portable path derived from the checkout instead of /Users/mochi/....
Suggested change
| "cd /tmp/qwen-slash-phase3 && /Users/mochi/code/qwen-code-test/dist/cli.js --approval-mode yolo" | |
| tmux new-session -d -s qwen-slash-phase3 -x 200 -y 50 \ | |
| "cd /tmp/qwen-slash-phase3 && node $(pwd)/dist/cli.js --approval-mode yolo" | |
| sleep 3 |
— gpt-5.5 via Qwen Code /review
wenshao
requested changes
May 3, 2026
wenshao
left a comment
Collaborator
There was a problem hiding this comment.
Also found 2 additional Critical issues that span whole files and cannot be mapped to individual lines:
[Critical] packages/cli/src/services/commandMetadata.ts — new 142-line service file has zero tests
getCommandSourceBadge, getCommandSourceGroup, formatSupportedModes, getCommandDisplayName, getCommandSubcommandNames and formatCommandSourceLabel are all untested. These functions are used by Help, SuggestionsDisplay and completion hooks, with 7+ branches in getCommandSourceBadge alone. Add packages/cli/src/services/commandMetadata.test.ts covering every exported function and branch.
[Critical] packages/cli/src/ui/utils/commandUtils.ts:247-292 — getBestSlashCommandMatch not directly tested
This function is the core matching engine for mid-input ghost text. Its sort logic is new in this PR, yet commandUtils.test.ts only references it in a comment. Add direct tests covering: recentCommands scoring, completionPriority ordering, argumentHint return path, alias matching, and null return on empty partialCommand.
— deepseek-v4-pro via Qwen Code /review
| ({ | ||
| name: cmd.name, | ||
| description: cmd.description, | ||
| input: cmd.argumentHint ? { hint: cmd.argumentHint } : null, |
Collaborator
There was a problem hiding this comment.
[Critical] This replaces PR #3618's acceptsInput logic with a simple cmd.argumentHint check, causing a functional regression in ACP mode.
Commands with subCommands but no argumentHint (e.g., /memory) will get input: null and auto-submit before the user can choose a subcommand. The same applies to commands with completion functions.
Suggested change
| input: cmd.argumentHint ? { hint: cmd.argumentHint } : null, | |
| input: (() => { | |
| const acceptsInput = | |
| cmd.kind !== CommandKind.BUILT_IN || | |
| cmd.completion != null || | |
| cmd.argumentHint != null || | |
| (cmd.subCommands != null && cmd.subCommands.length > 0); | |
| return acceptsInput | |
| ? { hint: cmd.argumentHint ?? '' } | |
| : null; | |
| })(), |
— deepseek-v4-pro via Qwen Code /review
| }); | ||
|
|
||
| const best = matches[0]; | ||
| if (!best) return null; |
Collaborator
There was a problem hiding this comment.
[Critical] Sort order in getBestSlashCommandMatch prioritizes recentOrder over completionPriority, but compareRankedCommandMatches in useSlashCompletion.ts does the opposite (completionPriority before recentScore).
This causes the mid-input ghost text to recommend a different command than what appears first in the dropdown completion menu, producing an inconsistent UX.
Suggested change
| if (!best) return null; | |
| return ( | |
| (right.completionPriority ?? 0) - (left.completionPriority ?? 0) || | |
| recentOrder || | |
| left.name.localeCompare(right.name) | |
| ); |
— deepseek-v4-pro via Qwen Code /review
Collaborator
Code ReviewOverviewPhase 3 of the slash-command experience overhaul (~2,400 LOC). The PR does four things: (1) reworks Strengths
Issues / SuggestionsCorrectness
UX / consistency
Code quality
Performance
SecurityNothing concerning. All new code is read-side display logic; no shell, FS, or network surface added. The regex changes ( RiskLow. The behavior changes are gated to UI surfaces ( VerdictApprove with the fixes above. The two correctness items worth fixing before merge are the decay-constant mismatch and the missing |
- Fix getBestSlashCommandMatch sort order: completionPriority first, recentOrder second — consistent with compareRankedCommandMatches in useSlashCompletion.ts so ghost text and dropdown agree on best match - Fix findSlashCommandTokens to index altNames into commandMap so alias tokens (e.g. /usage for /stats) are highlighted as valid instead of being marked invalid - Fix getRecentScore decay formula: 10 * Math.max(0, 1 - ageMs / RECENT_DECAY_MS) so the recent boost truly decays to 0 within the 10-minute window named by RECENT_DECAY_MS - Fix Help.tsx CommandsHelp scroll indicator to show command count range (e.g. 1-12/49) instead of raw line count (18/108), which was confusing because each command expands into 2-3 render lines - Fix Help.tsx CommandLine key prop: use stable type:text:index key instead of scrollOffset-index to avoid remounting every line on scroll - Internationalize Help.tsx tab labels via t() instead of hardcoded English strings - Add Tab/Shift+Tab to switch tabs hint in Help footer alongside Esc - Add commandMetadata.test.ts with full branch coverage for all 6 exported functions (getCommandSourceBadge, getCommandSourceGroup, formatSupportedModes, getCommandDisplayName, getCommandSubcommandNames, formatCommandSourceLabel) - Add direct unit tests for getBestSlashCommandMatch covering: null on empty input, null on no match, null for non-modelInvocable commands, completionPriority ordering, recentCommands tie-breaking, argumentHint return path, exact-match exclusion without hint, inclusion with hint - Update Session.test.ts expectation for sendAvailableCommandsUpdate to include _meta field that was added in this PR
Collaborator
Author
|
Thanks for the detailed review! All feedback has been addressed in commit c8331ba. Here's a summary: Critical fixesSort order inconsistency (
Recent decay formula mismatch ( Bug fixScroll indicator shows line count instead of command count ( Code quality
Direct tests for
UXTab label i18n — tab labels now go through Footer hint — footer now shows Test update
|
wenshao
previously requested changes
May 8, 2026
wenshao
left a comment
Collaborator
There was a problem hiding this comment.
测试文件类型错误（6 处 Critical，无法映射到 diff 行）
tsc --noEmit 在以下位置报类型错误——这些是 PR 新增接口变更导致测试 mock 对象未同步更新：
packages/cli/src/ui/components/InputPrompt.test.tsx:184—UseCommandCompletionReturnmock 缺少midInputGhostText属性（TS2741）packages/cli/src/ui/components/InputPrompt.test.tsx:2026—vimModeEnabled不存在于InputPromptProps类型（TS2339）packages/cli/src/ui/components/InputPrompt.test.tsx:2042— 同上（TS2339）packages/cli/src/ui/components/InputPrompt.test.tsx:3480—UIState类型转换缺少isHelpDialogOpen、activeHelpTab等新字段（TS2352）packages/cli/src/ui/hooks/useCommandCompletion.test.ts:49— 找不到命名空间vi，需添加import { vi } from 'vitest'（TS2503）packages/cli/src/ui/hooks/useCommandCompletion.test.ts:65— 同上（TS2503）
测试覆盖缺失
useCommandCompletion.test.ts— 缺少midInputGhostText在只匹配到非modelInvocable命令时返回null的测试useSlashCompletion.test.ts— 缺少recentCommands在非根查询（如/mod）中影响排序的测试
— DeepSeek/deepseek-v4-pro via Qwen Code /review
|
|
||
| try { | ||
| if (commandToExecute) { | ||
| if (!commandToExecute.hidden) { |
Collaborator
There was a problem hiding this comment.
[Critical] setRecentCommands（本次 PR 新增的近期命令追踪逻辑）完全没有测试覆盖。这是整个近期排序功能的基础——如果 count 递增、usedAt 更新或 hidden 命令过滤逻辑出错，补全排序和 ghost text 都会异常。
[Suggestion] setRecentCommands 在 commandToExecute.action() 之前执行（504 行），即使 action 抛出异常，失败的命令也会被记为 recent。建议移到 action 成功返回之后。
— DeepSeek/deepseek-v4-pro via Qwen Code /review
| ); | ||
| } | ||
|
|
||
| export function formatCommandSourceLabel( |
Collaborator
There was a problem hiding this comment.
[Suggestion] formatCommandSourceLabel 已导出并有完整测试，但在整个代码库中无任何调用方，是死代码。建议：接入 SuggestionsDisplay 或 Help 组件，或从本次 PR 中移除。
— DeepSeek/deepseek-v4-pro via Qwen Code /review
| .sort((left, right) => { | ||
| const leftRecent = recentCommands?.get(left.name); | ||
| const rightRecent = recentCommands?.get(right.name); | ||
| const recentOrder = |
Collaborator
There was a problem hiding this comment.
[Suggestion] getBestSlashCommandMatch 与 compareRankedCommandMatches（useSlashCompletion.ts）使用两套不同的排序模型：前者用原始 usedAt 时间戳对比，后者用 count*10 + 10*decay 公式。两者服务于不同场景（ghost text vs 补全菜单）可以理解，但没有注释说明此差异。建议添加注释明确此处为何不同于 compareRankedCommandMatches。
— DeepSeek/deepseek-v4-pro via Qwen Code /review
| type = index === 0 ? 'command' : 'default'; | ||
| } else { | ||
| // Backwards-compatible fallback when no command metadata is provided. | ||
| type = 'command'; |
Collaborator
There was a problem hiding this comment.
[Suggestion] 当 slashCommands 未提供时的回退路径存在行为变更：mid-input 位置的 slash token 现在无条件标记为 command 类型，旧代码无 group 2 因此 mid-input token 从未被高亮。注释标注 "Backwards-compatible fallback" 但实际是新行为。当前生产调用方 InputPrompt.tsx 始终传入 slashCommands，不受影响。
— DeepSeek/deepseek-v4-pro via Qwen Code /review
| [activeTab, onTabChange], | ||
| ); | ||
|
|
||
| useKeypress( |
Collaborator
There was a problem hiding this comment.
[Suggestion] Help 组件（从 83 行扩展到 508 行）缺少错误边界。如果 groupCommands 因异常数据崩溃，用户被锁在 Help 模态框内，唯一恢复方式是 Ctrl+C。建议在 DialogManager 中包裹 Error Boundary，或在 groupCommands 中做 defensive null check。
— DeepSeek/deepseek-v4-pro via Qwen Code /review
- Add test: midInputGhostText is null when only non-modelInvocable commands match - Add test: recentCommands boosts non-root prefix suggestions via recentScore Both tests address coverage gaps identified in code review.
Collaborator
Author
|
已处理 2026-05-08 新增 review 中的剩余两项测试覆盖问题，并已随最新提交推送：1712d946c。\n\n本次补充：\n- ：新增“仅匹配到 non-modelInvocable 命令时 返回 ”测试。\n- ：新增“ 在非根前缀查询下参与排序”测试（使用同等 ，验证 recent 分数生效）。\n\n本地验证：\n- + 共 51 tests 全通过。\n\n之前指出的 6 处类型错误已在更早提交中修复（当前分支 通过）。 |
Collaborator
Author
|
补充说明（上一条评论因 shell 转义导致格式损坏，这里给出完整版本）： 已处理 2026-05-08 新增 review 中的剩余两项测试覆盖问题，并已随最新提交推送：1712d946c。 本次补充：
本地验证：
之前指出的 6 处类型错误已在更早提交中修复（当前分支 |
yiliang114
approved these changes
May 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
/helpinto an interactive tabbed dialog with General, Commands, and Custom Commands views; command lists are scrollable with keyboard navigation._metaand documented the Phase 3 slash command design./help, tab switching, scrolling, and Esc close.Validation
/helpTabDownPageDownTabDownUpPageDownPageUpEsc/help commands/doctor/helpopens a General tab dialog rather than adding a static history item.TabandShift+Tabswitch tabs,Esccloses the dialog, and long command lists scroll with↑/↓andPageUp/PageDown./help commandsdoes not act as a help subcommand and still opens the default General dialog./doctorremains available./helpopens General,Tabswitches to Commands, command list scrolling displaysUse ↑/↓ to scroll (18/105),DownandPageDownscroll, Custom Commands resets scroll position and scrolls independently, andEsccloses the dialog./doctorreturns a Doctor Report with11 passed, 0 warnings, 0 failures.Scope / Risk
/helpsurface now behaves as an interactive dialog, so keyboard routing for Tab, Shift+Tab, arrows, PageUp/PageDown, and Esc is more important than before./helpintentionally has no subcommands;/help commandsis treated like normal/helpand does not directly select a tab./release-notesis intentionally not added.Testing Matrix
Testing matrix notes:
node dist/cli.jsTUI E2E.Linked Issues / Bugs
No linked issues
🤖 Generated with Qwen Code