input-kanban 0.0.13 → 0.0.15

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/ENVIRONMENT.md CHANGED
@@ -40,7 +40,7 @@ input-kanban \
40
40
  ## Notes
41
41
 
42
42
  - `KANBAN_DEFAULT_WORKSPACE` / `--workspace` should point to the local directory where work should run; `KANBAN_DEFAULT_REPO` / `--repo` remain compatibility aliases.
43
- - `input-kanban serve` starts a lightweight background scheduler that uses the same orchestrator auto-advance path as CLI `submit --auto` / `input-kanban auto <runId>`. It advances planned runs, serial batches, final judge startup, and bounded automatic retries without relying on an open browser tab.
43
+ - `input-kanban serve` starts a lightweight background scheduler that uses the same orchestrator auto-advance path as CLI `submit --auto` / `input-kanban auto <runId>`. It advances planned runs, serial batches, final judge startup, and bounded automatic retries without relying on an open browser tab. If a run was created with `--plan-approval` / plan approval gate enabled, the scheduler stops after planning until the user confirms the plan.
44
44
  - `KANBAN_RUNNER` / `--runner tmux` runs Codex tasks inside tmux windows while keeping scheduling and status tracking in the Node.js orchestrator.
45
45
  - `KANBAN_RUNNER=tmux` is optional. Use it when you want live terminal visibility into planner, worker, and final judge sessions.
46
46
  - With `KANBAN_RUNNER=tmux`, stopping and restarting `input-kanban serve` does not interrupt already-running Codex sessions; tmux keeps them alive and the scheduler resumes after restart. Do not assume the same safety for `headless` runner child processes.
package/PROJECT_GUIDE.md CHANGED
@@ -121,6 +121,7 @@ Supported submit options:
121
121
  --task-file <path|->
122
122
  --max-parallel <n>
123
123
  --worker-sandbox <read-only|workspace-write|danger-full-access>
124
+ --plan-approval
124
125
  --runner <headless|tmux>
125
126
  --runs-dir <path>
126
127
  --auto
@@ -130,7 +131,7 @@ Supported submit options:
130
131
  --poll-ms <ms>
131
132
  ```
132
133
 
133
- `input-kanban submit` creates a run and starts the planner. Task content can come from `--task <text>` or `--task-file <path|->`; omitting `--workspace` uses the current working directory as the target workspace, and `--repo` remains a compatibility alias. Omitting `--label` derives the run label from the first non-empty task line. Auto mode is the default for submit: it keeps polling the run through the shared orchestrator auto-advance path, dispatches batches when the plan is ready, and starts the final judge once all batches complete. `--no-auto` keeps submit to create + plan only. `-d` / `--detach` starts a background supervisor process for the same auto loop and lets the submitting terminal return immediately. The Web server also starts a lightweight scheduler that uses this shared path, so serial batch advancement does not depend on an open browser tab. The submit output includes `input-kanban status <runId> --watch` for terminal-side observation. Because it writes to the same runs directory as the Web server, CLI-created runs are visible in the 8787 dashboard when both processes use the same `--runs-dir`.
134
+ `input-kanban submit` creates a run and starts the planner. Task content can come from `--task <text>` or `--task-file <path|->`; omitting `--workspace` uses the current working directory as the target workspace, and `--repo` remains a compatibility alias. Omitting `--label` derives the run label from the first non-empty task line. Auto mode is the default for submit: it keeps polling the run through the shared orchestrator auto-advance path, dispatches batches when the plan is ready, and starts the final judge once all batches complete. `--plan-approval` adds a durable Planner → Worker gate: auto advances through planning, then pauses at the completed plan until the user confirms it from the Web dashboard by clicking `开始执行`. `--no-auto` keeps submit to create + plan only for the current CLI process, but a running Web server scheduler can still advance the run unless a durable gate such as `--plan-approval` is configured. `-d` / `--detach` starts a background supervisor process for the same auto loop and lets the submitting terminal return immediately. The Web server also starts a lightweight scheduler that uses this shared path, so serial batch advancement does not depend on an open browser tab. The submit output includes `input-kanban status <runId> --watch` for terminal-side observation. Because it writes to the same runs directory as the Web server, CLI-created runs are visible in the 8787 dashboard when both processes use the same `--runs-dir`.
134
135
 
135
136
  Default behavior:
136
137
 
@@ -160,8 +161,8 @@ Key points:
160
161
  - `result` prefers `judge/verdict.json` and falls back to `judge/last_message.md`; `--copy` copies the result to the clipboard.
161
162
  - `stop` requires an explicit `runId` and uses the same stop path as the Web dashboard.
162
163
  - `retry` retries failed/unknown workers while preserving the failed attempt directory.
163
- - `submit` defaults to auto mode: planner -> dispatch -> final judge, with one automatic retry for `batch_blocked` by default. `--no-auto` keeps create + plan only, and `-d/--detach` moves the auto loop to a background supervisor.
164
- - The Web dashboard now follows the same default auto behavior while the page is open: after planning it auto-dispatches planned runs and auto-starts the final judge once all batches complete.
164
+ - `submit` defaults to auto mode: planner -> dispatch -> final judge, with one automatic retry for `batch_blocked` by default. `--plan-approval` changes this to planner -> wait for plan confirmation -> dispatch -> final judge. `--no-auto` keeps create + plan only for the CLI process, and `-d/--detach` moves the auto loop to a background supervisor.
165
+ - The Web server scheduler follows the same shared auto behavior: after planning it auto-dispatches planned runs and auto-starts the final judge once all batches complete, unless a plan approval gate is required and still unapproved.
165
166
 
166
167
  Example agent loop:
167
168
 
package/README.md CHANGED
@@ -56,7 +56,15 @@ input-kanban submit --task "修复登录问题,并补充回归测试" --label
56
56
 
57
57
  `submit` 默认会创建任务批次、发起拆分、自动派发所有批次,并在全部完成后自动发起最终验收。默认 workspace 是当前目录;如果不传 `--label`,任务批次名称会从任务内容自动生成。它使用同一个 runs 目录,所以只要 8787 Web 看板也使用相同的 `--runs-dir`,CLI 创建的任务会在 Web 界面里可见。
58
58
 
59
- `input-kanban serve` 会启动一个轻量后台 scheduler,持续刷新并推进未完成的 runplan ready 后派发 batch、串行 batch 完成后启动下一批、全部 batch 完成后启动 final judge。CLI `submit --auto` / `input-kanban auto <runId>` 与 Web server 共用同一套 orchestrator 自动推进逻辑,因此任务推进不再依赖浏览器页面是否打开或刷新。
59
+ 如果希望 Planner 拆分完成后先看一眼计划,再放行执行,可以加计划确认 Gate
60
+
61
+ ```bash
62
+ input-kanban submit --task-file task.md --plan-approval
63
+ ```
64
+
65
+ 开启后,auto 会推进到 Planner 完成并停在“已拆分,待确认”;用户在 Web 上点一次“开始执行”即确认计划,随后 worker 批次和 final judge 继续自动推进。
66
+
67
+ `input-kanban serve` 会启动一个轻量后台 scheduler,持续刷新并推进未完成的 run:plan ready 后派发 batch、串行 batch 完成后启动下一批、全部 batch 完成后启动 final judge。CLI `submit --auto` / `input-kanban auto <runId>` 与 Web server 共用同一套 orchestrator 自动推进逻辑,因此任务推进不再依赖浏览器页面是否打开或刷新。若 run 配置了计划确认 Gate,auto/scheduler 会停在该 Gate,不会越过未确认的计划。
60
68
 
61
69
  如果希望提交后立即返回,让任务在后台自动执行,可以加 `-d` / `--detach`:
62
70
 
package/RELEASE_NOTES.md CHANGED
@@ -1,5 +1,41 @@
1
1
  # Release Notes
2
2
 
3
+ ## v0.0.15
4
+
5
+ ### Highlights
6
+
7
+ - Absorb PR #3 recovery hardening without directly merging its older `v0.0.13` branch, preserving the `v0.0.14` Plan Approval Gate and Web layout changes.
8
+ - Route blocked-run dashboard execution actions to `/api/runs/:id/retry`, so `batch_blocked` runs retry failed/unknown tasks instead of hitting the dispatch endpoint.
9
+ - Remove stale Web `workers_completed` / `workers_failed` UI state handling now that backend run status uses `batches_completed` and `batch_blocked`.
10
+ - Harden final judge starts: reject archived/stopped runs, duplicate running judges, and completed judges; failed judges are archived to `judge_attempts/` before retrying.
11
+ - Add short `/api/codex` detection caching to avoid repeatedly spawning Codex detection during frequent dashboard refreshes.
12
+ - Keep task-table Codex session IDs and their copy buttons on one line by widening the session column and using a compact inline layout.
13
+
14
+ ### Verification
15
+
16
+ - `npm run check` passed locally with 84 tests.
17
+ - `npm pack --dry-run` passed for `input-kanban@0.0.15`.
18
+ - Windows release-candidate validation on `zhangxing_win` passed with 84 tests.
19
+ - Windows Web smoke confirmed `/api/health`, `/api/codex`, Plan Approval UI, compact header copy tools, and one-line Codex session copy layout.
20
+
21
+ ## v0.0.14
22
+
23
+ ### Highlights
24
+
25
+ - Add a durable Plan Approval Gate: runs can now pause after planning until the generated plan is manually confirmed.
26
+ - Add `input-kanban submit --plan-approval`, which lets auto advance through planning and then stop at the unapproved plan gate instead of dispatching workers immediately.
27
+ - Make `dispatchRun()` confirm the plan gate before starting workers, so Web `开始执行` means “approve this plan and continue execution.”
28
+ - Clarify auto semantics in docs: auto advances to completion, failure, or the first unapproved gate; `--no-auto` is not a durable scheduler gate.
29
+ - Update the Web create form wording to `计划生成后手动确认后执行` and show planned gated runs as `已拆分,待确认`.
30
+ - Rework the run detail header layout: keep title/status on the left, move `Run ID ⧉` and `tmux ⧉` copy tools to a lightweight right-side tool group, and keep long unreadable IDs out of metadata chips.
31
+ - Record the Agent Profile / Candidate / Reviewer design direction in the private KB while keeping the current executor Codex-only.
32
+
33
+ ### Verification
34
+
35
+ - `npm run check` passed locally with 78 tests.
36
+ - `npm run check` passed on the remote Windows validation host `zhangxing_win` with 78 tests for the release-candidate working tree.
37
+ - Windows Web smoke confirmed the plan approval input, compact title copy tools, `/api/health`, and `/api/codex`.
38
+
3
39
  ## v0.0.13
4
40
 
5
41
  ### Highlights
@@ -171,7 +171,7 @@ function parseSubmitArgs(argv) {
171
171
  const args = {
172
172
  host: '127.0.0.1', port: 8787, workspace: undefined, repo: undefined, runsDir: undefined, codexBin: undefined,
173
173
  runner: undefined, label: undefined, taskText: undefined, taskFile: undefined, maxParallel: 3,
174
- workerSandbox: 'workspace-write', auto: true, detach: false, watch: true, json: false, pollMs: 3000, maxRetries: 1, help: false
174
+ workerSandbox: 'workspace-write', planApproval: false, auto: true, detach: false, watch: true, json: false, pollMs: 3000, maxRetries: 1, help: false
175
175
  };
176
176
  for (let i = 0; i < argv.length; i++) {
177
177
  const arg = argv[i];
@@ -190,6 +190,7 @@ function parseSubmitArgs(argv) {
190
190
  else if (arg === '--task-file') args.taskFile = next();
191
191
  else if (arg === '--max-parallel') args.maxParallel = Number(next());
192
192
  else if (arg === '--worker-sandbox') args.workerSandbox = validateSandbox(next(), '--worker-sandbox');
193
+ else if (arg === '--plan-approval') args.planApproval = true;
193
194
  else if (arg === '--auto') { args.auto = true; args.watch = true; }
194
195
  else if (arg === '--no-auto') { args.auto = false; args.watch = false; }
195
196
  else if (arg === '--detach' || arg === '-d') args.detach = true;
@@ -290,6 +291,7 @@ Submit options:
290
291
  --task-file <path> Read task description from file, use - for stdin
291
292
  --max-parallel <n> Default max parallel workers, default 3
292
293
  --worker-sandbox <mode> read-only, workspace-write, or danger-full-access
294
+ --plan-approval Pause after planning until the generated plan is confirmed
293
295
  --runner <mode> Runner mode: headless or tmux
294
296
  --runs-dir <path> Runtime runs directory shared with the Web UI
295
297
  --auto Plan, dispatch all batches, judge, and watch, default for submit
@@ -318,6 +320,7 @@ Options:
318
320
  --task-file <path> Read task description from file, use - for stdin
319
321
  --max-parallel <n> Default max parallel workers, default 3
320
322
  --worker-sandbox <mode> read-only, workspace-write, or danger-full-access
323
+ --plan-approval Pause after planning until the generated plan is confirmed
321
324
  --runner <mode> Runner mode: headless or tmux
322
325
  --runs-dir <path> Runtime runs directory shared with the Web UI
323
326
  --auto Plan, dispatch all batches, judge, and watch, default for submit
@@ -735,7 +738,8 @@ async function submit(args) {
735
738
  workspace: process.env.KANBAN_DEFAULT_WORKSPACE || process.env.KANBAN_DEFAULT_REPO,
736
739
  repo: process.env.KANBAN_DEFAULT_REPO,
737
740
  maxParallel: args.maxParallel,
738
- workerSandbox: args.workerSandbox
741
+ workerSandbox: args.workerSandbox,
742
+ planApproval: args.planApproval
739
743
  });
740
744
  if (!args.json) {
741
745
  console.log(`已创建任务批次: ${state.runId}`);
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "input-kanban",
3
- "version": "0.0.13",
3
+ "version": "0.0.15",
4
4
  "type": "module",
5
5
  "bin": {
6
6
  "input-kanban": "bin/input-kanban.js"
package/public/index.html CHANGED
@@ -41,20 +41,20 @@
41
41
  @media (prefers-reduced-motion: reduce) { button, button.state-active, button.state-pending::after, .refresh-pulse-chip.pulse .refresh-pulse-dot { animation: none !important; transition: none !important; } }
42
42
  table { width: 100%; border-collapse: collapse; font-size: 13px; table-layout: fixed; }
43
43
  th, td { border-bottom: 1px solid var(--line); padding: 9px 8px; text-align: left; vertical-align: top; }
44
- th:nth-child(1), td:nth-child(1) { width: 34%; }
44
+ th:nth-child(1), td:nth-child(1) { width: 32%; }
45
45
  th:nth-child(2), td:nth-child(2) { width: 92px; white-space: nowrap; }
46
46
  th:nth-child(3), td:nth-child(3) { width: 132px; }
47
47
  th:nth-child(4), td:nth-child(4) { width: 64px; white-space: nowrap; }
48
48
  th:nth-child(5), td:nth-child(5) { width: 96px; }
49
- th:nth-child(6), td:nth-child(6) { width: 92px; }
49
+ th:nth-child(6), td:nth-child(6) { width: 116px; }
50
50
  th:nth-child(7), td:nth-child(7) { width: 66px; }
51
51
  th:nth-child(8), td:nth-child(8) { width: 94px; }
52
52
  th { color: #cbd5e1; font-size: 12px; text-transform: uppercase; letter-spacing: .06em; }
53
53
  tr:hover { background: #162033; cursor: pointer; }
54
54
  .pill { display: inline-block; padding: 3px 8px; border-radius: 999px; font-size: 12px; font-weight: 800; background: var(--gray); line-height: 1.3; }
55
- .completed, .judged, .workers_completed, .planned, .batches_completed { background: var(--green); }
55
+ .completed, .judged, .planned, .batches_completed { background: var(--green); }
56
56
  .running, .planning, .judging { background: var(--blue); }
57
- .failed, .workers_failed, .plan_failed, .judge_failed, .unknown, .batch_blocked { background: var(--red); }
57
+ .failed, .plan_failed, .judge_failed, .unknown, .batch_blocked { background: var(--red); }
58
58
  .plan_empty { background: var(--orange); }
59
59
  .pending, .created { background: var(--gray); }
60
60
  .batch-row td { border-top: 3px solid var(--line-strong); background: #101827; font-weight: 800; font-size: 14px; padding-top: 13px; }
@@ -91,8 +91,11 @@
91
91
  .run-card-meta .meta-chip { padding: 4px 7px; font-size: 11px; }
92
92
  .run-card-meta .meta-chip.long .meta-value { max-width: 248px; }
93
93
  .build-header { border: 1px solid var(--line); border-radius: 12px; padding: 14px; background: var(--panel-2); margin-bottom: 14px; }
94
- .build-title { display: flex; align-items: center; gap: 10px; font-size: 22px; font-weight: 900; }
95
- .build-meta { margin-top: 10px; display: flex; flex-wrap: wrap; gap: 7px; align-items: center; }
94
+ .build-title { display: grid; grid-template-columns: minmax(0, 1fr) auto; gap: 10px; align-items: center; font-size: 22px; font-weight: 900; }
95
+ .build-title-main { min-width: 0; display: flex; align-items: center; gap: 10px; }
96
+ .build-title-text { min-width: 0; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
97
+ .build-title-tools { display: inline-flex; align-items: center; gap: 6px; justify-self: end; }
98
+ .build-meta { margin-top: 12px; display: flex; flex-wrap: wrap; gap: 7px; align-items: center; }
96
99
  .meta-chip { display: inline-flex; align-items: center; gap: 6px; max-width: 100%; padding: 5px 8px; border: 1px solid var(--line); border-radius: 999px; background: #020617; color: #cbd5e1; font-size: 12px; line-height: 1.2; }
97
100
  .meta-chip.danger { border-color: #f97316; color: #fed7aa; background: rgba(180,83,9,.14); }
98
101
  .meta-label { color: var(--muted); font-weight: 800; white-space: nowrap; }
@@ -106,12 +109,17 @@
106
109
  .log-panel { margin-top: 16px; }
107
110
  .file-tabs button { font-size: 13px; }
108
111
  .copy-btn { padding: 2px 6px; margin: 0 0 0 6px; border-radius: 6px; font-size: 12px; line-height: 1.2; background: var(--gray); vertical-align: middle; }
112
+ .title-copy-btn { display: inline-flex; align-items: center; gap: 5px; opacity: .78; font-size: 12px; padding: 5px 8px; margin: 0; }
113
+ .title-copy-btn:hover, .title-copy-btn:focus { opacity: 1; }
114
+ .title-copy-btn .icon-svg { width: 12px; height: 12px; }
109
115
  .rename-btn { opacity: 0; pointer-events: none; transition: opacity .15s ease; }
110
116
  .run-card:hover .rename-btn, .run-card:focus-within .rename-btn, .build-title:hover .rename-btn, .build-title:focus-within .rename-btn, .rename-btn:focus { opacity: 1; pointer-events: auto; }
111
117
  .run-card-title-actions { display: inline-flex; align-items: center; gap: 3px; }
112
118
  .archive-confirm-btn { min-width: 46px; padding: 4px 10px; border-color: rgba(248,113,113,.85); background: var(--red) !important; color: white; font-weight: 900; }
113
119
  .icon-svg { width: 14px; height: 14px; display: block; }
114
- .session-cell { word-break: break-all; }
120
+ .session-cell-wrap { display: inline-flex; align-items: center; gap: 5px; white-space: nowrap; }
121
+ .session-cell { min-width: 0; font-variant-numeric: tabular-nums; }
122
+ .session-cell-wrap .copy-btn { flex: 0 0 auto; margin-left: 0; }
115
123
  .row-actions { display: flex; align-items: center; justify-content: flex-end; gap: 6px; }
116
124
  .row-actions .danger, .row-actions .secondary { margin: 0; }
117
125
  .status-stack { display: inline-flex; flex-direction: column; align-items: flex-start; gap: 5px; }
@@ -169,6 +177,8 @@
169
177
  <option value="danger-full-access">danger-full-access(高风险,跳过沙箱限制)</option>
170
178
  </select>
171
179
  <div class="muted">仅影响 worker;任务拆分和汇总验收仍保持 read-only。若执行过程提示 Permission denied / sandbox denied,这通常不是任务本身失败,而是当前沙箱能力不足;在可信工作区可改用 danger-full-access。DNS / 网络失败则通常需要检查代理、VPN 或本地 evidence。</div>
180
+ <label><input id="planApproval" type="checkbox" style="width:auto; margin-right:6px;" />计划生成后手动确认后执行</label>
181
+ <div class="muted">开启后会停在“已拆分,待确认”;点击“开始执行”后继续自动执行到验收。</div>
172
182
  <label>任务说明</label><textarea id="taskText" placeholder="粘贴任务说明"></textarea>
173
183
  <div class="toolbar">
174
184
  <button onclick="createRun()">创建批次</button>
@@ -265,7 +275,6 @@ function userFacingErrorMessage(error) {
265
275
  const statusText = {
266
276
  created: '已创建', pending: '等待中', planning: '拆分中', planned: '已拆分',
267
277
  running: '执行中', completed: '已完成', failed: '失败', unknown: '未知',
268
- workers_completed: '子任务完成', workers_failed: '子任务失败',
269
278
  batches_completed: '批次完成', batch_blocked: '批次阻塞', plan_empty: '拆分为空', stopped: '已停止',
270
279
  judging: '验收中', judged: '已验收', plan_failed: '拆分失败', judge_failed: '验收失败'
271
280
  };
@@ -273,6 +282,9 @@ const roleText = { planner: '任务拆分', judge: '汇总验收' };
273
282
  function displayStatus(s) { return statusText[s] || s || '-'; }
274
283
  function displayRole(s) { return roleText[s] || s || '-'; }
275
284
  function pill(s) { return `<span class="pill ${s || ''}">${displayStatus(s)}</span>`; }
285
+ function planApprovalPending(state = currentState) { return !!(state?.gates?.planApproval?.required && !state.gates.planApproval.approved); }
286
+ function runStatusLabel(state = currentState) { return planApprovalPending(state) ? '已拆分,待确认' : displayStatus(state?.status); }
287
+ function statusPill(state = currentState) { return `<span class="pill ${state?.status || ''}">${esc(runStatusLabel(state))}</span>`; }
276
288
  function esc(s) { return String(s ?? '').replace(/[&<>]/g, c => ({'&':'&amp;','<':'&lt;','>':'&gt;'}[c])); }
277
289
  function formatDateTime(s) {
278
290
  if (!s) return '-';
@@ -314,6 +326,12 @@ function gitChip() { return '<span class="meta-chip" title="Git 工作区"><span
314
326
  function editIcon() {
315
327
  return '<svg class="icon-svg" viewBox="0 0 24 24" focusable="false" aria-hidden="true"><path d="M4 16.5V20h3.5L18.1 9.4l-3.5-3.5L4 16.5Z" fill="currentColor"/><path d="m16 4.5 3.5 3.5" stroke="currentColor" stroke-width="2" stroke-linecap="round"/></svg>';
316
328
  }
329
+ function copyIcon() {
330
+ return '<svg class="icon-svg" viewBox="0 0 24 24" focusable="false" aria-hidden="true"><path d="M8 8h11v11H8V8Z" fill="none" stroke="currentColor" stroke-width="2"/><path d="M5 16H4V4h12v1" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round"/></svg>';
331
+ }
332
+ function titleCopyLabel(kind) {
333
+ return kind === 'tmux' ? `<span>tmux</span>${copyIcon()}` : `<span>Run ID</span>${copyIcon()}`;
334
+ }
317
335
  function archiveIcon() {
318
336
  return '<svg class="icon-svg" viewBox="0 0 24 24" focusable="false" aria-hidden="true"><path d="M4 5h16v4H4V5Z" fill="currentColor"/><path d="M6 10h12v9H6v-9Z" fill="currentColor" opacity=".72"/><path d="M9 13h6" stroke="#020617" stroke-width="2" stroke-linecap="round"/></svg>';
319
337
  }
@@ -407,7 +425,7 @@ function initializeWorkerSandboxPreference() {
407
425
  }
408
426
  async function createRun() {
409
427
  saveWorkerSandboxPreference();
410
- const body = { label: label.value, workspace: repo.value, repo: repo.value, maxParallel: maxParallel.value, workerSandbox: workerSandbox.value, taskText: taskText.value };
428
+ const body = { label: label.value, workspace: repo.value, repo: repo.value, maxParallel: maxParallel.value, workerSandbox: workerSandbox.value, planApproval: planApproval.checked, taskText: taskText.value };
411
429
  const r = await api('/api/runs', { method: 'POST', body: JSON.stringify(body) });
412
430
  selectedRun = r.runId; selectedTask = null; selectedFileName = null;
413
431
  clearFileView();
@@ -444,7 +462,7 @@ function renderRunList() {
444
462
  const visibleRuns = latestRuns.slice(0, runListVisibleCount);
445
463
  const cards = visibleRuns.map(r => `
446
464
  <div class="run-card ${selectedRun === r.runId ? 'active' : ''}" onclick="selectRun('${r.runId}')" onmouseleave="clearArchiveConfirm('${r.runId}')">
447
- <div class="run-card-title"><span class="run-card-name-wrap"><span class="run-card-name" title="${esc(r.label)}">${esc(r.label)}</span><span class="run-card-title-actions"><button class="secondary copy-btn rename-btn" title="修改任务批次名称" onclick="renameRunLabel(event, '${r.runId}')">${editIcon()}</button><button class="secondary copy-btn rename-btn ${pendingArchiveRunId === r.runId ? 'archive-confirm-btn' : ''}" title="${pendingArchiveRunId === r.runId ? '再次点击确认归档' : '归档任务批次(运行中请先停止)'}" onclick="archiveRunFromCard(event, '${r.runId}')">${pendingArchiveRunId === r.runId ? '确认' : archiveIcon()}</button></span></span><span>${pill(r.status)}</span></div>
465
+ <div class="run-card-title"><span class="run-card-name-wrap"><span class="run-card-name" title="${esc(r.label)}">${esc(r.label)}</span><span class="run-card-title-actions"><button class="secondary copy-btn rename-btn" title="修改任务批次名称" onclick="renameRunLabel(event, '${r.runId}')">${editIcon()}</button><button class="secondary copy-btn rename-btn ${pendingArchiveRunId === r.runId ? 'archive-confirm-btn' : ''}" title="${pendingArchiveRunId === r.runId ? '再次点击确认归档' : '归档任务批次(运行中请先停止)'}" onclick="archiveRunFromCard(event, '${r.runId}')">${pendingArchiveRunId === r.runId ? '确认' : archiveIcon()}</button></span></span><span>${statusPill(r)}</span></div>
448
466
  <div class="run-card-meta">
449
467
  ${metaChip('工作区', basenamePath(r.workspacePath || r.repo), { title: r.workspacePath || r.repo, long: true, extra: `<button class="secondary copy-btn" title="复制工作区地址" onclick="copyRunRepoPath(event, '${r.runId}')">⧉</button>` })}
450
468
  ${r.git?.isGit ? gitChip() : ''}
@@ -522,8 +540,11 @@ async function refreshSelected({auto=false} = {}) {
522
540
  function renderSelectedHeader() {
523
541
  if (!currentState) return '<div class="muted">未选择任务批次</div>';
524
542
  const sandbox = currentState.workerSandbox || 'workspace-write';
543
+ const titleActions = [`<button class="secondary copy-btn title-copy-btn" title="复制 Run ID:${esc(currentState.runId)}" data-copy-kind="run-id" onclick="copyRunId(event)">${titleCopyLabel('run-id')}</button>`];
544
+ if (currentState.runner === 'tmux' && hasRunTmuxMetadata(currentState)) {
545
+ titleActions.push(`<button class="secondary copy-btn title-copy-btn" title="复制 tmux attach 指令" data-copy-kind="tmux" onclick="copyTmuxRunCommand(event)">${titleCopyLabel('tmux')}</button>`);
546
+ }
525
547
  const chips = [
526
- metaChip('Run ID', currentState.runId, { long: true }),
527
548
  metaChip('工作区', basenamePath(currentState.workspacePath || currentState.repo), {
528
549
  title: currentState.workspacePath || currentState.repo,
529
550
  long: true,
@@ -536,18 +557,8 @@ function renderSelectedHeader() {
536
557
  if (currentState.git?.isGit || currentState.workspace?.git?.isGit) {
537
558
  chips.push(gitChip());
538
559
  }
539
- if (currentState.runner === 'tmux') {
540
- if (hasRunTmuxMetadata(currentState)) {
541
- chips.push(metaChip('终端', tmuxSessionName(currentState), {
542
- long: true,
543
- extra: `<button class="secondary copy-btn" onclick="copyTmuxRunCommand(event)">复制tmux attach指令</button>`
544
- }));
545
- } else {
546
- chips.push(metaChip('终端', 'tmux 现场尚未生成'));
547
- }
548
- }
549
560
  chips.push(refreshPulseChip());
550
- return `<div class="build-title"><span>${esc(currentState.label)}</span><button class="secondary copy-btn rename-btn" title="修改任务批次名称" onclick="renameRunLabel(event, currentState.runId)">${editIcon()}</button>${pill(currentState.status)}</div><div class="build-meta">${chips.join('')}</div>`;
561
+ return `<div class="build-title"><div class="build-title-main"><span class="build-title-text">${esc(currentState.label)}</span><button class="secondary copy-btn rename-btn" title="修改任务批次名称" onclick="renameRunLabel(event, currentState.runId)">${editIcon()}</button>${statusPill(currentState)}</div><div class="build-title-tools">${titleActions.join('')}</div></div><div class="build-meta">${chips.join('')}</div>`;
551
562
  }
552
563
  async function loadTaskDescription() {
553
564
  if (!selectedRun) { document.getElementById('taskDescription').textContent = '未选择任务批次'; return; }
@@ -576,22 +587,22 @@ function runActionState(key) {
576
587
  const anyWorkerRunning = (currentState.tasks || []).some(t => t.status === 'running');
577
588
  if (key === 'plan') {
578
589
  if (status === 'planning' || currentState.planner?.status === 'running') return { label: '拆分中', disabled: true, state: 'active' };
579
- if (status === 'planned' || status === 'running' || status === 'workers_completed' || status === 'batches_completed' || status === 'judging' || status === 'judged') return { label: '已拆分', disabled: true, state: 'done' };
590
+ if (status === 'planned' || status === 'running' || status === 'batches_completed' || status === 'judging' || status === 'judged') return { label: '已拆分', disabled: true, state: 'done' };
580
591
  if (status === 'plan_failed' || status === 'plan_empty') return { label: '重试拆分', disabled: false, state: 'retry' };
581
592
  return { label: '拆分', disabled: false, state: '' };
582
593
  }
583
594
  if (key === 'dispatch') {
584
595
  if (status === 'running' || anyWorkerRunning) return { label: '执行中', disabled: true, state: 'active' };
585
- if (status === 'planned' || status === 'batch_blocked') return { label: '执行', disabled: false, state: status === 'batch_blocked' ? 'retry' : '' };
586
- if (status === 'workers_failed') return { label: '重试执行', disabled: false, state: 'retry' };
587
- if (['workers_completed','batches_completed','judging','judged'].includes(status)) return { label: '已完成', disabled: true, state: 'done' };
596
+ if (status === 'planned') return { label: planApprovalPending() ? '开始执行' : '执行', disabled: false, state: '' };
597
+ if (status === 'batch_blocked') return { label: '重试执行', disabled: false, state: 'retry' };
598
+ if (['batches_completed','judging','judged'].includes(status)) return { label: '已完成', disabled: true, state: 'done' };
588
599
  return { label: '执行', disabled: true, state: 'done' };
589
600
  }
590
601
  if (key === 'judge') {
591
602
  if (status === 'judging' || currentState.judge?.status === 'running') return { label: '验收中', disabled: true, state: 'active' };
592
603
  if (status === 'judged') return { label: '已验收', disabled: true, state: 'done' };
593
604
  if (status === 'judge_failed') return { label: '重试验收', disabled: false, state: 'retry' };
594
- if (status === 'batches_completed' || status === 'workers_completed') return { label: '验收', disabled: false, state: '' };
605
+ if (status === 'batches_completed') return { label: '验收', disabled: false, state: '' };
595
606
  return { label: '验收', disabled: true, state: 'done' };
596
607
  }
597
608
  if (key === 'stop') {
@@ -705,7 +716,7 @@ function shortSessionId(thread) {
705
716
  }
706
717
  function sessionCell(thread) {
707
718
  if (!thread) return '-';
708
- return `<span class="session-cell" title="${esc(thread)}">…${esc(shortSessionId(thread))}</span><button class="copy-btn" title="复制完整 Codex 会话ID" onclick="copySessionId(event, '${esc(thread)}')">⧉</button>`;
719
+ return `<span class="session-cell-wrap"><span class="session-cell" title="${esc(thread)}">…${esc(shortSessionId(thread))}</span><button class="copy-btn" title="复制完整 Codex 会话ID" onclick="copySessionId(event, '${esc(thread)}')">⧉</button></span>`;
709
720
  }
710
721
  function taskStartedCell(t) {
711
722
  return t?.startedAt ? formatDateTime(t.startedAt) : '-';
@@ -844,6 +855,19 @@ function hideExecutionSummary() {
844
855
  el.classList.add('hidden');
845
856
  el.innerHTML = '';
846
857
  }
858
+ async function copyRunId(event) {
859
+ event.stopPropagation();
860
+ const runId = currentState?.runId || selectedRun || '';
861
+ if (!runId) return;
862
+ try {
863
+ await navigator.clipboard.writeText(runId);
864
+ const kind = event.currentTarget.dataset.copyKind || 'run-id';
865
+ event.currentTarget.textContent = '✓';
866
+ setTimeout(() => { event.currentTarget.innerHTML = titleCopyLabel(kind); }, 900);
867
+ } catch {
868
+ prompt('复制 Run ID', runId);
869
+ }
870
+ }
847
871
  async function copyRepoPath(event, repoPath = currentState?.workspacePath || currentState?.repo || '') {
848
872
  event.stopPropagation();
849
873
  if (!repoPath) return;
@@ -864,9 +888,10 @@ async function copyTmuxRunCommand(event) {
864
888
  const command = runAttachCommand(currentState);
865
889
  if (!command) return;
866
890
  try {
891
+ const kind = event.currentTarget.dataset.copyKind || 'tmux';
867
892
  await navigator.clipboard.writeText(command);
868
- event.currentTarget.textContent = '已复制';
869
- setTimeout(() => { event.currentTarget.textContent = '复制tmux attach指令'; }, 900);
893
+ event.currentTarget.textContent = '';
894
+ setTimeout(() => { event.currentTarget.innerHTML = titleCopyLabel(kind); }, 900);
870
895
  } catch {
871
896
  prompt('复制 tmux attach 命令', command);
872
897
  }
@@ -1024,7 +1049,11 @@ async function renameRunLabel(event, runId = selectedRun) {
1024
1049
  });
1025
1050
  }
1026
1051
  async function planRun() { if (selectedRun) await runActionWithPending('plan', async () => { await api(`/api/runs/${selectedRun}/plan`, {method:'POST'}); await refreshSelected(); }); }
1027
- async function dispatchRun() { if (selectedRun) await runActionWithPending('dispatch', async () => { await api(`/api/runs/${selectedRun}/dispatch`, {method:'POST'}); await refreshSelected(); }); }
1052
+ async function dispatchRun() {
1053
+ if (!selectedRun) return;
1054
+ const endpoint = currentState?.status === 'batch_blocked' ? 'retry' : 'dispatch';
1055
+ await runActionWithPending('dispatch', async () => { await api(`/api/runs/${selectedRun}/${endpoint}`, {method:'POST'}); await refreshSelected(); });
1056
+ }
1028
1057
  async function judgeRun() { if (selectedRun) await runActionWithPending('judge', async () => { await api(`/api/runs/${selectedRun}/judge`, {method:'POST'}); await refreshSelected(); }); }
1029
1058
  async function stopSelectedRun() {
1030
1059
  if (!selectedRun) return;
@@ -220,8 +220,28 @@ function deriveRunLabel(label, taskText) {
220
220
  return truncateDisplayWidth(cleaned, MAX_DERIVED_LABEL_DISPLAY_WIDTH) || 'task';
221
221
  }
222
222
 
223
+ function normalizePlanApprovalGate(value = false) {
224
+ const required = !!value;
225
+ return { required, approved: !required, approvedAt: null, approvedBy: null };
226
+ }
227
+
228
+ function planApprovalGate(state) {
229
+ return state?.gates?.planApproval || { required: false, approved: true, approvedAt: null, approvedBy: null };
230
+ }
231
+
232
+ function requiresPlanApproval(state) {
233
+ const gate = planApprovalGate(state);
234
+ return !!gate.required && !gate.approved;
235
+ }
223
236
 
224
- export async function createRun({ label = '', taskText = '', workspace = '', repo = DEFAULT_REPO, maxParallel = 3, workerSandbox = 'workspace-write' } = {}) {
237
+ function approvePlanGate(state, approvedBy = 'local-user') {
238
+ const gate = planApprovalGate(state);
239
+ if (!gate.required || gate.approved) return false;
240
+ state.gates = { ...(state.gates || {}), planApproval: { ...gate, approved: true, approvedAt: nowIso(), approvedBy } };
241
+ return true;
242
+ }
243
+
244
+ export async function createRun({ label = '', taskText = '', workspace = '', repo = DEFAULT_REPO, maxParallel = 3, workerSandbox = 'workspace-write', planApproval = false, requiresPlanApproval = false } = {}) {
225
245
  const resolvedWorkspace = await assertWorkspacePath(workspace || repo || DEFAULT_WORKSPACE);
226
246
  const workspaceMeta = await detectWorkspaceMetadata(resolvedWorkspace);
227
247
  const runLabel = deriveRunLabel(label, taskText);
@@ -243,6 +263,7 @@ export async function createRun({ label = '', taskText = '', workspace = '', rep
243
263
  repo: resolvedWorkspace,
244
264
  maxParallel: Number(maxParallel) || 3,
245
265
  workerSandbox: normalizeSandbox(workerSandbox),
266
+ gates: { planApproval: normalizePlanApprovalGate(planApproval || requiresPlanApproval) },
246
267
  runner: RUNNER,
247
268
  status: 'created', createdAt: nowIso(), updatedAt: nowIso(),
248
269
  planner: { status: 'pending' }, batches: [], tasks: [], judge: { status: 'pending' }
@@ -479,6 +500,27 @@ async function rotatePlannerAttempt(state, runDir) {
479
500
  }];
480
501
  }
481
502
 
503
+ async function rotateJudgeAttempt(state, runDir) {
504
+ const judgeDir = roleDir(runDir, 'judge');
505
+ if (!fs.existsSync(judgeDir)) return;
506
+ const attemptsDir = path.join(runDir, 'judge_attempts');
507
+ await ensureDir(attemptsDir);
508
+ const previousAttempts = (state.judgeAttempts || []).map(item => Number(item.attempt || 0)).filter(Number.isFinite);
509
+ const attempt = Number(state.judge?.attempt || 0) || Math.max(1, 1 + Math.max(0, ...previousAttempts));
510
+ const archivedDir = path.join(attemptsDir, `attempt-${String(attempt).padStart(2, '0')}`);
511
+ await fsp.rm(archivedDir, { recursive: true, force: true });
512
+ await fsp.rename(judgeDir, archivedDir);
513
+ state.judgeAttempts = [...(state.judgeAttempts || []), {
514
+ attempt,
515
+ status: state.judge?.status,
516
+ exitCode: state.judge?.exitCode ?? null,
517
+ startedAt: state.judge?.startedAt,
518
+ endedAt: state.judge?.endedAt,
519
+ archivedDir,
520
+ archivedAt: nowIso()
521
+ }];
522
+ }
523
+
482
524
  async function rotateWorkerAttempt(state, task) {
483
525
  const runDir = pathForRun(state.runId);
484
526
  const workerDir = roleDir(runDir, 'worker', task.id);
@@ -580,6 +622,7 @@ export async function dispatchRun(runId) {
580
622
  if (!state.tasks?.length) throw new Error('no tasks in plan');
581
623
  if (state.status === 'batch_blocked') throw new Error('current batch is blocked by failed/unknown tasks');
582
624
  if (allBatchesCompleted(state)) throw new Error('all batches completed; run final judge next');
625
+ approvePlanGate(state);
583
626
  state.status = 'running';
584
627
  await scheduleMoreWorkers(state);
585
628
  recomputeRunStatus(state);
@@ -770,16 +813,26 @@ export async function startJudge(runId) {
770
813
  return await withRunStateLock(runId, async () => {
771
814
  const state = await loadRun(runId);
772
815
  if (!state) throw new Error(`run not found: ${runId}`);
816
+ if (state.archived) throw new Error('archived run cannot be judged');
817
+ if (state.status === 'stopped') throw new Error('stopped run cannot be judged');
773
818
  recomputeRunStatus(state);
819
+ if (state.judge?.status === 'running') throw new Error('judge already running');
820
+ if (state.judge?.status === 'completed') throw new Error('judge already completed');
774
821
  if (!allBatchesCompleted(state) && state.tasks?.length) throw new Error('final judge is allowed only after all batches completed');
775
- const outDir = roleDir(pathForRun(runId), 'judge');
822
+ const runDir = pathForRun(runId);
823
+ if (state.judge?.status === 'failed') await rotateJudgeAttempt(state, runDir);
824
+ const outDir = roleDir(runDir, 'judge');
776
825
  await ensureDir(outDir);
777
826
  const judgeInputPath = path.join(outDir, 'judge_input.json');
778
827
  const judgeInput = await buildJudgeInput(state);
779
828
  await writeJsonAtomic(judgeInputPath, judgeInput);
780
829
  const prompt = defaultJudgePrompt(state, judgeInputPath);
781
- const child = await runner.startCodexTask({ runId: state.runId, taskId: 'judge', batchId: 'judge', runStatePath: statePath(pathForRun(runId)), prompt, sandbox: 'read-only', cwd: workspacePathOf(state), outDir });
782
- state.judge = { status: 'running', pid: child.pid, startedAt: nowIso(), dir: outDir };
830
+ const child = await runner.startCodexTask({ runId: state.runId, taskId: 'judge', batchId: 'judge', runStatePath: statePath(runDir), prompt, sandbox: 'read-only', cwd: workspacePathOf(state), outDir });
831
+ const previousJudgeAttempts = [
832
+ ...(state.judgeAttempts || []).map(item => Number(item.attempt || 0)),
833
+ Number(state.judge?.attempt || 0)
834
+ ].filter(Number.isFinite);
835
+ state.judge = { status: 'running', pid: child.pid, startedAt: nowIso(), dir: outDir, attempt: 1 + Math.max(0, ...previousJudgeAttempts) };
783
836
  state.status = 'judging';
784
837
  await saveRun(state);
785
838
  child.onExit(async code => {
@@ -816,6 +869,7 @@ export async function autoAdvanceRun(runId, { appClient = null, startCreated = f
816
869
  return await refreshRun(runId, appClient) || state;
817
870
  }
818
871
  if (state.status === 'planned') {
872
+ if (requiresPlanApproval(state)) return state;
819
873
  try { state = await dispatchRun(runId); }
820
874
  catch (error) { if (!/all batches completed|current batch is blocked/i.test(error.message || '')) throw error; }
821
875
  return await refreshRun(runId, appClient) || state;
@@ -1230,7 +1284,7 @@ export function summaryOfRun(s) {
1230
1284
  const tasks = s.tasks || [];
1231
1285
  const workspacePath = s.workspacePath || s.repo || '';
1232
1286
  const git = s.git || s.workspace?.git || null;
1233
- return { runId: s.runId, label: s.label, repo: s.repo || workspacePath, workspacePath, workspaceName: s.workspaceName || path.basename(workspacePath || ''), git, status: s.status, runner: s.runner || RUNNER, workerSandbox: s.workerSandbox || 'workspace-write', archived: !!s.archived, createdAt: s.createdAt, updatedAt: s.updatedAt, durationEnd: runDurationEndOfState(s), total: tasks.length, completed: tasks.filter(t => t.status === 'completed').length, failed: tasks.filter(t => ['failed','unknown'].includes(t.status)).length, running: tasks.filter(t => t.status === 'running').length, batches: (s.batches || []).map(b => ({ id: b.id, name: b.name, status: b.status, total: b.tasks?.length || 0, completed: (b.tasks || []).filter(t => t.status === 'completed').length })) };
1287
+ return { runId: s.runId, label: s.label, repo: s.repo || workspacePath, workspacePath, workspaceName: s.workspaceName || path.basename(workspacePath || ''), git, status: s.status, runner: s.runner || RUNNER, workerSandbox: s.workerSandbox || 'workspace-write', gates: s.gates || {}, archived: !!s.archived, createdAt: s.createdAt, updatedAt: s.updatedAt, durationEnd: runDurationEndOfState(s), total: tasks.length, completed: tasks.filter(t => t.status === 'completed').length, failed: tasks.filter(t => ['failed','unknown'].includes(t.status)).length, running: tasks.filter(t => t.status === 'running').length, batches: (s.batches || []).map(b => ({ id: b.id, name: b.name, status: b.status, total: b.tasks?.length || 0, completed: (b.tasks || []).filter(t => t.status === 'completed').length })) };
1234
1288
  }
1235
1289
 
1236
1290
  export async function readRunTaskText(runId) {
package/src/server.js CHANGED
@@ -8,6 +8,8 @@ import { createRun, listRuns, startPlanner, dispatchRun, startJudge, refreshRun,
8
8
  import { startAutoScheduler } from './scheduler.js';
9
9
 
10
10
  const PUBLIC_DIR = path.join(APP_ROOT, 'public');
11
+ const CODEX_INFO_TTL_MS = 30000;
12
+ let codexInfoCache = null;
11
13
 
12
14
  function send(res, status, body, type = 'application/json') {
13
15
  const data = type === 'application/json' ? JSON.stringify(body, null, 2) : body;
@@ -26,6 +28,13 @@ async function readBody(req) {
26
28
  function notFound(res) { send(res, 404, { error: 'not found' }); }
27
29
  function methodNotAllowed(res) { send(res, 405, { error: 'method not allowed' }); }
28
30
 
31
+ async function cachedCodexInfo(nowMs = Date.now()) {
32
+ if (codexInfoCache && codexInfoCache.expiresAt > nowMs) return codexInfoCache.value;
33
+ const value = await detectCodexInfo();
34
+ codexInfoCache = { value, expiresAt: nowMs + CODEX_INFO_TTL_MS };
35
+ return value;
36
+ }
37
+
29
38
  async function serveStatic(req, res, pathname) {
30
39
  let file = pathname === '/' ? path.join(PUBLIC_DIR, 'index.html') : path.join(PUBLIC_DIR, pathname.replace(/^\/+/, ''));
31
40
  file = path.resolve(file);
@@ -46,7 +55,7 @@ async function handleApi(req, res, url, appClient) {
46
55
  return send(res, 200, { ok: true, version: PACKAGE_VERSION, appRoot: APP_ROOT, runsDir: RUNS_DIR, defaultWorkspace: DEFAULT_WORKSPACE, defaultRepo: DEFAULT_REPO, runner: RUNNER, codexBin: CODEX_BIN });
47
56
  }
48
57
  if (req.method === 'GET' && url.pathname === '/api/codex') {
49
- return send(res, 200, { ok: true, codex: await detectCodexInfo() });
58
+ return send(res, 200, { ok: true, codex: await cachedCodexInfo() });
50
59
  }
51
60
  if (parts[1] === 'runs' && parts.length === 2) {
52
61
  if (req.method === 'GET') return send(res, 200, { runs: await listRuns({ includeArchived: url.searchParams.get('includeArchived') === '1', workspace: url.searchParams.get('workspace') || '' }) });