agentworkshop 0.2.1

package/scripts/omp-real-test.py

@@ -0,0 +1,328 @@
+# -*- coding: utf-8 -*-
+"""真实 OMP 多 Agent 验收测试(python -u 运行,需已启动服务端 3001 + omp CLI)。
+
+三个真实场景(全部走真实 omp 子进程 + 真实 LLM):
+  场景 A —— 简单任务执行:lead 自动分解 → worker 执行 → 父任务 COMPLETED
+  场景 B —— A2A 通信:worker1 ↔ worker2 经 send_message_to_agent / poll_messages 往返一问一答
+  场景 C —— 多任务队列:FIFO 队列 + 单 WORKING 不变量(每个 worker 串行消费,顺序完成)
+
+用法:
+  python -u scripts/omp-real-test.py            # 全部场景
+  python -u scripts/omp-real-test.py simple     # 仅场景 A
+  python -u scripts/omp-real-test.py a2a        # 仅场景 B
+  python -u scripts/omp-real-test.py queue      # 仅场景 C
+环境变量: AW_BASE / AW_MODEL_PROVIDER / AW_MODEL(缺省 zhipu-coding-plan/glm-5-turbo)
+"""
+import json
+import os
+import sys
+import time
+import urllib.request
+from datetime import datetime, timezone
+
+sys.stdout.reconfigure(encoding='utf-8', errors='replace')
+
+BASE = os.environ.get('AW_BASE', 'http://127.0.0.1:3001')
+PROVIDER = os.environ.get('AW_MODEL_PROVIDER', 'zhipu-coding-plan')
+MODEL = os.environ.get('AW_MODEL', 'glm-5-turbo')
+
+PASS = 0
+FAIL = 0
+CHANNEL_IDS = []
+
+
+def check(name, ok, detail=''):
+    global PASS, FAIL
+    tag = 'PASS' if ok else 'FAIL'
+    if ok:
+        PASS += 1
+    else:
+        FAIL += 1
+    print(f'  [{tag}]  {name}{f" — {detail}" if detail else ""}', flush=True)
+
+
+def log(msg):
+    print(f'[{time.strftime("%H:%M:%S")}] {msg}', flush=True)
+
+
+def api(method, path, body=None, token=None, timeout=30):
+    headers = {'Content-Type': 'application/json'}
+    if token:
+        headers['Authorization'] = 'Bearer ' + token
+    req = urllib.request.Request(
+        BASE + path,
+        data=json.dumps(body).encode() if body is not None else None,
+        method=method,
+        headers=headers,
+    )
+    with urllib.request.urlopen(req, timeout=timeout) as resp:
+        d = json.loads(resp.read().decode())
+    if isinstance(d, dict) and d.get('code') not in (None, 0, 'ok', 'success'):
+        raise RuntimeError(f'{method} {path} -> {str(d)[:240]}')
+    return d.get('data', d) if isinstance(d, dict) else d
+
+
+def ptext(parts):
+    if not isinstance(parts, list):
+        return ''
+    return ' '.join(p.get('text', '') for p in parts if isinstance(p, dict) and p.get('text'))
+
+
+def wait_until(name, cond, timeout_s, interval_s=3):
+    deadline = time.monotonic() + timeout_s
+    last = None
+    while time.monotonic() < deadline:
+        try:
+            last = cond()
+            if last:
+                return last
+        except Exception as e:  # noqa: BLE001
+            last = e
+        time.sleep(interval_s)
+    raise TimeoutError(f'wait_until 超时: {name} (last={str(last)[:200]})')
+
+
+def all_events(token, channel_id):
+    data = api('GET', f'/api/workshop/channels/{channel_id}/events?limit=1000', token=token)
+    return data.get('items', [])
+
+
+def all_tasks(token, channel_id):
+    return api('GET', f'/api/workshop/channels/{channel_id}/tasks', token=token)
+
+
+def all_messages(token, channel_id, limit=500):
+    msgs = api('GET', f'/api/workshop/channels/{channel_id}/messages?limit={limit}', token=token)
+    return sorted(msgs, key=lambda m: m['createdAt'])
+
+
+# ═══════════════════════════════════════════════════════════════
+
+def make_team(token, name, n_workers=2, worker_prefix='w'):
+    """新建 channel:1 lead + n_workers 个 OMP worker,全部真实 omp 子进程。"""
+    ch = api('POST', '/api/workshop/channels', body={
+        'name': name,
+        'description': '真实 OMP 验收测试',
+        'leadAgent': {'name': f'{name}-lead', 'harness': 'omp',
+                      'config': {'provider': PROVIDER, 'model': MODEL}},
+    }, token=token)
+    channel_id = ch['channelId']
+    CHANNEL_IDS.append(channel_id)
+    lead_id = ch.get('leadAgentId')
+    workers = []
+    for i in range(n_workers):
+        w = api('POST', f'/api/workshop/channels/{channel_id}/agents', body={
+            'name': f'{worker_prefix}{i + 1}-{name}',
+            'harness': 'omp',
+            'role': 'worker',
+            'config': {'provider': PROVIDER, 'model': MODEL,
+                       'systemPromptPrefix': f'你是 worker{i + 1},本场景测试中的执行者。'},
+        }, token=token)
+        workers.append(w['id'])
+    log(f'团队就绪 channel={channel_id[:8]} lead={lead_id[:8]} workers={[w[:8] for w in workers]} '
+        f'(模型 {PROVIDER}/{MODEL})')
+    return {'channelId': channel_id, 'leadId': lead_id, 'workers': workers}
+
+
+def submit(token, channel_id, title, description, assignee=None):
+    body = {'title': title, 'description': description}
+    if assignee:
+        body['assigneeId'] = assignee
+    t = api('POST', f'/api/workshop/channels/{channel_id}/tasks', body=body, token=token)
+    return t['id']
+
+
+def terminal_state(tasks, tid):
+    t = next((x for x in tasks if x['id'] == tid), None)
+    return t['state'] if t else None
+
+
+TERMINAL = ('COMPLETED', 'FAILED', 'CANCELED')
+
+
+def wait_terminal(token, channel_id, tid, timeout_s):
+    """仅当任务进入终态时返回状态字符串,否则返回 None(供 wait_until 轮询)。"""
+    st = terminal_state(all_tasks(token, channel_id), tid)
+    return st if st in TERMINAL else None
+
+
+# ═══════════ 场景 A:简单任务执行 ═══════════
+
+def scenario_simple(token, channel_id):
+    log('── 场景 A:简单任务执行(lead 自动分解 → worker 执行 → 父任务收口) ──')
+    t0 = time.monotonic()
+    pid = submit(token, channel_id,
+                 '简单任务:法国首都',
+                 '请完成这个小任务:写出句子「巴黎是法国的首都」,然后用 complete_task 工具提交交付物(summary=该句子),不要做其他多余操作。')
+    final = wait_until('A 父任务终态', lambda: wait_terminal(token, channel_id, pid, 900),
+                       timeout_s=900, interval_s=5)
+    elapsed = int(time.monotonic() - t0)
+    check('父任务 COMPLETED', final == 'COMPLETED', f'state={final} 耗时={elapsed}s')
+    tasks = all_tasks(token, channel_id)
+    children = [t for t in tasks if t.get('parentId') == pid]
+    check('lead 已分解出子任务', len(children) >= 1, f'children={len(children)}')
+    done_child = next((c for c in children if c['state'] == 'COMPLETED'), None)
+    check('子任务完成且带交付', done_child is not None and len(done_child.get('artifacts', [])) > 0,
+          f"child={done_child['id'][:8] if done_child else '-'} artifacts={len(done_child.get('artifacts', [])) if done_child else 0}")
+    parent = next(t for t in tasks if t['id'] == pid)
+    parent_text = json.dumps(parent.get('artifacts', ''), ensure_ascii=False)
+    check('父任务含最终交付(含关键词)', '巴黎' in parent_text, parent_text[:100])
+    return children, parent
+
+
+# ═══════════ 场景 B:A2A 通信(worker ↔ worker) ═══════════
+
+def scenario_a2a(token, channel_id, workers):
+    w1, w2 = workers[0], workers[1]
+    log('── 场景 B:A2A 通信(w1 向空闲 w2 发起 require_reply 提问,w2 回执,验证 in_reply_to 关联) ──')
+    base_msgs = {m['id'] for m in all_messages(token, channel_id)}
+    t0 = time.monotonic()
+
+    # 仅给 w1 派任务;w2 保持空闲(idle 时收到消息 → 独立 peer 回合 → 自动回执关联)
+    t1 = submit(token, channel_id, 'A2A-发送方:向同事提问并回报',
+                f'你是 A2A 通信测试的发起方。步骤:\n'
+                f'1. 用 send_message_to_agent 发消息给同事 <{w2}>,参数 message="请只回复:RED-77",'
+                f'require_reply=true(要求对方回执)。\n'
+                f'2. 调用 poll_messages(wait_seconds=90) 阻塞等待对方回复。\n'
+                f'3. 收到回复后调用 complete_task,summary 参数填你收到的回复全文。\n'
+                f'只与同事 [{w2}] 通信,不要联系 lead。',
+                assignee=w1)
+
+    # w1 完成即结束(同步检查 reply 链路证据)
+    wait_until('A2A 发送方任务终态', lambda: wait_terminal(token, channel_id, t1, 900),
+               timeout_s=900, interval_s=5)
+    elapsed = int(time.monotonic() - t0)
+    st = terminal_state(all_tasks(token, channel_id), t1)
+    check('发送方任务 COMPLETED(拿到回执后收口)', st == 'COMPLETED', f'state={st} 耗时={elapsed}s')
+
+    # 通信链路证据:w1 → w2 提问(带 require_reply)+ w2 → w1 回复(带 in_reply_to 关联)
+    msgs = [m for m in all_messages(token, channel_id) if m['id'] not in base_msgs]
+    asked = [m for m in msgs if m.get('fromAgentId') == w1 and m.get('toAgentId') == w2]
+    replies = [m for m in msgs if m.get('fromAgentId') == w2 and m.get('toAgentId') == w1
+               and m.get('metadata', {}).get('x-aw-in-reply-to')]
+    check('w1→w2 提问已投递(带 require_reply)',
+          len(asked) >= 1 and any(m.get('metadata', {}).get('x-aw-require-reply') == 'true' for m in asked),
+          f'asked={len(asked)}')
+    check('w2→w1 回执已投递(带 in_reply_to 关联)', len(replies) >= 1, f'replies={len(replies)}')
+    if asked and replies:
+        check('回执内容含 RED-77', 'RED-77' in ptext(replies[-1].get('parts', [])),
+              ptext(replies[-1].get('parts', []))[:60])
+        check('in_reply_to 精确关联提问消息',
+              replies[-1].get('metadata', {}).get('x-aw-in-reply-to') == asked[-1]['id'],
+              f"reply→{str(replies[-1].get('metadata', {}).get('x-aw-in-reply-to'))[:8]} ask={asked[-1]['id'][:8]}")
+    # 信箱排空:新消息最终全部消费(轮询直到无 pending)
+    wait_until('A2A 信箱排空', lambda: not [m for m in all_messages(token, channel_id, 500)
+                                            if m['id'] not in base_msgs and m.get('state') in ('pending', 'consuming')],
+               timeout_s=120, interval_s=3)
+    msgs_final = [m for m in all_messages(token, channel_id, 500) if m['id'] not in base_msgs]
+    check('全部新消息已消费(无残留 pending)', all(m.get('state') == 'consumed' for m in msgs_final),
+          f'consumed={sum(1 for m in msgs_final if m.get("state") == "consumed")}/{len(msgs_final)}')
+    return msgs_final
+
+
+# ═══════════ 场景 C:多任务队列 ═══════════
+
+def scenario_queue(token, channel_id, workers):
+    log('── 场景 C:多任务队列(直发 worker 信箱,验证 FIFO + 单 WORKING 不变量) ──')
+    w1, w2 = workers[0], workers[1]
+    # 直发:任务绕过 lead,直接进 worker 自己的信箱队列 —— 队列 FIFO 的确定性测试
+    assigns = [w1, w2, w1, w2]
+    titles = [f'队列任务-{i + 1}(回复 Q{i + 1}-DONE)' for i in range(len(assigns))]
+    descs = [f'请只用一句话回复:{f"Q{i + 1}-DONE"}(不要多余文字),然后 complete_task。' for i in range(len(assigns))]
+    ids = [submit(token, channel_id, titles[i], descs[i], assignee=assigns[i])
+           for i in range(len(assigns))]
+    log(f'已直发 {len(ids)} 个任务: w1←[{",".join(str(i + 1) for i, a in enumerate(assigns) if a == w1)}] '
+        f'w2←[{",".join(str(i + 1) for i, a in enumerate(assigns) if a == w2)}]')
+
+    t0 = time.monotonic()
+    final_states = {}
+    deadline = time.monotonic() + 900
+    while time.monotonic() < deadline:
+        tasks = all_tasks(token, channel_id)
+        final_states = {tid: terminal_state(tasks, tid) for tid in ids}
+        if all(v == 'COMPLETED' for v in final_states.values()):
+            break
+        time.sleep(4)
+    elapsed = int(time.monotonic() - t0)
+
+    check('全部任务 COMPLETED', all(v == 'COMPLETED' for v in final_states.values()),
+          f'states={list(final_states.values())} 耗时={elapsed}s')
+
+    # 单 WORKING 不变量:任一时刻每个 assignee 至多 1 个 WORKING(事件重放)
+    evs = all_events(token, channel_id)
+    tasks_final = all_tasks(token, channel_id)
+    assignee_of = {t['id']: t['assigneeId'] for t in tasks_final}
+    sim = {}
+    max_concurrent = 0
+    for e in evs:
+        if e['type'] != 'task.status':
+            continue
+        payload = e.get('payload') or {}
+        state = payload.get('state')
+        tid = e.get('taskId')
+        ag = e.get('agentId') or ''
+        if not tid or not ag or assignee_of.get(tid) != ag:
+            continue
+        if state == 'WORKING':
+            sim[ag] = sim.get(ag, 0) + 1
+            max_concurrent = max(max_concurrent, sim[ag])
+        elif state in ('COMPLETED', 'FAILED', 'CANCELED', 'WAITING', 'ASSIGNED'):
+            sim[ag] = max(0, sim.get(ag, 0) - 1)
+    check('单 WORKING 不变量(每 assignee 并发 WORKING≤1)', max_concurrent <= 1,
+          f'max_concurrent={max_concurrent}')
+
+    # FIFO 顺序:每个 worker 直发队列按提交序完成(信箱 FIFO)
+    tasks = all_tasks(token, channel_id)
+    by_assignee = {}
+    for t in tasks:
+        if t['id'] in ids and t['state'] == 'COMPLETED':
+            by_assignee.setdefault(t['assigneeId'], []).append(t)
+    fifo_ok = True
+    orders = []
+    for ag, lst in by_assignee.items():
+        order = [int(t['title'].split('-')[1].split('(')[0]) for t in lst]
+        orders.append(f"{ag[:8]}:{order}")
+        fifo_ok = fifo_ok and order == sorted(order)
+    check('每 worker 直发队列 FIFO 顺序完成', fifo_ok, '; '.join(orders) or '无')
+    return ids, by_assignee
+
+
+# ═══════════════════════════════════════════════════════════════
+
+def main():
+    which = sys.argv[1] if len(sys.argv) > 1 else 'all'
+    log(f'目标: {BASE} | 模型 {PROVIDER}/{MODEL} | 场景: {which}')
+    user = api('POST', '/api/workshop/users/register', body={
+        'name': f'omp-real-{datetime.now(timezone.utc).strftime("%H%M%S")}',
+        'email': f'ompreal{int(time.time() * 1000) % 10**8}@test.local',
+        'password': 'Test1234!',
+    })
+    token = user['token']
+
+    if which in ('all', 'simple'):
+        team = make_team(token, 'simple', n_workers=1)
+        try:
+            scenario_simple(token, team['channelId'])
+        finally:
+            pass  # 汇聚终局一起清理
+
+    if which in ('all', 'a2a'):
+        team = make_team(token, 'a2a', n_workers=2, worker_prefix='w')
+        try:
+            scenario_a2a(token, team['channelId'], team['workers'])
+        finally:
+            pass
+
+    if which in ('all', 'queue'):
+        team = make_team(token, 'queue', n_workers=2)
+        try:
+            scenario_queue(token, team['channelId'], team['workers'])
+        finally:
+            pass
+
+    log(f'★ 结果: {PASS} passed, {FAIL} failed')
+    sys.exit(1 if FAIL else 0)
+
+
+if __name__ == '__main__':
+    main()

package/scripts/smoke-optimization-verify.ts

@@ -0,0 +1,72 @@
+import { openWorkshopDb } from '../server/services/workshop/db/database'
+import { createTaskRepo } from '../server/services/workshop/db/task.repo'
+import { TaskEngine } from '../server/services/workshop/runtime/task-engine'
+import { createMessageRepo } from '../server/services/workshop/db/message.repo'
+
+const db = openWorkshopDb(':memory:')
+const now = new Date().toISOString()
+db.prepare('INSERT INTO channels (id, name, enabled, created_at, updated_at) VALUES (\'ch1\', \'test\', 1, ?, ?)').run(now, now)
+const tasks = createTaskRepo(db)
+const messages = createMessageRepo(db)
+const engine = new TaskEngine({ tasks, messages })
+
+let failed = 0
+const okIf = (m: string, c: boolean) => {
+  if (c) console.log(`PASS ${m}`)
+  else {
+    console.log(`FAIL ${m}`)
+    failed++
+  }
+}
+
+const parent = engine.create({ channelId: 'ch1', creatorId: 'u1', assigneeId: 'lead', title: 'parent', description: '[mode:goal] do it' })
+// 调度器同款前置:SUBMITTED → WORKING 后 dispatch(WORKING → WAITING 合法)
+engine.transition(parent.id, 'WORKING', 'lead')
+engine.dispatch(parent, { assigneeId: 'w1', title: 'child one' })
+
+// 判重:同父 + 归一化同题(大小写/空白)+ 非终态 → 409,不创建新任务
+let caught409 = false
+try {
+  engine.dispatch(parent, { assigneeId: 'w2', title: 'Child  ONE' })
+}
+catch (e) {
+  const err = e as { code?: string, statusCode?: number, status?: number }
+  caught409 = err.code === 'DUPLICATE_DISPATCH' || err.statusCode === 409 || err.status === 409
+}
+okIf('duplicate-dispatch-409', caught409)
+
+// 409 后无新任务:parent + child one = 2 行;children = 1
+const lite = engine.listLite('ch1')
+const views = engine.queueViewsOfLite('ch1')
+okIf(`listLite rows after 409: ${lite.length} (expect 2: parent+child one)`, lite.length === 2)
+okIf('lite fields ok(状态/进度/数组字段齐备,大列置空)', lite.every(t => typeof t.id === 'string' && t.state && typeof t.progress === 'number' && Array.isArray(t.history) && Array.isArray(t.artifacts)))
+okIf(`queueViewsLite assignees: ${[...views.keys()].sort().join(',')} (expect lead,w1)`, [...views.keys()].sort().join(',') === 'lead,w1')
+okIf(`queued in w1 view: ${views.get('w1')?.queued.length ?? -1} (expect 1)`, views.get('w1')?.queued.length === 1)
+okIf(`listChildrenMeta after 409: ${tasks.listChildrenMeta('ch1', parent.id).length} (expect 1)`, tasks.listChildrenMeta('ch1', parent.id).length === 1)
+
+// 异题派发合法:第二个子任务可建
+const c3 = engine.dispatch(parent, { assigneeId: 'w2', title: 'child two' })
+okIf('second-distinct-child', !!c3)
+okIf(`listLite rows after child two: ${engine.listLite('ch1').length} (expect 3)`, engine.listLite('ch1').length === 3)
+okIf(`listChildrenMeta after child two: ${tasks.listChildrenMeta('ch1', parent.id).length} (expect 2)`, tasks.listChildrenMeta('ch1', parent.id).length === 2)
+
+// 完成闸门:仅一个子完成 → parent 仍 WAITING;双子终态后显式 complete(parent) 放行
+const childrenIds = tasks.listChildrenMeta('ch1', parent.id).map(r => r.id)
+engine.transition(childrenIds[0]!, 'WORKING', 'w1')
+engine.complete(childrenIds[0]!, [{ artifactId: 'a1', name: 'output', parts: [{ text: 'done' }] }])
+okIf(`complete-child: ${engine.get(childrenIds[0]!)?.state} (expect COMPLETED)`, engine.get(childrenIds[0]!)?.state === 'COMPLETED')
+okIf(`parent state(单子完成): ${engine.get(parent.id)?.state} (expect WAITING)`, engine.get(parent.id)?.state === 'WAITING')
+engine.transition(childrenIds[1]!, 'WORKING', 'w2')
+engine.complete(childrenIds[1]!, [{ artifactId: 'a2', name: 'output2', parts: [{ text: 'done2' }] }])
+// 父任务收口由调度器(goal 宽限窗/规则引擎)负责;TaskEngine 层验证 done-check 闸门:
+// 全部子任务终态后 complete(parent) 应放行(WAITING → COMPLETED 合法迁移)
+engine.complete(parent.id, [{ artifactId: 'a3', name: 'summary', parts: [{ text: 'all done' }] }])
+okIf(`parent state(双子终态后显式 complete): ${engine.get(parent.id)?.state} (expect COMPLETED)`, engine.get(parent.id)?.state === 'COMPLETED')
+
+// F1:legacy 迁移列与索引
+const cols = (db.prepare('PRAGMA table_info(tasks)').all() as Array<{ name: string }>).map(c => c.name)
+okIf('route_reason column present', cols.includes('route_reason'))
+okIf('idx_tasks_parent present', !!db.prepare('SELECT name FROM sqlite_master WHERE type=\'index\' AND name=\'idx_tasks_parent\'').get())
+
+console.log(failed ? `SMOKE FAILED(${failed})` : 'SMOKE ALL PASS')
+process.exit(failed ? 1 : 0)

package/scripts/start.mjs

@@ -0,0 +1,60 @@
+// ============================================================
+// 生产启动入口（配置驱动 + 双模式）
+// 模式判定（shared/config/home.mjs 单一入口）：
+//   repo 模式：cwd 在项目检出内 → config.yml/data 都在项目根（开发/部署 checkout 场景）
+//   home 模式：全局安装（aw start）→ cwd=AW Home,配置=AW Home/config.yml,
+//              运行时覆盖=AW Home/runtime-settings.json,数据=AW Home/data
+// 另在启动前预载 .env（cwd 的 .env;NUXT_SESSION_PASSWORD 等密钥只经环境
+// 注入,绝不写入 config.yml/仓库;已导出的真实环境变量优先）。
+// ============================================================
+import { existsSync, readFileSync } from 'node:fs'
+import { dirname, join, resolve } from 'node:path'
+import { fileURLToPath } from 'node:url'
+import { resolveRunMode } from '../shared/config/home.mjs'
+
+const packageRoot = resolve(dirname(fileURLToPath(import.meta.url)), '..')
+
+// ---- .env 预载（KEY=VALUE 行,# 注释;不覆盖已存在的环境变量;cwd 优先,包根兜底） ----
+for (const envPath of [resolve(process.cwd(), '.env'), join(packageRoot, '.env')]) {
+  if (!existsSync(envPath)) continue
+  for (const line of readFileSync(envPath, 'utf8').split(/\r?\n/)) {
+    const m = line.match(/^\s*([A-Za-z_][A-Za-z0-9_]*)\s*=\s*(.*)\s*$/)
+    if (!m || line.trim().startsWith('#')) continue
+    const key = m[1]
+    let val = m[2] ?? ''
+    if ((val.startsWith('"') && val.endsWith('"')) || (val.startsWith('\'') && val.endsWith('\''))) val = val.slice(1, -1)
+    if (process.env[key] === undefined) process.env[key] = val
+  }
+}
+
+// 共享配置引擎：config.yml + runtime-settings + env（模式感知路径）
+const { loadEffective } = await import('../shared/config/engine.mjs')
+
+// CLI 传参优先（node scripts/start.mjs --port 8080）
+let argPort
+{
+  const eq = process.argv.find(a => a.startsWith('--port='))
+  const idx = process.argv.indexOf('--port')
+  argPort = eq ? eq.split('=')[1] : (idx >= 0 ? process.argv[idx + 1] : undefined)
+}
+
+const rm = resolveRunMode({ cwd: process.cwd(), packageRoot, env: process.env })
+const eff = loadEffective({
+  configPath: rm.configPath,
+  settingsPath: rm.settingsPath,
+  env: { ...process.env, ...(argPort ? { PORT: argPort } : {}) },
+  mode: 'prod',
+})
+
+const host = eff.effective['server.host'] ?? '0.0.0.0'
+const prodPort = eff.effective['server.prod.port'] ?? 3000
+const portSource = eff.sources['server.prod.port']
+
+process.env.HOST = process.env.HOST || process.env.NITRO_HOST || String(host)
+process.env.PORT = process.env.PORT || process.env.NITRO_PORT || String(prodPort)
+process.env.NITRO_HOST = process.env.HOST
+process.env.NITRO_PORT = process.env.PORT
+
+console.log(`[config] 生产服务启动 -> http://${process.env.HOST}:${process.env.PORT}  (模式: ${rm.mode}, port source: ${portSource})`)
+
+await import('../.output/server/index.mjs')