feat: 提升并发容量默认配置

ai_knowledge_base_v2/apps/admin-web/src/App.vue

@@ -1067,6 +1067,12 @@ async function clearFeishuCache() {
                     v-model="systemSettingValues[setting.key]"
                   />
                   <small>{{ setting.description }}</small>
+                  <small
+                    v-if="setting.valueHint"
+                    class="setting-value-hint"
+                  >
+                    {{ setting.valueHint(systemSettingValues[setting.key]) }}
+                  </small>
                 </label>
               </div>
             </div>

ai_knowledge_base_v2/apps/admin-web/src/config/systemSettings.ts

@@ -16,6 +16,7 @@ export interface SystemSettingDefinition {
   placeholder?: string;
   options?: SystemSettingOption[];
   description: string;
+  valueHint?: (value: SystemSettingValue) => string;
 }
 
 export interface SystemSettingSection {
@@ -181,28 +182,29 @@ export const systemSettingSections: SystemSettingSection[] = [
         key: "chat_max_active_requests",
         label: "问答最大并发数",
         type: "number",
-        defaultValue: 2,
+        defaultValue: 30,
         min: 1,
         max: 1000,
-        description: "同一后端进程内同时进入飞书和模型生成链路的最大请求数。",
+        description: "用户正式问答的全局执行上限；Redis 可用时由所有后端实例共享，超过后进入排队。",
+        valueHint: concurrentCapacityHint,
       },
       {
         key: "chat_max_queue_size",
         label: "问答最大排队数",
         type: "number",
-        defaultValue: 20,
+        defaultValue: 90,
         min: 0,
         max: 10000,
-        description: "超过最大并发后允许等待的请求数量，超过后直接提示稍后再试。",
+        description: "超过最大并发后允许等待的请求数量。默认按并发数的 3 倍预留，避免课程结束后的集中提问被直接拒绝。",
       },
       {
         key: "chat_queue_timeout_seconds",
         label: "问答排队超时（秒）",
         type: "number",
-        defaultValue: 60,
+        defaultValue: 90,
         min: 1,
         max: 3600,
-        description: "请求在排队中超过该时间后自动结束并提示用户稍后再试。",
+        description: "请求在排队中超过该时间后自动结束。默认 90 秒可覆盖 30 并发下约三轮请求释放。",
       },
       {
         key: "chat_active_lease_seconds",
@@ -292,4 +294,16 @@ export const systemSettingSections: SystemSettingSection[] = [
   },
 ];
 
+function concurrentCapacityHint(value: SystemSettingValue) {
+  const concurrency = Math.max(1, Number(value) || 1);
+  const lower = roundDownToTen((concurrency * 120) / 25 / 1.4);
+  const upper = roundDownToTen((concurrency * 120) / 25 / 1.2);
+  const capacity = lower === upper ? `约 ${lower} 人` : `约 ${lower}–${upper} 人`;
+  return `容量估算：按单次问答约 25 秒并预留 20%–40% 波动，当前 ${concurrency} 并发可承接 2 分钟内${capacity}集中提问。实际能力还受模型服务商额度和回答时长影响。`;
+}
+
+function roundDownToTen(value: number) {
+  return Math.max(1, Math.floor(value / 10) * 10);
+}
+
 export const systemSettingDefinitions = systemSettingSections.flatMap((section) => section.settings);

ai_knowledge_base_v2/apps/admin-web/src/styles.css

@@ -1725,6 +1725,14 @@ textarea {
   line-height: 1.5;
 }
 
+.setting-field > .setting-value-hint {
+  padding: 9px 11px;
+  border: 1px solid #cde6dc;
+  border-radius: 8px;
+  background: #f3faf7;
+  color: #246b52;
+}
+
 .setting-field .el-input-number,
 .setting-field .el-input,
 .setting-field .el-select {

ai_knowledge_base_v2/apps/backend/alembic/versions/0028_chat_concurrency_default.py

@@ -0,0 +1,84 @@
+"""raise the default chat concurrency to thirty
+
+Revision ID: 0028_chat_concurrency
+Revises: 0027_content_generation
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "0028_chat_concurrency"
+down_revision = "0027_content_generation"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    config = sa.table(
+        "sys_system_config",
+        sa.column("config_key", sa.String(length=100)),
+        sa.column("config_value", sa.Text()),
+        sa.column("description", sa.String(length=255)),
+    )
+    connection = op.get_bind()
+    _set_default(
+        connection,
+        config,
+        key="chat_max_active_requests",
+        old_default="2",
+        new_default="30",
+        description="用户正式问答的全局执行上限；Redis 可用时由所有后端实例共享，超过后进入排队。",
+    )
+    _set_default(
+        connection,
+        config,
+        key="chat_max_queue_size",
+        old_default="20",
+        new_default="90",
+        description="超过最大并发后允许等待的请求数量。默认按并发数的 3 倍预留。",
+    )
+    _set_default(
+        connection,
+        config,
+        key="chat_queue_timeout_seconds",
+        old_default="60",
+        new_default="90",
+        description="请求在排队中超过该时间后自动结束。默认可覆盖约三轮请求释放。",
+    )
+
+
+def downgrade() -> None:
+    # 运行时配置可能已经被管理员调整，降级代码版本时不覆盖现场值。
+    pass
+
+
+def _set_default(
+    connection,
+    config,
+    *,
+    key: str,
+    old_default: str,
+    new_default: str,
+    description: str,
+) -> None:
+    current = connection.execute(
+        sa.select(config.c.config_value).where(config.c.config_key == key)
+    ).scalar_one_or_none()
+    if current is None:
+        connection.execute(
+            config.insert().values(
+                config_key=key,
+                config_value=new_default,
+                description=description,
+            )
+        )
+        return
+    if str(current).strip() == old_default:
+        connection.execute(
+            config.update()
+            .where(config.c.config_key == key)
+            .values(config_value=new_default, description=description)
+        )

ai_knowledge_base_v2/apps/backend/app/core/config.py

@@ -59,9 +59,9 @@ class Settings(BaseSettings):
 
     default_daily_chat_limit: int = 100
     default_user_name_prefix: str = "用户"
-    chat_max_active_requests: int = 2
+    chat_max_active_requests: int = 30
-    chat_max_queue_size: int = 20
+    chat_max_queue_size: int = 90
-    chat_queue_timeout_seconds: int = 60
+    chat_queue_timeout_seconds: int = 90
     chat_active_lease_seconds: int = 900
     periodic_report_worker_enabled: bool = True
     periodic_report_poll_seconds: int = 5

ai_knowledge_base_v2/apps/backend/tests/test_production_readiness.py

@@ -27,6 +27,7 @@ from app.services.security_state_service import SecurityStateService
 from app.services.auth_service import AuthService
 from app.services.model_service import ModelClientService
 from app.core.observability import RequestObservabilityMiddleware
+from app.core.config import Settings
 
 
 @pytest.fixture(autouse=True)
@@ -43,6 +44,14 @@ def test_secret_is_encrypted_and_masked():
     assert SecretService.masked(encrypted) == MASKED_SECRET
 
 
+def test_production_chat_capacity_defaults_cover_expected_burst():
+    settings = Settings(_env_file=None)
+
+    assert settings.chat_max_active_requests == 30
+    assert settings.chat_max_queue_size == 90
+    assert settings.chat_queue_timeout_seconds == 90
+
+
 def test_migration_revision_ids_fit_default_alembic_version_column():
     versions_dir = Path(__file__).parents[1] / "alembic" / "versions"
     revision_pattern = re.compile(r'^revision\s*=\s*["\']([^"\']+)', re.MULTILINE)