feat: add per-model concurrency limits

frontend-web/src/components/page-sidebar.tsx

@@ -7,6 +7,7 @@ import {
   ChevronsUpDown,
   Folder,
   FolderOpen,
+  Gauge,
   ListTree,
   MousePointerClick,
   Settings,
@@ -231,6 +232,14 @@ function SecondaryNavSidebar() {
                       白名单管理
                     </DropdownMenuItem>
                   )}
+                  {isAdmin && (
+                    <DropdownMenuItem
+                      onClick={() => navigate("/page/strategy/settings/model-concurrency")}
+                    >
+                      <Gauge className="size-4" />
+                      模型并发限制
+                    </DropdownMenuItem>
+                  )}
                   {canManageSensitiveWords() && (
                     <DropdownMenuItem
                       onClick={() => navigate("/page/strategy/admin/sensitive-words")}

frontend-web/src/core/system-settings/api.ts

@@ -342,6 +342,18 @@ export interface SkillCompressionUpdate {
   keep_index_in_prompt?: boolean;
 }
 
+export interface ModelConcurrencySettings {
+  enabled: boolean;
+  default_limit: number;
+  model_limits: Record<string, number>;
+}
+
+export interface ModelConcurrencyUpdate {
+  enabled?: boolean;
+  default_limit?: number;
+  model_limits?: Record<string, number>;
+}
+
 export async function loadSkillCompressionSettings(): Promise<SkillCompressionSettings> {
   const response = await apiFetch(
     `${getBackendBaseURL()}/api/system-settings/skill-compression`,
@@ -361,6 +373,53 @@ export async function loadSkillCompressionSettings(): Promise<SkillCompressionSe
   return data;
 }
 
+export async function loadModelConcurrencySettings(): Promise<ModelConcurrencySettings> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/model-concurrency`,
+  );
+  if (!response.ok) {
+    throw new Error(
+      `Failed to load model concurrency settings: ${response.status}`,
+    );
+  }
+  const data = (await response.json()) as ModelConcurrencySettings;
+  writeYLog({
+    optType: OptType.view,
+    optDetail: "CM助手-全量-系统设置-模型并发限制查看",
+    url: "/api/system-settings/model-concurrency",
+    returnParam: data,
+  });
+  return data;
+}
+
+export async function updateModelConcurrencySettings(
+  body: ModelConcurrencyUpdate,
+): Promise<ModelConcurrencySettings> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/model-concurrency`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(body),
+    },
+  );
+  if (!response.ok) {
+    const detail = await response
+      .json()
+      .catch(() => ({ detail: "Failed to update model concurrency settings" }));
+    throw new Error(detail.detail ?? "Failed to update model concurrency settings");
+  }
+  const data = (await response.json()) as ModelConcurrencySettings;
+  writeYLog({
+    optType: OptType.edit,
+    optDetail: "CM助手-全量-系统设置-模型并发限制编辑",
+    url: "/api/system-settings/model-concurrency",
+    reqParam: body,
+    returnParam: data,
+  });
+  return data;
+}
+
 export async function updateSkillCompressionSettings(
   body: SkillCompressionUpdate,
 ): Promise<SkillCompressionSettings> {

frontend-web/src/core/system-settings/hooks.ts

@@ -7,6 +7,7 @@ import {
   loadAppearanceSettings,
   loadCitationDisplaySettings,
   loadLeaderboardSettings,
+  loadModelConcurrencySettings,
   loadPrivateSquareStatus,
   loadPromptPrefixSettings,
   loadPublishSquares,
@@ -15,6 +16,7 @@ import {
   updateAppearanceSettings,
   updateCitationDisplaySettings,
   updateLeaderboardSettings,
+  updateModelConcurrencySettings,
   updatePrivateSquareSettings,
   updatePromptPrefixSettings,
   updatePublishSquares,
@@ -22,6 +24,7 @@ import {
   type AppearanceUpdate,
   type CitationDisplayUpdate,
   type LeaderboardSettingsUpdate,
+  type ModelConcurrencyUpdate,
   type PrivateSquareUpdate,
   type PromptPrefixUpdate,
   type PublishSquaresUpdate,
@@ -37,6 +40,7 @@ const APPEARANCE_KEY = ["systemSettings", "appearance"] as const;
 const CITATION_DISPLAY_KEY = ["systemSettings", "citationDisplay"] as const;
 const LEADERBOARD_KEY = ["systemSettings", "leaderboard"] as const;
 const SKILL_COMPRESSION_KEY = ["systemSettings", "skillCompression"] as const;
+const MODEL_CONCURRENCY_KEY = ["systemSettings", "modelConcurrency"] as const;
 
 export function usePromptPrefixSettings() {
   return useQuery({
@@ -238,6 +242,27 @@ export function useUpdateSkillCompressionSettings() {
   });
 }
 
+export function useModelConcurrencySettings() {
+  return useQuery({
+    queryKey: MODEL_CONCURRENCY_KEY,
+    queryFn: () => loadModelConcurrencySettings(),
+    staleTime: 60_000,
+    refetchOnWindowFocus: false,
+    retry: false,
+  });
+}
+
+export function useUpdateModelConcurrencySettings() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: (body: ModelConcurrencyUpdate) =>
+      updateModelConcurrencySettings(body),
+    onSuccess: (data) => {
+      queryClient.setQueryData(MODEL_CONCURRENCY_KEY, data);
+    },
+  });
+}
+
 export function useLeaderboardSettings() {
   return useQuery({
     queryKey: LEADERBOARD_KEY,

frontend-web/src/pages/ModelConcurrencySettingsPage.tsx

@@ -0,0 +1,189 @@
+"use client";
+
+import { Loader2Icon } from "lucide-react";
+import { useEffect, useMemo, useRef, useState } from "react";
+import { toast } from "sonner";
+
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { Switch } from "@/components/ui/switch";
+import { useModels } from "@/core/models/hooks";
+import {
+  useModelConcurrencySettings,
+  useUpdateModelConcurrencySettings,
+} from "@/core/system-settings/hooks";
+
+export default function ModelConcurrencySettingsPage({
+  embedded = false,
+}: {
+  embedded?: boolean;
+} = {}) {
+  const { data, isLoading, error } = useModelConcurrencySettings();
+  const { models } = useModels();
+  const update = useUpdateModelConcurrencySettings();
+
+  const [enabled, setEnabled] = useState(false);
+  const [defaultLimit, setDefaultLimit] = useState(10);
+  const [modelLimits, setModelLimits] = useState<Record<string, string>>({});
+  const initializedRef = useRef(false);
+
+  useEffect(() => {
+    if (!data || initializedRef.current) return;
+    setEnabled(data.enabled);
+    setDefaultLimit(data.default_limit);
+    setModelLimits(
+      Object.fromEntries(
+        Object.entries(data.model_limits ?? {}).map(([name, limit]) => [
+          name,
+          String(limit),
+        ]),
+      ),
+    );
+    initializedRef.current = true;
+  }, [data]);
+
+  const modelNames = useMemo(() => models.map((model) => model.name), [models]);
+
+  const handleModelLimitChange = (modelName: string, value: string) => {
+    setModelLimits((prev) => ({ ...prev, [modelName]: value }));
+  };
+
+  const handleSave = async () => {
+    const nextDefault = Math.max(1, Math.min(Math.round(defaultLimit) || 10, 10000));
+    const cleaned: Record<string, number> = {};
+    for (const [name, rawValue] of Object.entries(modelLimits)) {
+      const trimmed = rawValue.trim();
+      if (!trimmed) continue;
+      const parsed = Math.round(Number(trimmed));
+      if (!Number.isFinite(parsed) || parsed < 1) {
+        toast.error(`模型 ${name} 的并发上限必须是大于 0 的整数`);
+        return;
+      }
+      cleaned[name] = Math.min(parsed, 10000);
+    }
+
+    try {
+      await update.mutateAsync({
+        enabled,
+        default_limit: nextDefault,
+        model_limits: cleaned,
+      });
+      setDefaultLimit(nextDefault);
+      setModelLimits(
+        Object.fromEntries(
+          Object.entries(cleaned).map(([name, limit]) => [name, String(limit)]),
+        ),
+      );
+      toast.success("已保存");
+    } catch (e) {
+      toast.error(e instanceof Error ? e.message : "保存失败");
+    }
+  };
+
+  return (
+    <div
+      className={
+        embedded
+          ? "flex w-full flex-col gap-6"
+          : "mx-auto flex w-full max-w-3xl flex-col gap-6 px-6 py-10"
+      }
+    >
+      <div>
+        <h1 className={embedded ? "text-base font-semibold" : "text-xl font-semibold"}>
+          模型并发限制
+        </h1>
+        <p className="text-muted-foreground mt-1 text-sm">
+          按模型限制同时进行中的问答数。达到上限后，新请求会收到“当前并发已达到上限，请稍后再试”的提示。
+        </p>
+      </div>
+
+      {isLoading ? (
+        <div className="flex items-center justify-center py-16">
+          <Loader2Icon className="text-muted-foreground size-6 animate-spin" />
+        </div>
+      ) : error ? (
+        <div className="text-destructive rounded-xl border border-dashed py-16 text-center text-sm">
+          加载失败：{error.message}
+        </div>
+      ) : (
+        <div className="flex flex-col gap-5 rounded-xl border bg-white px-5 py-5 dark:border-[#213D5E] dark:bg-[#02244C] dark:text-white">
+          <div className="flex items-center justify-between gap-4">
+            <div>
+              <div className="text-sm font-medium">开启限制</div>
+              <div className="text-muted-foreground mt-0.5 text-xs">
+                关闭后，不再按模型并发数拦截问答请求。
+              </div>
+            </div>
+            <Switch checked={enabled} onCheckedChange={setEnabled} />
+          </div>
+
+          <div className="flex flex-col gap-1.5">
+            <p className="text-sm font-medium">默认上限</p>
+            <p className="text-muted-foreground text-xs">
+              未单独配置的模型，使用这个并发上限。
+            </p>
+            <Input
+              type="number"
+              min={1}
+              max={10000}
+              value={defaultLimit}
+              disabled={!enabled}
+              onChange={(e) => setDefaultLimit(Number(e.target.value))}
+              className="w-32"
+            />
+          </div>
+
+          <div className="flex flex-col gap-3">
+            <div>
+              <p className="text-sm font-medium">按模型单独设置</p>
+              <p className="text-muted-foreground mt-0.5 text-xs">
+                留空表示使用默认上限。
+              </p>
+            </div>
+
+            <div className="grid gap-3">
+              {modelNames.length === 0 ? (
+                <div className="text-muted-foreground rounded-lg border border-dashed px-4 py-6 text-sm">
+                  当前没有可配置的模型。
+                </div>
+              ) : (
+                modelNames.map((modelName) => (
+                  <div
+                    key={modelName}
+                    className="flex items-center justify-between gap-4 rounded-lg border px-4 py-3"
+                  >
+                    <div className="min-w-0">
+                      <div className="truncate text-sm font-medium">{modelName}</div>
+                      <div className="text-muted-foreground text-xs">
+                        留空时使用默认上限 {defaultLimit || 10}
+                      </div>
+                    </div>
+                    <Input
+                      type="number"
+                      min={1}
+                      max={10000}
+                      value={modelLimits[modelName] ?? ""}
+                      disabled={!enabled}
+                      onChange={(e) => handleModelLimitChange(modelName, e.target.value)}
+                      className="w-32"
+                      placeholder={String(defaultLimit || 10)}
+                    />
+                  </div>
+                ))
+              )}
+            </div>
+          </div>
+
+          <div className="flex justify-end">
+            <Button onClick={handleSave} disabled={update.isPending}>
+              {update.isPending && (
+                <Loader2Icon className="mr-1.5 size-4 animate-spin" />
+              )}
+              保存
+            </Button>
+          </div>
+        </div>
+      )}
+    </div>
+  );
+}

frontend-web/src/pages/StrategyRoutes.tsx

@@ -51,6 +51,7 @@ import MemoryPage from "./MemoryPage";
 import NewAgentPage from "./NewAgentPage";
 import PromptPrefixSettingsPage from "./PromptPrefixSettingsPage";
 import PrivateSquareSettingsPage from "./PrivateSquareSettingsPage";
+import ModelConcurrencySettingsPage from "./ModelConcurrencySettingsPage";
 import RecommendedQuestionsPage from "./RecommendedQuestionsPage";
 import ScheduledTaskRunDetailPage from "./ScheduledTaskRunDetailPage";
 import ScheduledTasksPage from "./ScheduledTasksPage";
@@ -346,6 +347,16 @@ export default function StrategyRoutes() {
                 )
               }
             />
+            <Route
+              path="settings/model-concurrency"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <ModelConcurrencySettingsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
             <Route
               path="settings/llm-metrics"
               element={

frontend-web/src/pages/WorkspaceRoutes.tsx

@@ -31,6 +31,7 @@ import CheckpointMigrationPage from './CheckpointMigrationPage';
 import MemoryPage from './MemoryPage';
 import PromptPrefixSettingsPage from './PromptPrefixSettingsPage';
 import PrivateSquareSettingsPage from './PrivateSquareSettingsPage';
+import ModelConcurrencySettingsPage from './ModelConcurrencySettingsPage';
 import PublishSquaresSettingsPage from './PublishSquaresSettingsPage';
 import SkillCompressionSettingsPage from './SkillCompressionSettingsPage';
 import SkillEvolutionPage from './SkillEvolutionPage';
@@ -562,6 +563,18 @@ export default function WorkspaceRoutes({
             )
           }
         />
+        <Route
+          path="settings/model-concurrency"
+          element={
+            getStoredAuth()?.system_role === 'admin' ? (
+              <WorkspaceLayout {...layoutProps}>
+                <ModelConcurrencySettingsPage />
+              </WorkspaceLayout>
+            ) : (
+              <Navigate to="/page/workspace/chats/new" replace />
+            )
+          }
+        />
         <Route
           path="settings/publish-squares"
           element={