llm-simple-router 0.2.0 → 0.3.6

package/README.md

@@ -22,6 +22,9 @@
 | 多供应商 | 配置多个后端供应商，按模型映射路由 |
 | 多密钥 (Router Keys) | 为不同使用方创建独立密钥，支持模型白名单 |
 | 流式代理 | 完整支持 SSE 流式和非流式请求 |
+| 供应商并发控制 | 按 Provider 维度限制并发数、队列长度和超时，防止单一供应商过载 |
+| 实时监控 | SSE 推送活跃请求、延迟热力图、Token 吞吐、运行时资源指标 |
+| 代理增强 (实验性) | 注入系统指令、会话记忆、模型锁定等增强功能 |
 | 管理后台 | Vue 3 + shadcn-vue Web UI，管理供应商、映射、密钥 |
 | 请求日志 | 结构化展示完整四阶段链路（客户端请求/上游请求/上游响应/客户端响应），适配 Claude Code 请求格式 |
 | 性能指标 | TTFT、吞吐量、Token 用量、缓存命中率，支持按模型/密钥筛选 |
@@ -34,6 +37,14 @@
 |-----------|-------------|
 | ![Dashboard](docs/screenshot/dashboard.png) | ![Provider](docs/screenshot/provider.png) |
 
+| 供应商并发控制 | 实时监控 |
+|--------------|---------|
+| ![Provider Concurrency](docs/screenshot/provider_concurrency.png) | ![Monitor](docs/screenshot/monitor.png) |
+
+| 代理增强 (实验性) |
+|-----------------|
+| ![Proxy Enhancement](docs/screenshot/proxy_enhance.png) |
+
 | 模型映射 | 重试规则 |
 |---------|--------|
 | ![Model Mapping](docs/screenshot/model_mapping.png) | ![Retry](docs/screenshot/retry.png) |

package/dist/admin/monitor.d.ts

@@ -0,0 +1,7 @@
+import { FastifyPluginCallback } from "fastify";
+import type { RequestTracker } from "../monitor/request-tracker.js";
+interface MonitorRoutesOptions {
+    tracker?: RequestTracker;
+}
+export declare const adminMonitorRoutes: FastifyPluginCallback<MonitorRoutesOptions>;
+export {};

package/dist/admin/monitor.js

@@ -0,0 +1,25 @@
+const HTTP_OK = 200;
+export const adminMonitorRoutes = (app, options, done) => {
+    const { tracker } = options;
+    if (!tracker) {
+        done();
+        return;
+    }
+    app.get("/admin/api/monitor/active", async () => tracker.getActive());
+    app.get("/admin/api/monitor/recent", async () => tracker.getRecent());
+    app.get("/admin/api/monitor/stats", async () => tracker.getStats());
+    app.get("/admin/api/monitor/concurrency", async () => tracker.getConcurrency());
+    app.get("/admin/api/monitor/runtime", async () => tracker.getRuntime());
+    app.get("/admin/api/monitor/stream", (request, reply) => {
+        reply.raw.writeHead(HTTP_OK, {
+            "Content-Type": "text/event-stream",
+            "Cache-Control": "no-cache",
+            Connection: "keep-alive",
+        });
+        tracker.addClient(reply.raw);
+        request.raw.on("close", () => {
+            tracker.removeClient(reply.raw);
+        });
+    });
+    done();
+};

package/dist/admin/providers.d.ts

@@ -1,7 +1,11 @@
 import { FastifyPluginCallback } from "fastify";
 import Database from "better-sqlite3";
+import { ProviderSemaphoreManager } from "../proxy/semaphore.js";
+import type { RequestTracker } from "../monitor/request-tracker.js";
 interface ProviderRoutesOptions {
     db: Database.Database;
+    semaphoreManager?: ProviderSemaphoreManager;
+    tracker?: RequestTracker;
 }
 export declare const adminProviderRoutes: FastifyPluginCallback<ProviderRoutesOptions>;
 export {};

package/dist/admin/providers.js

@@ -1,9 +1,9 @@
 import { Type } from "@sinclair/typebox";
-import { getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, getAllMappingGroups } from "../db/index.js";
+import { getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, getAllMappingGroups, PROVIDER_CONCURRENCY_DEFAULTS } from "../db/index.js";
 import { encrypt, decrypt } from "../utils/crypto.js";
 import { getSetting } from "../db/settings.js";
-import { HTTP_CREATED, HTTP_NOT_FOUND, HTTP_CONFLICT } from "./constants.js";
-const API_KEY_PREVIEW_MIN_LEN = 8;
+import { HTTP_CREATED, HTTP_NOT_FOUND, HTTP_CONFLICT, HTTP_BAD_REQUEST } from "./constants.js";
+const API_KEY_PREVIEW_MIN_LENGTH = 8;
 const API_KEY_PREVIEW_PREFIX_LEN = 4;
 const PROVIDER_NAME_RE = /^[a-zA-Z0-9_-]+$/;
 const CreateProviderSchema = Type.Object({
@@ -13,6 +13,9 @@ const CreateProviderSchema = Type.Object({
     api_key: Type.String({ minLength: 1 }),
     models: Type.Optional(Type.Array(Type.String())),
     is_active: Type.Optional(Type.Number()),
+    max_concurrency: Type.Optional(Type.Integer({ minimum: 0 })),
+    queue_timeout_ms: Type.Optional(Type.Integer({ minimum: 0 })),
+    max_queue_size: Type.Optional(Type.Integer({ minimum: 1 })),
 });
 const UpdateProviderSchema = Type.Object({
     name: Type.Optional(Type.String({ minLength: 1 })),
@@ -21,9 +24,12 @@ const UpdateProviderSchema = Type.Object({
     api_key: Type.Optional(Type.String({ minLength: 1 })),
     models: Type.Optional(Type.Array(Type.String())),
     is_active: Type.Optional(Type.Number()),
+    max_concurrency: Type.Optional(Type.Integer({ minimum: 0 })),
+    queue_timeout_ms: Type.Optional(Type.Integer({ minimum: 0 })),
+    max_queue_size: Type.Optional(Type.Integer({ minimum: 1 })),
 });
 export const adminProviderRoutes = (app, options, done) => {
-    const { db } = options;
+    const { db, semaphoreManager, tracker } = options;
     app.get("/admin/api/providers", async (_request, reply) => {
         const encryptionKey = getSetting(db, "encryption_key");
         const providers = getAllProviders(db);
@@ -35,6 +41,10 @@ export const adminProviderRoutes = (app, options, done) => {
             api_key: s.api_key ? decrypt(s.api_key, encryptionKey) : "",
             models: JSON.parse(s.models || "[]"),
             is_active: s.is_active,
+            max_concurrency: s.max_concurrency,
+            queue_timeout_ms: s.queue_timeout_ms,
+            max_queue_size: s.max_queue_size,
+            concurrency_status: semaphoreManager?.getStatus(s.id) ?? { active: 0, queued: 0 },
             created_at: s.created_at,
             updated_at: s.updated_at,
         })));
@@ -42,7 +52,7 @@ export const adminProviderRoutes = (app, options, done) => {
     app.post("/admin/api/providers", { schema: { body: CreateProviderSchema } }, async (request, reply) => {
         const body = request.body;
         if (!PROVIDER_NAME_RE.test(body.name)) {
-            return reply.status(400).send({ error: { message: "Provider 名称仅允许英文大小写字母、数字、横线和下划线" } });
+            return reply.status(HTTP_BAD_REQUEST).send({ error: { message: "Provider 名称仅允许英文大小写字母、数字、横线和下划线" } });
         }
         const encryptedKey = encrypt(body.api_key, getSetting(db, "encryption_key"));
         const id = createProvider(db, {
@@ -50,9 +60,23 @@ export const adminProviderRoutes = (app, options, done) => {
             api_type: body.api_type,
             base_url: body.base_url,
             api_key: encryptedKey,
-            api_key_preview: body.api_key.length > 8 ? `${body.api_key.slice(0, 4)}...${body.api_key.slice(-4)}` : "****",
+            api_key_preview: body.api_key.length > API_KEY_PREVIEW_MIN_LENGTH ? `${body.api_key.slice(0, API_KEY_PREVIEW_PREFIX_LEN)}...${body.api_key.slice(-API_KEY_PREVIEW_PREFIX_LEN)}` : "****",
             models: JSON.stringify(body.models ?? []),
             is_active: body.is_active ?? 1,
+            max_concurrency: body.max_concurrency ?? PROVIDER_CONCURRENCY_DEFAULTS.max_concurrency,
+            queue_timeout_ms: body.queue_timeout_ms ?? PROVIDER_CONCURRENCY_DEFAULTS.queue_timeout_ms,
+            max_queue_size: body.max_queue_size ?? PROVIDER_CONCURRENCY_DEFAULTS.max_queue_size,
+        });
+        semaphoreManager?.updateConfig(id, {
+            maxConcurrency: body.max_concurrency ?? PROVIDER_CONCURRENCY_DEFAULTS.max_concurrency,
+            queueTimeoutMs: body.queue_timeout_ms ?? PROVIDER_CONCURRENCY_DEFAULTS.queue_timeout_ms,
+            maxQueueSize: body.max_queue_size ?? PROVIDER_CONCURRENCY_DEFAULTS.max_queue_size,
+        });
+        tracker?.updateProviderConfig(id, {
+            name: body.name,
+            maxConcurrency: body.max_concurrency ?? PROVIDER_CONCURRENCY_DEFAULTS.max_concurrency,
+            queueTimeoutMs: body.queue_timeout_ms ?? PROVIDER_CONCURRENCY_DEFAULTS.queue_timeout_ms,
+            maxQueueSize: body.max_queue_size ?? PROVIDER_CONCURRENCY_DEFAULTS.max_queue_size,
         });
         return reply.code(HTTP_CREATED).send({ id });
     });
@@ -64,7 +88,7 @@ export const adminProviderRoutes = (app, options, done) => {
         }
         const body = request.body;
         if (body.name !== undefined && !PROVIDER_NAME_RE.test(body.name)) {
-            return reply.status(400).send({ error: { message: "Provider 名称仅允许英文大小写字母、数字、横线和下划线" } });
+            return reply.status(HTTP_BAD_REQUEST).send({ error: { message: "Provider 名称仅允许英文大小写字母、数字、横线和下划线" } });
         }
         const fields = {};
         if (body.name !== undefined)
@@ -77,11 +101,31 @@ export const adminProviderRoutes = (app, options, done) => {
             fields.is_active = body.is_active;
         if (body.models !== undefined)
             fields.models = JSON.stringify(body.models);
+        if (body.max_concurrency !== undefined)
+            fields.max_concurrency = body.max_concurrency;
+        if (body.queue_timeout_ms !== undefined)
+            fields.queue_timeout_ms = body.queue_timeout_ms;
+        if (body.max_queue_size !== undefined)
+            fields.max_queue_size = body.max_queue_size;
         if (body.api_key) {
             fields.api_key = encrypt(body.api_key, getSetting(db, "encryption_key"));
-            fields.api_key_preview = body.api_key.length > 8 ? `${body.api_key.slice(0, 4)}...${body.api_key.slice(-4)}` : "****";
+            fields.api_key_preview = body.api_key.length > API_KEY_PREVIEW_MIN_LENGTH ? `${body.api_key.slice(0, API_KEY_PREVIEW_PREFIX_LEN)}...${body.api_key.slice(-API_KEY_PREVIEW_PREFIX_LEN)}` : "****";
         }
         updateProvider(db, id, fields);
+        const updated = getProviderById(db, id);
+        if (body.max_concurrency !== undefined || body.queue_timeout_ms !== undefined || body.max_queue_size !== undefined) {
+            semaphoreManager?.updateConfig(id, {
+                maxConcurrency: updated.max_concurrency,
+                queueTimeoutMs: updated.queue_timeout_ms,
+                maxQueueSize: updated.max_queue_size,
+            });
+        }
+        tracker?.updateProviderConfig(id, {
+            name: body.name ?? existing.name,
+            maxConcurrency: updated.max_concurrency,
+            queueTimeoutMs: updated.queue_timeout_ms,
+            maxQueueSize: updated.max_queue_size,
+        });
         return reply.send({ success: true });
     });
     app.delete("/admin/api/providers/:id", async (request, reply) => {
@@ -95,9 +139,13 @@ export const adminProviderRoutes = (app, options, done) => {
                     return reply.code(HTTP_CONFLICT).send({ error: { message: `Provider is referenced by mapping group '${g.client_model}'` } });
                 }
             }
-            catch { /* rule format invalid, skip */ }
+            catch {
+                continue;
+            }
         }
         deleteProvider(db, id);
+        semaphoreManager?.remove(id);
+        tracker?.removeProviderConfig(id);
         return reply.send({ success: true });
     });
     done();

package/dist/admin/retry-rules.js

@@ -1,6 +1,9 @@
 import { Type } from "@sinclair/typebox";
 import { getAllRetryRules, createRetryRule, updateRetryRule, deleteRetryRule, } from "../db/index.js";
 import { HTTP_BAD_REQUEST, HTTP_CREATED } from "./constants.js";
+const DEFAULT_RETRY_DELAY_MS = 5000;
+const DEFAULT_MAX_RETRIES = 10;
+const DEFAULT_MAX_DELAY_MS = 60000;
 const CreateRetryRuleSchema = Type.Object({
     name: Type.String({ minLength: 1 }),
     status_code: Type.Number({ minimum: 100, maximum: 599 }),
@@ -52,9 +55,9 @@ export const adminRetryRuleRoutes = (app, options, done) => {
             body_pattern: body.body_pattern,
             is_active: body.is_active ?? 1,
             retry_strategy: body.retry_strategy ?? "exponential",
-            retry_delay_ms: body.retry_delay_ms ?? 5000,
-            max_retries: body.max_retries ?? 10,
-            max_delay_ms: body.max_delay_ms ?? 60000,
+            retry_delay_ms: body.retry_delay_ms ?? DEFAULT_RETRY_DELAY_MS,
+            max_retries: body.max_retries ?? DEFAULT_MAX_RETRIES,
+            max_delay_ms: body.max_delay_ms ?? DEFAULT_MAX_DELAY_MS,
         });
         refreshMatcher(matcher, db);
         return reply.code(HTTP_CREATED).send({ id });

package/dist/admin/routes.d.ts

@@ -1,9 +1,13 @@
 import { FastifyPluginCallback } from "fastify";
 import Database from "better-sqlite3";
 import { RetryRuleMatcher } from "../proxy/retry-rules.js";
+import type { RequestTracker } from "../monitor/request-tracker.js";
+import { ProviderSemaphoreManager } from "../proxy/semaphore.js";
 interface AdminRoutesOptions {
     db: Database.Database;
     matcher: RetryRuleMatcher | null;
+    tracker?: RequestTracker;
+    semaphoreManager?: ProviderSemaphoreManager;
 }
 export declare const adminRoutes: FastifyPluginCallback<AdminRoutesOptions>;
 export {};

package/dist/admin/routes.js

@@ -9,12 +9,13 @@ import { adminMetricsRoutes } from "./metrics.js";
 import { adminProxyEnhancementRoutes } from "./proxy-enhancement.js";
 import { adminRouterKeyRoutes } from "./router-keys.js";
 import { adminSetupRoutes } from "./setup.js";
+import { adminMonitorRoutes } from "./monitor.js";
 export const adminRoutes = (app, options, done) => {
     // Setup 路由不需要 auth
     app.register(adminSetupRoutes, { db: options.db });
     app.register(adminAuthPlugin, { db: options.db });
     app.register(adminLoginRoutes, { db: options.db });
-    app.register(adminProviderRoutes, { db: options.db });
+    app.register(adminProviderRoutes, { db: options.db, semaphoreManager: options.semaphoreManager, tracker: options.tracker });
     app.register(adminMappingRoutes, { db: options.db });
     app.register(adminGroupRoutes, { db: options.db });
     app.register(adminRetryRuleRoutes, { db: options.db, matcher: options.matcher });
@@ -23,5 +24,6 @@ export const adminRoutes = (app, options, done) => {
     app.register(adminStatsRoutes, { db: options.db });
     app.register(adminMetricsRoutes, { db: options.db });
     app.register(adminProxyEnhancementRoutes, { db: options.db });
+    app.register(adminMonitorRoutes, { tracker: options.tracker });
     done();
 };

package/dist/admin/setup.js

@@ -2,6 +2,9 @@ import { randomBytes } from "node:crypto";
 import jwt from "jsonwebtoken";
 import { getSetting, setSetting, isInitialized } from "../db/settings.js";
 import { hashPassword } from "../utils/password.js";
+import { HTTP_BAD_REQUEST, HTTP_CONFLICT } from "./constants.js";
+const CRYPTO_BYTES_LENGTH = 32;
+const MIN_PASSWORD_LENGTH = 6;
 export const adminSetupRoutes = (app, options, done) => {
     const { db } = options;
     app.get("/admin/api/setup/status", async () => {
@@ -9,15 +12,15 @@ export const adminSetupRoutes = (app, options, done) => {
     });
     app.post("/admin/api/setup/initialize", async (request, reply) => {
         const { password } = request.body;
-        if (!password || password.length < 6) { // eslint-disable-line no-magic-numbers
-            return reply.code(400).send({ error: { message: "Password must be at least 6 characters" } });
+        if (!password || password.length < MIN_PASSWORD_LENGTH) {
+            return reply.code(HTTP_BAD_REQUEST).send({ error: { message: `Password must be at least ${MIN_PASSWORD_LENGTH} characters` } });
         }
         // 事务中原子检查防竞态
         const alreadyInitialized = db.transaction(() => {
             if (isInitialized(db))
                 return true;
-            const encryptionKey = randomBytes(32).toString("hex");
-            const jwtSecret = randomBytes(32).toString("hex");
+            const encryptionKey = randomBytes(CRYPTO_BYTES_LENGTH).toString("hex");
+            const jwtSecret = randomBytes(CRYPTO_BYTES_LENGTH).toString("hex");
             setSetting(db, "admin_password_hash", hashPassword(password));
             setSetting(db, "encryption_key", encryptionKey);
             setSetting(db, "jwt_secret", jwtSecret);
@@ -25,7 +28,7 @@ export const adminSetupRoutes = (app, options, done) => {
             return false;
         })();
         if (alreadyInitialized) {
-            return reply.code(409).send({ error: { message: "Already initialized" } });
+            return reply.code(HTTP_CONFLICT).send({ error: { message: "Already initialized" } });
         }
         // 自动登录：签发 JWT
         const TOKEN_EXPIRY_SECONDS = 172800; // 48 hours，与 admin-auth 保持一致

package/dist/db/index.d.ts

@@ -1,6 +1,6 @@
 import Database from "better-sqlite3";
 export declare function initDatabase(dbPath: string): Database.Database;
-export { getActiveProviders, getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, } from "./providers.js";
+export { getActiveProviders, getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, PROVIDER_CONCURRENCY_DEFAULTS, } from "./providers.js";
 export type { Provider } from "./providers.js";
 export { getModelMapping, getAllModelMappings, createModelMapping, updateModelMapping, deleteModelMapping, getMappingGroup, getMappingGroupById, getAllMappingGroups, createMappingGroup, updateMappingGroup, deleteMappingGroup, getActiveProviderModels, resolveByProviderModel, } from "./mappings.js";
 export type { ModelMapping, MappingGroup, ProviderModelEntry } from "./mappings.js";

package/dist/db/index.js

@@ -36,7 +36,7 @@ export function initDatabase(dbPath) {
     return db;
 }
 // --- Re-export from per-table modules ---
-export { getActiveProviders, getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, } from "./providers.js";
+export { getActiveProviders, getAllProviders, getProviderById, createProvider, updateProvider, deleteProvider, PROVIDER_CONCURRENCY_DEFAULTS, } from "./providers.js";
 export { getModelMapping, getAllModelMappings, createModelMapping, updateModelMapping, deleteModelMapping, getMappingGroup, getMappingGroupById, getAllMappingGroups, createMappingGroup, updateMappingGroup, deleteMappingGroup, getActiveProviderModels, resolveByProviderModel, } from "./mappings.js";
 export { getActiveRetryRules, getAllRetryRules, createRetryRule, updateRetryRule, deleteRetryRule, seedDefaultRules, } from "./retry-rules.js";
 export { insertRequestLog, getRequestLogs, getRequestLogById, deleteLogsBefore, insertMetrics, } from "./logs.js";

package/dist/db/mappings.js

@@ -64,7 +64,9 @@ export function getActiveProviderModels(db) {
                 results.push({ provider_name: p.name, backend_model: m });
             }
         }
-        catch { /* 忽略解析失败 */ }
+        catch {
+            continue;
+        }
     }
     return results;
 }
@@ -120,7 +122,9 @@ export function resolveByProviderModel(db, providerName, backendModel) {
                 return { client_model: g.client_model, provider_id: providerRow.id, backend_model: backendModel };
             }
         }
-        catch { /* continue */ }
+        catch {
+            continue;
+        }
     }
     // provider 有这个模型但没有 mapping group，直接返回 provider 维度信息
     return { client_model: backendModel, provider_id: providerRow.id, backend_model: backendModel };

package/dist/db/migrations/017_add_provider_concurrency.sql

@@ -0,0 +1,3 @@
+ALTER TABLE providers ADD COLUMN max_concurrency INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE providers ADD COLUMN queue_timeout_ms INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE providers ADD COLUMN max_queue_size INTEGER NOT NULL DEFAULT 100;

package/dist/db/providers.d.ts

@@ -8,9 +8,17 @@ export interface Provider {
     api_key_preview?: string;
     models: string;
     is_active: number;
+    max_concurrency: number;
+    queue_timeout_ms: number;
+    max_queue_size: number;
     created_at: string;
     updated_at: string;
 }
+export declare const PROVIDER_CONCURRENCY_DEFAULTS: {
+    readonly max_concurrency: 0;
+    readonly queue_timeout_ms: 0;
+    readonly max_queue_size: 100;
+};
 export declare function getActiveProviders(db: Database.Database, apiType: "openai" | "anthropic"): Provider[];
 export declare function getAllProviders(db: Database.Database): Provider[];
 export declare function getProviderById(db: Database.Database, id: string): Provider | undefined;
@@ -22,6 +30,9 @@ export declare function createProvider(db: Database.Database, provider: {
     api_key_preview?: string;
     models?: string;
     is_active?: number;
+    max_concurrency?: number;
+    queue_timeout_ms?: number;
+    max_queue_size?: number;
 }): string;
-export declare function updateProvider(db: Database.Database, id: string, fields: Partial<Pick<Provider, "name" | "api_type" | "base_url" | "api_key" | "api_key_preview" | "is_active">>): void;
+export declare function updateProvider(db: Database.Database, id: string, fields: Partial<Pick<Provider, "name" | "api_type" | "base_url" | "api_key" | "api_key_preview" | "models" | "is_active" | "max_concurrency" | "queue_timeout_ms" | "max_queue_size">>): void;
 export declare function deleteProvider(db: Database.Database, id: string): void;

package/dist/db/providers.js

@@ -1,7 +1,12 @@
 import { randomUUID } from "crypto";
 import { buildUpdateQuery, deleteById } from "./helpers.js";
+export const PROVIDER_CONCURRENCY_DEFAULTS = {
+    max_concurrency: 0,
+    queue_timeout_ms: 0,
+    max_queue_size: 100,
+};
 const PROVIDER_FIELDS = new Set([
-    "name", "api_type", "base_url", "api_key", "api_key_preview", "models", "is_active",
+    "name", "api_type", "base_url", "api_key", "api_key_preview", "models", "is_active", "max_concurrency", "queue_timeout_ms", "max_queue_size",
 ]);
 export function getActiveProviders(db, apiType) {
     return db
@@ -17,8 +22,8 @@ export function getProviderById(db, id) {
 export function createProvider(db, provider) {
     const id = randomUUID();
     const now = new Date().toISOString();
-    db.prepare(`INSERT INTO providers (id, name, api_type, base_url, api_key, api_key_preview, models, is_active, created_at, updated_at)
-     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`).run(id, provider.name, provider.api_type, provider.base_url, provider.api_key, provider.api_key_preview ?? null, provider.models ?? "[]", provider.is_active ?? 1, now, now);
+    db.prepare(`INSERT INTO providers (id, name, api_type, base_url, api_key, api_key_preview, models, is_active, max_concurrency, queue_timeout_ms, max_queue_size, created_at, updated_at)
+     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`).run(id, provider.name, provider.api_type, provider.base_url, provider.api_key, provider.api_key_preview ?? null, provider.models ?? "[]", provider.is_active ?? 1, provider.max_concurrency ?? PROVIDER_CONCURRENCY_DEFAULTS.max_concurrency, provider.queue_timeout_ms ?? PROVIDER_CONCURRENCY_DEFAULTS.queue_timeout_ms, provider.max_queue_size ?? PROVIDER_CONCURRENCY_DEFAULTS.max_queue_size, now, now);
     return id;
 }
 export function updateProvider(db, id, fields) {

package/dist/db/retry-rules.js

@@ -1,6 +1,9 @@
 import { randomUUID } from "crypto";
 import { buildUpdateQuery, deleteById } from "./helpers.js";
 const RETRY_FIELDS = new Set(["name", "status_code", "body_pattern", "is_active", "retry_strategy", "retry_delay_ms", "max_retries", "max_delay_ms"]);
+const DEFAULT_RETRY_DELAY_MS = 5000;
+const DEFAULT_MAX_RETRIES = 10;
+const DEFAULT_MAX_DELAY_MS = 60000;
 export function getActiveRetryRules(db) {
     return db
         .prepare("SELECT * FROM retry_rules WHERE is_active = 1 ORDER BY created_at DESC")
@@ -15,7 +18,7 @@ export function createRetryRule(db, rule) {
     const id = randomUUID();
     const now = new Date().toISOString();
     db.prepare(`INSERT INTO retry_rules (id, name, status_code, body_pattern, is_active, created_at, retry_strategy, retry_delay_ms, max_retries, max_delay_ms)
-     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`).run(id, rule.name, rule.status_code, rule.body_pattern, rule.is_active ?? 1, now, rule.retry_strategy ?? "exponential", rule.retry_delay_ms ?? 5000, rule.max_retries ?? 10, rule.max_delay_ms ?? 60000);
+     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`).run(id, rule.name, rule.status_code, rule.body_pattern, rule.is_active ?? 1, now, rule.retry_strategy ?? "exponential", rule.retry_delay_ms ?? DEFAULT_RETRY_DELAY_MS, rule.max_retries ?? DEFAULT_MAX_RETRIES, rule.max_delay_ms ?? DEFAULT_MAX_DELAY_MS);
     return id;
 }
 export function updateRetryRule(db, id, fields) {

package/dist/db/router-keys.js

@@ -30,7 +30,9 @@ export function getAvailableModels(db) {
         try {
             JSON.parse(r.models || "[]").forEach((m) => set.add(m));
         }
-        catch { /* skip invalid JSON */ }
+        catch {
+            continue;
+        }
     }
     return [...set].sort();
 }

package/dist/index.js

@@ -6,6 +6,10 @@ import { randomUUID } from "crypto";
 import Fastify from "fastify";
 import { insertRequestLog } from "./db/logs.js";
 const HTTP_NOT_FOUND = 404;
+const HTTP_INTERNAL_ERROR = 500;
+const HTTP_BAD_REQUEST = 400;
+const PROVIDER_DEFAULT_QUEUE_TIMEOUT_MS = 5000;
+const PROVIDER_DEFAULT_MAX_QUEUE_SIZE = 100;
 // 代理路由路径 → api_type，用于在全局 hook/errorHandler 中识别代理请求
 const PROXY_API_TYPES = {
     "/v1/chat/completions": "openai",
@@ -19,12 +23,14 @@ function getProxyApiType(url) {
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
 import { getConfig } from "./config.js";
-import { initDatabase, seedDefaultRules } from "./db/index.js";
+import { initDatabase, seedDefaultRules, getAllProviders } from "./db/index.js";
 import { authMiddleware } from "./middleware/auth.js";
 import { openaiProxy } from "./proxy/openai.js";
 import { anthropicProxy } from "./proxy/anthropic.js";
 import { adminRoutes } from "./admin/routes.js";
 import { RetryRuleMatcher } from "./proxy/retry-rules.js";
+import { ProviderSemaphoreManager } from "./proxy/semaphore.js";
+import { RequestTracker } from "./monitor/request-tracker.js";
 import { modelState } from "./proxy/model-state.js";
 import fastifyStatic from "@fastify/static";
 export async function buildApp(options) {
@@ -72,7 +78,7 @@ export async function buildApp(options) {
     // 统一 schema validation 错误响应格式，代理路由的错误也记录到 request_logs
     app.setErrorHandler((error, request, reply) => {
         const fastifyError = error;
-        const status = fastifyError.statusCode ?? 500;
+        const status = fastifyError.statusCode ?? HTTP_INTERNAL_ERROR;
         const proxyApiType = getProxyApiType(request.url);
         if (proxyApiType) {
             request.log.error({ statusCode: status, err: error }, `Proxy request error: ${fastifyError.message}`);
@@ -91,8 +97,8 @@ export async function buildApp(options) {
                 router_key_id: request.routerKey?.id ?? null,
             });
         }
-        if (status === 400 && fastifyError.validation) {
-            return reply.code(400).send({ error: { message: fastifyError.message } });
+        if (status === HTTP_BAD_REQUEST && fastifyError.validation) {
+            return reply.code(HTTP_BAD_REQUEST).send({ error: { message: fastifyError.message } });
         }
         return reply.code(status).send({ error: { message: fastifyError.message } });
     });
@@ -102,6 +108,26 @@ export async function buildApp(options) {
     modelState.init(db);
     const matcher = new RetryRuleMatcher();
     matcher.load(db);
+    const semaphoreManager = new ProviderSemaphoreManager();
+    const tracker = new RequestTracker({ semaphoreManager });
+    tracker.startPushInterval();
+    // 从 DB 读取已有 provider 的并发配置，初始化信号量管理器和 tracker
+    const allProviders = getAllProviders(db);
+    for (const p of allProviders) {
+        if (p.max_concurrency > 0) {
+            semaphoreManager.updateConfig(p.id, {
+                maxConcurrency: p.max_concurrency,
+                queueTimeoutMs: p.queue_timeout_ms,
+                maxQueueSize: p.max_queue_size,
+            });
+        }
+        tracker.updateProviderConfig(p.id, {
+            name: p.name,
+            maxConcurrency: p.max_concurrency ?? 0,
+            queueTimeoutMs: p.queue_timeout_ms ?? PROVIDER_DEFAULT_QUEUE_TIMEOUT_MS,
+            maxQueueSize: p.max_queue_size ?? PROVIDER_DEFAULT_MAX_QUEUE_SIZE,
+        });
+    }
     app.register(authMiddleware, { db });
     app.register(openaiProxy, {
         db,
@@ -109,6 +135,8 @@ export async function buildApp(options) {
         retryMaxAttempts: config.RETRY_MAX_ATTEMPTS,
         retryBaseDelayMs: config.RETRY_BASE_DELAY_MS,
         matcher,
+        semaphoreManager,
+        tracker,
     });
     app.register(anthropicProxy, {
         db,
@@ -116,8 +144,10 @@ export async function buildApp(options) {
         retryMaxAttempts: config.RETRY_MAX_ATTEMPTS,
         retryBaseDelayMs: config.RETRY_BASE_DELAY_MS,
         matcher,
+        semaphoreManager,
+        tracker,
     });
-    app.register(adminRoutes, { db, matcher });
+    app.register(adminRoutes, { db, matcher, tracker, semaphoreManager });
     // 前端静态文件服务（生产环境）
     const frontendDist = path.resolve(process.env.FRONTEND_DIST || path.join(__dirname, "../frontend-dist"));
     if (existsSync(frontendDist)) {
@@ -145,6 +175,7 @@ export async function buildApp(options) {
         app,
         db,
         close: async () => {
+            tracker.stopPushInterval();
             await app.close();
             db.close();
         },

package/dist/metrics/sse-metrics-transform.d.ts

@@ -1,5 +1,14 @@
 import { Transform, TransformCallback } from "stream";
 import { MetricsExtractor } from "./metrics-extractor.js";
+import type { MetricsResult } from "./metrics-extractor.js";
+export interface MetricsTransformOptions {
+    /** 每次处理 SSE 事件后触发的回调，附带当前指标快照 */
+    onMetrics?: (metrics: MetricsResult) => void;
+    /** 每收到一个 SSE data 行时触发，传入原始文本行 */
+    onChunk?: (rawLine: string) => void;
+    /** 回调节流间隔（毫秒），默认 5000 */
+    throttleMs?: number;
+}
 /**
  * 旁路采集 SSE 指标的 Transform stream
  *
@@ -9,8 +18,15 @@ import { MetricsExtractor } from "./metrics-extractor.js";
 export declare class SSEMetricsTransform extends Transform {
     private parser;
     private extractor;
-    constructor(apiType: "openai" | "anthropic", requestStartTime: number);
+    private onMetrics?;
+    private onChunk?;
+    private throttleMs;
+    private lastCallbackTime;
+    private flushed;
+    constructor(apiType: "openai" | "anthropic", requestStartTime: number, options?: MetricsTransformOptions);
     _transform(chunk: Buffer, _encoding: BufferEncoding, callback: TransformCallback): void;
     _flush(callback: TransformCallback): void;
     getExtractor(): MetricsExtractor;
+    /** 节流逻辑：首次或距上次回调超过 throttleMs 时触发 */
+    private emitMetricsIfReady;
 }

package/dist/metrics/sse-metrics-transform.js

@@ -1,6 +1,7 @@
 import { Transform } from "stream";
 import { SSEParser } from "./sse-parser.js";
 import { MetricsExtractor } from "./metrics-extractor.js";
+const DEFAULT_THROTTLE_MS = 5000;
 /**
  * 旁路采集 SSE 指标的 Transform stream
  *
@@ -10,16 +11,30 @@ import { MetricsExtractor } from "./metrics-extractor.js";
 export class SSEMetricsTransform extends Transform {
     parser;
     extractor;
-    constructor(apiType, requestStartTime) {
+    onMetrics;
+    onChunk;
+    throttleMs;
+    lastCallbackTime = 0;
+    flushed = false;
+    constructor(apiType, requestStartTime, options) {
         super();
         this.parser = new SSEParser();
         this.extractor = new MetricsExtractor(apiType, requestStartTime);
+        this.onMetrics = options?.onMetrics;
+        this.onChunk = options?.onChunk;
+        this.throttleMs = options?.throttleMs ?? DEFAULT_THROTTLE_MS;
     }
     _transform(chunk, _encoding, callback) {
-        const events = this.parser.feed(chunk.toString("utf-8"));
+        const text = chunk.toString("utf-8");
+        const events = this.parser.feed(text);
         for (const event of events) {
             this.extractor.processEvent(event);
+            // 将解析后的事件还原为 SSE data 行格式传给 onChunk
+            if (event.data != null && this.onChunk) {
+                this.onChunk(`data: ${event.data}`);
+            }
         }
+        this.emitMetricsIfReady();
         callback(null, chunk);
     }
     _flush(callback) {
@@ -27,9 +42,25 @@ export class SSEMetricsTransform extends Transform {
         for (const event of events) {
             this.extractor.processEvent(event);
         }
+        // flush 无条件推送最终状态，确保消费者能拿到完整指标
+        if (this.onMetrics && !this.flushed) {
+            this.flushed = true;
+            this.lastCallbackTime = Date.now();
+            this.onMetrics(this.extractor.getMetrics());
+        }
         callback();
     }
     getExtractor() {
         return this.extractor;
     }
+    /** 节流逻辑：首次或距上次回调超过 throttleMs 时触发 */
+    emitMetricsIfReady() {
+        if (!this.onMetrics)
+            return;
+        const now = Date.now();
+        if (now - this.lastCallbackTime >= this.throttleMs) {
+            this.lastCallbackTime = now;
+            this.onMetrics(this.extractor.getMetrics());
+        }
+    }
 }