feat(cluster,api,models,cli): add cluster-aware model catalog deployments and request routing

ts/api/server.ts

@@ -7,13 +7,15 @@
 import * as http from 'node:http';
 import type { IApiConfig } from '../interfaces/config.ts';
 import type { IHealthResponse } from '../interfaces/api.ts';
+import { ClusterCoordinator } from '../cluster/coordinator.ts';
 import { logger } from '../logger.ts';
-import { API_SERVER } from '../constants.ts';
+import { VERSION } from '../constants.ts';
 import { ApiRouter } from './router.ts';
 import { ContainerManager } from '../containers/container-manager.ts';
 import { ModelRegistry } from '../models/registry.ts';
 import { ModelLoader } from '../models/loader.ts';
 import { GpuDetector } from '../hardware/gpu-detector.ts';
+import { ClusterHandler } from './handlers/cluster.ts';
 
 /**
  * API Server for ModelGrid
@@ -26,22 +28,29 @@ export class ApiServer {
   private modelRegistry: ModelRegistry;
   private modelLoader: ModelLoader;
   private gpuDetector: GpuDetector;
+  private clusterCoordinator: ClusterCoordinator;
+  private clusterHandler: ClusterHandler;
   private startTime: number = 0;
 
   constructor(
     config: IApiConfig,
     containerManager: ContainerManager,
     modelRegistry: ModelRegistry,
+    modelLoader: ModelLoader,
+    clusterCoordinator: ClusterCoordinator,
   ) {
     this.config = config;
     this.containerManager = containerManager;
     this.modelRegistry = modelRegistry;
     this.gpuDetector = new GpuDetector();
-    this.modelLoader = new ModelLoader(modelRegistry, containerManager, true);
+    this.modelLoader = modelLoader;
+    this.clusterCoordinator = clusterCoordinator;
+    this.clusterHandler = new ClusterHandler(clusterCoordinator);
     this.router = new ApiRouter(
       containerManager,
       modelRegistry,
       this.modelLoader,
+      clusterCoordinator,
       config.apiKeys,
     );
   }
@@ -120,6 +129,11 @@ export class ApiServer {
     const url = new URL(req.url || '/', `http://${req.headers.host || 'localhost'}`);
     const path = url.pathname;
 
+    if (path.startsWith('/_cluster')) {
+      await this.clusterHandler.handle(req, res, path, url);
+      return;
+    }
+
     // Health check endpoint (no auth required)
     if (path === '/health' || path === '/healthz') {
       await this.handleHealthCheck(res);
@@ -194,7 +208,7 @@ export class ApiServer {
 
       const response: IHealthResponse = {
         status,
-        version: '1.0.0', // TODO: Get from config
+        version: VERSION,
         uptime: Math.floor((Date.now() - this.startTime) / 1000),
         containers: statuses.size,
         models: models.size,
@@ -276,7 +290,6 @@ export class ApiServer {
       error: {
         message,
         type,
-        code: null,
       },
     }));
   }