fix: pull corresponding engine base on model

Author: louis-jan

cortex-js/src/infrastructure/commanders/models/model-pull.command.ts

@@ -37,6 +37,14 @@ export class ModelPullCommand extends CommandRunner {
 
     const existingModel = await this.modelsCliUsecases.getModel(modelId);
     const engine = existingModel?.engine || 'cortex.llamacpp';
+
+    await this.modelsCliUsecases.pullModel(modelId).catch((e: Error) => {
+      if (e instanceof ModelNotFoundException)
+        console.error('Model does not exist.');
+      else console.error(e);
+      exit(1);
+    });
+
     // Pull engine if not exist
     if (
       !existsSync(
@@ -45,16 +53,11 @@ export class ModelPullCommand extends CommandRunner {
     ) {
       await this.initUsecases.installEngine(
         await this.initUsecases.defaultInstallationOptions(),
+        'latest',
+        engine,
       );
     }
 
-    await this.modelsCliUsecases.pullModel(modelId).catch((e: Error) => {
-      if (e instanceof ModelNotFoundException)
-        console.error('Model does not exist.');
-      else console.error(e);
-      exit(1);
-    });
-
     console.log('\nDownload complete!');
     exit(0);
   }

cortex-js/src/infrastructure/commanders/models/model-start.command.ts

@@ -69,6 +69,8 @@ export class ModelStartCommand extends CommandRunner {
     ) {
       await this.initUsecases.installEngine(
         await this.initUsecases.defaultInstallationOptions(),
+        'latest',
+        engine,
       );
     }
     await this.cortexUsecases

cortex-js/src/infrastructure/commanders/usecases/init.cli.usecases.ts

@@ -52,12 +52,35 @@ export class InitCliUsecases {
   installEngine = async (
     options: InitOptions,
     version: string = 'latest',
+    engine: string = 'default',
     force: boolean = true,
   ): Promise<any> => {
     const configs = await this.fileManagerService.getConfig();
 
     if (configs.initialized && !force) return;
 
+    if (engine === 'default' || engine === 'cortex.llamacpp')
+      // Ship Llama.cpp engine by default
+      await this.installLlamaCppEngine(options, version);
+
+    if (engine === 'cortex.onnx')
+      if (process.platform === 'win32')
+        // Ship ONNX Runtime on Windows by default
+        await this.installONNXEngine();
+
+    configs.initialized = true;
+    await this.fileManagerService.writeConfigFile(configs);
+  };
+
+  /**
+   * Install Llama.cpp engine
+   * @param options
+   * @param version
+   */
+  private installLlamaCppEngine = async (
+    options: InitOptions,
+    version: string = 'latest',
+  ) => {
     const engineFileName = this.parseEngineFileName(options);
 
     const res = await firstValueFrom(
@@ -148,12 +171,6 @@ export class InitCliUsecases {
     if (options.runMode === 'GPU' && !(await cudaVersion())) {
       await this.installCudaToolkitDependency(options.cudaVersion);
     }
-
-    // Ship ONNX Runtime on Windows by default
-    if (process.platform === 'win32') await this.installONNXEngine();
-
-    configs.initialized = true;
-    await this.fileManagerService.writeConfigFile(configs);
   };
 
   /**