@aws/ml-container-creator 0.2.3 → 0.2.5

package/README.md

@@ -15,7 +15,7 @@ A CLI tool that creates SageMaker-compatible Docker containers for deploying ML
 
 | Deployment Target | Description |
 |---|---|
-| Managed Inference | SageMaker real-time endpoints |
+| Real-Time Inference | SageMaker real-time endpoints |
 | Async Inference | SageMaker async endpoints with S3 output |
 | Batch Transform | SageMaker batch processing |
 | HyperPod EKS | Kubernetes-based deployment |

package/bin/cli.js

@@ -27,7 +27,8 @@ program
 
     // --- General ---
     .addOption(new Option('--skip-prompts', 'Skip interactive prompts and use configuration from other sources'))
-    .addOption(new Option('--config <path>', 'Path to configuration file'))
+    .addOption(new Option('--auto-prompt', 'Fill defaults, prompt only for missing required values'))
+    .addOption(new Option('--config <path>', 'Path to JSON configuration file'))
     .addOption(new Option('--project-name <name>', 'Project name'))
     .addOption(new Option('--project-dir <dir>', 'Output directory path'))
     .addOption(new Option('--force', 'Overwrite existing output directory without prompting'))
@@ -41,7 +42,7 @@ program
     .addOption(new Option('--base-image <image>', 'Base container image for Dockerfile'))
 
     // --- Build & Infrastructure ---
-    .addOption(new Option('--deployment-target <target>', 'Deployment target (managed-inference, async-inference, batch-transform, hyperpod-eks)'))
+    .addOption(new Option('--deployment-target <target>', 'Deployment target (realtime-inference, async-inference, batch-transform, hyperpod-eks)'))
     .addOption(new Option('--instance-type <type>', 'SageMaker instance type (e.g. ml.g5.xlarge, ml.m5.large)'))
     .addOption(new Option('--region <region>', 'AWS region'))
     .addOption(new Option('--role-arn <arn>', 'IAM role ARN for SageMaker execution'))
@@ -154,7 +155,7 @@ program.configureHelp({
 
         for (const opt of allOptions) {
             const long = opt.long || '';
-            if (['--skip-prompts', '--config', '--project-name', '--project-dir', '--force', '--version', '--help'].includes(long)) {
+            if (['--skip-prompts', '--auto-prompt', '--config', '--project-name', '--project-dir', '--force', '--version', '--help'].includes(long)) {
                 groups.general.push(opt);
             } else if (['--deployment-config', '--framework', '--model-format', '--model-name', '--model-server', '--base-image'].includes(long)) {
                 groups.model.push(opt);

package/config/parameter-schema.json

@@ -1,7 +1,7 @@
 {
     "schemaVersion": "1.0.0",
     "deploymentTargets": {
-        "managed-inference": {
+        "realtime-inference": {
             "endpoint": {
                 "initialInstanceCount": {
                     "type": "integer",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aws/ml-container-creator",
-  "version": "0.2.3",
+  "version": "0.2.5",
   "description": "Generator for SageMaker AI BYOC paradigm for predictive inference use-cases.",
   "type": "module",
   "main": "src/app.js",

package/src/app.js

@@ -156,6 +156,22 @@ export async function run(projectName, options) {
             console.log('   If your model package lacks an InferenceSpecification, use the S3 path');
             console.log('   directly instead: --model-name="s3://bucket/path/model.tar.gz"\n');
         }
+    } else if (configManager.isAutoPrompt()) {
+        // Auto-prompt mode: run the wizard with all resolved values pre-filled.
+        // The wizard skips prompts for values already in explicitConfig and
+        // uses phase-level gates to skip irrelevant sections entirely.
+        // This gives context-aware prompting (correct MCP queries, filtered choices)
+        // while only asking for what's truly missing.
+        console.log('\n🔄 Auto-prompt mode — prompting only for missing values with full context');
+
+        const promptRunner = new PromptRunner({
+            configManager,
+            options: kebabOptions,
+            registryConfigManager,
+            baseConfig
+        });
+        const promptAnswers = await promptRunner.run();
+        answers = configManager.getFinalConfiguration(promptAnswers);
     } else {
         const promptRunner = new PromptRunner({
             configManager,
@@ -482,7 +498,7 @@ async function _ensureTemplateVariables(answers, registryConfigManager = null) {
         testTypes: [],
         buildTimestamp: new Date().toISOString(),
         buildTarget: 'codebuild',
-        deploymentTarget: 'managed-inference',
+        deploymentTarget: 'realtime-inference',
         hyperPodCluster: null,
         hyperPodNamespace: 'default',
         hyperPodReplicas: 1,

package/src/lib/auto-prompt-builder.js

@@ -0,0 +1,172 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+/**
+ * Auto-Prompt Builder — generates targeted prompts for missing required parameters.
+ * 
+ * Used by --auto-prompt mode to ask only for values that cannot be inferred
+ * or defaulted from the provided CLI flags.
+ */
+
+/**
+ * Builds a minimal set of prompts for the given missing parameters.
+ * Each prompt is self-contained and doesn't depend on multi-phase wizard state.
+ *
+ * @param {string[]} missingParams - Parameter names that need values
+ * @param {object} currentConfig - Current configuration (with defaults filled)
+ * @returns {Array} Array of prompt objects compatible with runPrompts()
+ */
+export function buildAutoPrompts(missingParams, currentConfig) {
+    const prompts = [];
+
+    for (const param of missingParams) {
+        const builder = PROMPT_BUILDERS[param];
+        if (builder) {
+            const prompt = builder(currentConfig);
+            if (prompt) {
+                prompts.push(prompt);
+            }
+        } else {
+            // Fallback: generic text input for unknown parameters
+            prompts.push({
+                type: 'input',
+                name: param,
+                message: `Enter value for ${param}:`
+            });
+        }
+    }
+
+    return prompts;
+}
+
+/**
+ * Map of parameter names to prompt builder functions.
+ * Each builder receives the current config and returns a prompt object.
+ */
+const PROMPT_BUILDERS = {
+    deploymentConfig: (_config) => ({
+        type: 'list',
+        name: 'deploymentConfig',
+        message: 'Select deployment configuration:',
+        choices: [
+            { type: 'separator', separator: '── Large Language Models ──' },
+            { name: 'Transformers with vLLM', value: 'transformers-vllm' },
+            { name: 'Transformers with SGLang', value: 'transformers-sglang' },
+            { name: 'Transformers with TensorRT-LLM', value: 'transformers-tensorrt-llm' },
+            { name: 'Transformers with LMI', value: 'transformers-lmi' },
+            { name: 'Transformers with DJL', value: 'transformers-djl' },
+            { type: 'separator', separator: '── HTTP Serving ──' },
+            { name: 'HTTP with Flask', value: 'http-flask' },
+            { name: 'HTTP with FastAPI', value: 'http-fastapi' },
+            { type: 'separator', separator: '── NVIDIA Triton ──' },
+            { name: 'Triton FIL (XGBoost, LightGBM)', value: 'triton-fil' },
+            { name: 'Triton ONNX Runtime', value: 'triton-onnxruntime' },
+            { name: 'Triton TensorFlow', value: 'triton-tensorflow' },
+            { name: 'Triton PyTorch', value: 'triton-pytorch' },
+            { name: 'Triton vLLM', value: 'triton-vllm' },
+            { name: 'Triton TensorRT-LLM', value: 'triton-tensorrtllm' },
+            { name: 'Triton Python Backend', value: 'triton-python' },
+            { type: 'separator', separator: '── Diffusion Models ──' },
+            { name: 'Diffusors with vLLM Omni', value: 'diffusors-vllm-omni' }
+        ]
+    }),
+
+    instanceType: (config) => {
+        const architecture = config.architecture || 'http';
+        const isGpu = architecture === 'transformers' || architecture === 'triton' || architecture === 'diffusors';
+
+        const gpuChoices = [
+            { name: 'ml.g5.xlarge  (1× A10G 24GB — small LLMs)', value: 'ml.g5.xlarge' },
+            { name: 'ml.g5.2xlarge (1× A10G 24GB — medium LLMs)', value: 'ml.g5.2xlarge' },
+            { name: 'ml.g5.4xlarge (1× A10G 24GB — larger models)', value: 'ml.g5.4xlarge' },
+            { name: 'ml.g5.12xlarge (4× A10G 96GB — large LLMs)', value: 'ml.g5.12xlarge' },
+            { name: 'ml.g5.48xlarge (8× A10G 192GB — very large)', value: 'ml.g5.48xlarge' },
+            { name: 'ml.g6.xlarge  (1× L4 24GB)', value: 'ml.g6.xlarge' },
+            { name: 'ml.g6.2xlarge (1× L4 24GB)', value: 'ml.g6.2xlarge' },
+            { name: 'ml.p4d.24xlarge (8× A100 320GB)', value: 'ml.p4d.24xlarge' },
+            { name: 'ml.p5.48xlarge (8× H100 640GB)', value: 'ml.p5.48xlarge' },
+            { name: 'Custom (enter manually)', value: '_custom' }
+        ];
+
+        const cpuChoices = [
+            { name: 'ml.m5.large   (2 vCPU, 8GB — lightweight)', value: 'ml.m5.large' },
+            { name: 'ml.m5.xlarge  (4 vCPU, 16GB — small models)', value: 'ml.m5.xlarge' },
+            { name: 'ml.m5.2xlarge (8 vCPU, 32GB — medium models)', value: 'ml.m5.2xlarge' },
+            { name: 'ml.m5.4xlarge (16 vCPU, 64GB — large models)', value: 'ml.m5.4xlarge' },
+            { name: 'ml.c5.xlarge  (4 vCPU, 8GB — compute-heavy)', value: 'ml.c5.xlarge' },
+            { name: 'ml.c5.2xlarge (8 vCPU, 16GB — compute-heavy)', value: 'ml.c5.2xlarge' },
+            { name: 'Custom (enter manually)', value: '_custom' }
+        ];
+
+        return {
+            type: 'list',
+            name: 'instanceType',
+            message: `Select instance type${isGpu ? ' (GPU recommended for this architecture)' : ''}:`,
+            choices: isGpu ? gpuChoices : cpuChoices
+        };
+    },
+
+    deploymentTarget: (_config) => ({
+        type: 'list',
+        name: 'deploymentTarget',
+        message: 'Select deployment target:',
+        choices: [
+            { name: 'Real-Time Inference', value: 'realtime-inference' },
+            { name: 'Async Inference', value: 'async-inference' },
+            { name: 'Batch Transform', value: 'batch-transform' },
+            { name: 'HyperPod EKS', value: 'hyperpod-eks' }
+        ]
+    }),
+
+    modelFormat: (config) => {
+        const engine = config.engine || 'sklearn';
+        const formatMap = {
+            sklearn: [
+                { name: 'pkl (pickle)', value: 'pkl' },
+                { name: 'joblib', value: 'joblib' }
+            ],
+            xgboost: [
+                { name: 'json', value: 'json' },
+                { name: 'model (binary)', value: 'model' },
+                { name: 'ubj (universal binary JSON)', value: 'ubj' }
+            ],
+            tensorflow: [
+                { name: 'keras', value: 'keras' },
+                { name: 'h5', value: 'h5' },
+                { name: 'SavedModel', value: 'SavedModel' }
+            ]
+        };
+
+        const choices = formatMap[engine] || formatMap.sklearn;
+
+        return {
+            type: 'list',
+            name: 'modelFormat',
+            message: `Select model format for ${engine}:`,
+            choices
+        };
+    },
+
+    awsRegion: (_config) => ({
+        type: 'list',
+        name: 'awsRegion',
+        message: 'Select AWS region:',
+        choices: [
+            { name: 'us-east-1 (N. Virginia)', value: 'us-east-1' },
+            { name: 'us-west-2 (Oregon)', value: 'us-west-2' },
+            { name: 'eu-west-1 (Ireland)', value: 'eu-west-1' },
+            { name: 'ap-northeast-1 (Tokyo)', value: 'ap-northeast-1' },
+            { name: 'ap-southeast-1 (Singapore)', value: 'ap-southeast-1' },
+            { name: 'Custom (enter manually)', value: '_custom' }
+        ]
+    }),
+
+    buildTarget: (_config) => ({
+        type: 'list',
+        name: 'buildTarget',
+        message: 'Select build target:',
+        choices: [
+            { name: 'CodeBuild (recommended)', value: 'codebuild' }
+        ]
+    })
+};

package/src/lib/ci-register-helpers.js

@@ -25,7 +25,7 @@ import { createHash } from 'node:crypto';
  * @param {string} modelName - e.g. "meta-llama/Llama-2-7b-chat-hf", defaults to "none"
  * @param {string} instanceType - e.g. "ml.g5.xlarge"
  * @param {string} region - e.g. "us-east-1"
- * @param {string} deploymentTarget - e.g. "managed-inference"
+ * @param {string} deploymentTarget - e.g. "realtime-inference"
  * @returns {string} 16-character lowercase hex string
  */
 export function computeConfigId(deploymentConfig, modelName, instanceType, region, deploymentTarget) {

package/src/lib/cli-handler.js

@@ -190,7 +190,7 @@ CLI OPTIONS:
   --instance-type=<type>      SageMaker instance type (e.g., ml.m5.large, ml.g5.xlarge)
   --region=<region>           AWS region
   --role-arn=<arn>            AWS IAM role ARN for SageMaker execution
-  --deployment-target=<target> Deployment target (managed-inference|hyperpod-eks)
+  --deployment-target=<target> Deployment target (realtime-inference|async-inference|batch-transform|hyperpod-eks)
   --hyperpod-cluster=<name> HyperPod EKS cluster name
   --hyperpod-namespace=<ns> Kubernetes namespace for HyperPod (default: default)
   --hyperpod-replicas=<n>   Number of replicas for HyperPod (default: 1)

package/src/lib/config-manager.js

@@ -78,6 +78,7 @@ export default class ConfigManager {
         this.args = args || [];
         this.config = {};
         this.skipPrompts = false;
+        this.autoPrompt = false;
         this.deploymentConfigResolver = new DeploymentConfigResolver();
         this.parameterMatrix = this._getParameterMatrix();
         this.schemaValidator = new ParameterSchemaValidator();
@@ -106,6 +107,9 @@ export default class ConfigManager {
         await this._loadCliArguments();
         await this._loadCliOptions();
 
+        // Normalize deprecated values to canonical equivalents
+        this._normalizeDeprecatedValues();
+
         // Query configured MCP servers for unbounded parameter values
         await this._queryMcpServers();
 
@@ -113,6 +117,18 @@ export default class ConfigManager {
         this.skipPrompts = this.options['skip-prompts'] || 
                           this._hasCompleteConfiguration();
 
+        // Auto-prompt mode: fill defaults like skip-prompts, but prompt for truly missing values
+        this.autoPrompt = this.options['auto-prompt'] === true;
+        if (this.autoPrompt) {
+            // In auto-prompt mode, we don't skip prompts entirely — we'll selectively prompt
+            this.skipPrompts = false;
+
+            // Pre-fill defaults for required parameters that can be auto-generated.
+            // Promote these into explicitConfig so the wizard skips them.
+            // This means the wizard only prompts for values that are truly ambiguous.
+            this._fillAutoPromptDefaults();
+        }
+
         return this.config;
     }
 
@@ -182,8 +198,8 @@ export default class ConfigManager {
             }
         }
 
-        // When skipping prompts, provide reasonable defaults for missing required parameters
-        if (this.skipPrompts) {
+        // When skipping prompts or in auto-prompt mode, provide reasonable defaults for missing required parameters
+        if (this.skipPrompts || this.autoPrompt) {
             Object.entries(this.parameterMatrix).forEach(([param, config]) => {
                 if (config.required && 
                     (finalConfig[param] === null || finalConfig[param] === undefined)) {
@@ -635,7 +651,7 @@ export default class ConfigManager {
                 mcp: false,
                 promptable: true,
                 required: true,
-                default: 'managed-inference',
+                default: 'realtime-inference',
                 valueSpace: 'bounded'
             },
             hyperPodCluster: {
@@ -1373,6 +1389,35 @@ export default class ConfigManager {
         this._parseEnvVarOptions('server-env', 'serverEnvVars');
     }
 
+    /**
+     * Normalizes deprecated parameter values to their canonical equivalents.
+     * Prints a deprecation warning when a deprecated value is encountered.
+     * @private
+     */
+    _normalizeDeprecatedValues() {
+        const DEPRECATED_VALUES = {
+            deploymentTarget: {
+                'managed-inference': {
+                    canonical: 'realtime-inference',
+                    message: '--deployment-target=managed-inference is deprecated, use realtime-inference instead'
+                }
+            }
+        };
+
+        for (const [param, aliases] of Object.entries(DEPRECATED_VALUES)) {
+            const currentValue = this.config[param];
+            if (currentValue && aliases[currentValue]) {
+                const { canonical, message } = aliases[currentValue];
+                console.log(`\n⚠️  Deprecation: ${message}`);
+                this.config[param] = canonical;
+                // Also update explicit config if it was set there
+                if (this.explicitConfig && this.explicitConfig[param] === currentValue) {
+                    this.explicitConfig[param] = canonical;
+                }
+            }
+        }
+    }
+
     /**
      * Parse --model-env or --server-env CLI options into env var collections.
      * Supports both array (multiple flags) and single string values.
@@ -1828,6 +1873,135 @@ export default class ConfigManager {
         return autoGeneratable.includes(param);
     }
 
+    /**
+     * Fills auto-prompt defaults for parameters that have sensible defaults
+     * or can be inferred from the current config. Promotes these into
+     * explicitConfig so the wizard skips them.
+     * 
+     * Only fills parameters that:
+     * - Have a non-null default in the parameter matrix, OR
+     * - Can be auto-generated (instanceType, modelFormat, etc.)
+     * 
+     * Does NOT fill parameters that are truly ambiguous and need user input
+     * (e.g., deploymentConfig when not provided).
+     * @private
+     */
+    _fillAutoPromptDefaults() {
+        if (!this.explicitConfig) {
+            this.explicitConfig = {};
+        }
+
+        // Derive architecture from deploymentConfig if available
+        let architecture = this.config.architecture;
+        if (!architecture && this.config.deploymentConfig) {
+            try {
+                const parts = this.deploymentConfigResolver.decompose(this.config.deploymentConfig);
+                architecture = parts.architecture;
+                this.config.architecture = parts.architecture;
+                this.config.backend = parts.backend;
+                this.config.engine = parts.engine;
+            } catch {
+                // Invalid deploymentConfig — will be caught by validation
+            }
+        }
+
+        Object.entries(this.parameterMatrix).forEach(([param, config]) => {
+            // Skip if already explicitly set
+            if (this.explicitConfig[param] !== undefined && this.explicitConfig[param] !== null) {
+                return;
+            }
+
+            // For optional parameters: mark them as explicit (with null) so the wizard skips them.
+            // The downstream template logic handles defaults for optional params.
+            if (!config.required) {
+                // Don't override if there's already a value in config
+                if (this.config[param] !== undefined && this.config[param] !== null) {
+                    this.explicitConfig[param] = this.config[param];
+                } else if (config.default !== null && config.default !== undefined) {
+                    this.config[param] = config.default;
+                    this.explicitConfig[param] = config.default;
+                }
+                return;
+            }
+
+            // For required parameters: fill auto-generatable values
+            if (this.config[param] === undefined || this.config[param] === null) {
+                if (param === 'instanceType') {
+                    const arch = architecture || 'http';
+                    this.config[param] = arch === 'http' ? 'ml.m5.large' : 'ml.g5.xlarge';
+                } else if (param === 'modelFormat') {
+                    if (architecture === 'transformers' || architecture === 'triton' || architecture === 'diffusors') {
+                        return; // Not needed for these architectures
+                    }
+                    const engine = this.config.engine || 'sklearn';
+                    const formatMap = { sklearn: 'pkl', xgboost: 'json', tensorflow: 'keras' };
+                    this.config[param] = formatMap[engine] || 'pkl';
+                } else if (param === 'projectName') {
+                    this.config[param] = this._generateProjectName(architecture);
+                } else {
+                    return; // Can't fill — leave for prompting
+                }
+            }
+
+            // Promote non-null values to explicitConfig so the wizard skips them
+            if (this.config[param] !== undefined && this.config[param] !== null) {
+                if (config.default !== null || this._canAutoGenerate(param)) {
+                    this.explicitConfig[param] = this.config[param];
+                }
+            }
+        });
+    }
+
+    /**
+     * Returns whether auto-prompt mode is active
+     * @returns {boolean}
+     */
+    isAutoPrompt() {
+        return this.autoPrompt;
+    }
+
+    /**
+     * Gets the list of required parameters that are truly missing and cannot be
+     * auto-generated or defaulted. Used by auto-prompt mode to determine which
+     * specific prompts to show.
+     * 
+     * @returns {string[]} Array of parameter names that need prompting
+     */
+    getMissingRequiredParameters() {
+        const missing = [];
+
+        Object.entries(this.parameterMatrix).forEach(([param, config]) => {
+            if (!config.required || !config.promptable) return;
+
+            const value = this.config[param];
+            const hasValue = value !== undefined && value !== null;
+
+            if (hasValue) return;
+
+            // Special case: modelFormat is not required for transformers/triton/diffusors
+            if (param === 'modelFormat') {
+                const architecture = this.config.architecture;
+                if (architecture === 'transformers' || architecture === 'triton' || architecture === 'diffusors') {
+                    return;
+                }
+                // Can be inferred from engine
+                if (this.config.engine || this.config.deploymentConfig) {
+                    return;
+                }
+            }
+
+            // Skip params that can be auto-generated
+            if (this._canAutoGenerate(param)) return;
+
+            // Skip params that have a non-null default
+            if (config.default !== null && config.default !== undefined) return;
+
+            missing.push(param);
+        });
+
+        return missing;
+    }
+
     /**
      * Generates a project name based on framework
      * @param {string} framework - The ML framework

package/src/lib/parameter-schema-validator.js

@@ -27,15 +27,15 @@ const SUPPORTED_SCHEMA_VERSION = '1.0.0';
  * Format: 'deploymentTarget.category.schemaKey'
  */
 const PARAMETER_NAME_MAP = {
-    endpointInitialInstanceCount: 'managed-inference.endpoint.initialInstanceCount',
-    endpointDataCapturePercent: 'managed-inference.endpoint.dataCapturePercent',
-    endpointVariantName: 'managed-inference.endpoint.variantName',
-    endpointVolumeSize: 'managed-inference.endpoint.volumeSize',
-    icCpuCount: 'managed-inference.inferenceComponent.cpuCount',
-    icMemorySize: 'managed-inference.inferenceComponent.memorySize',
-    icGpuCount: 'managed-inference.inferenceComponent.gpuCount',
-    icCopyCount: 'managed-inference.inferenceComponent.copyCount',
-    icModelWeight: 'managed-inference.inferenceComponent.modelWeight'
+    endpointInitialInstanceCount: 'realtime-inference.endpoint.initialInstanceCount',
+    endpointDataCapturePercent: 'realtime-inference.endpoint.dataCapturePercent',
+    endpointVariantName: 'realtime-inference.endpoint.variantName',
+    endpointVolumeSize: 'realtime-inference.endpoint.volumeSize',
+    icCpuCount: 'realtime-inference.inferenceComponent.cpuCount',
+    icMemorySize: 'realtime-inference.inferenceComponent.memorySize',
+    icGpuCount: 'realtime-inference.inferenceComponent.gpuCount',
+    icCopyCount: 'realtime-inference.inferenceComponent.copyCount',
+    icModelWeight: 'realtime-inference.inferenceComponent.modelWeight'
 };
 
 export default class ParameterSchemaValidator {
@@ -96,7 +96,7 @@ export default class ParameterSchemaValidator {
     /**
      * Resolve a parameter name to its schema constraint object.
      * @param {string} parameterName - ConfigManager key (e.g., 'endpointVolumeSize')
-     * @param {string} [deploymentTarget] - Deployment target override (e.g., 'managed-inference')
+     * @param {string} [deploymentTarget] - Deployment target override (e.g., 'realtime-inference')
      * @returns {Object|null} Constraint object or null if not found
      */
     _resolveConstraint(parameterName, deploymentTarget) {

package/src/lib/prompt-runner.js

@@ -81,9 +81,9 @@ export default class PromptRunner {
         const regionPreviousAnswers = bootstrapRegion ? { _bootstrapRegion: bootstrapRegion } : {};
         const regionAndTargetAnswers = await this._runPhase(infraRegionAndTargetPrompts, regionPreviousAnswers, explicitConfig, existingConfig);
 
-        // 1b. Instance type — query MCP and prompt for managed-inference, async-inference, batch-transform, and hyperpod-eks
+        // 1b. Instance type — query MCP and prompt for realtime-inference, async-inference, batch-transform, and hyperpod-eks
         let instanceAnswers = {};
-        if (regionAndTargetAnswers.deploymentTarget === 'managed-inference' ||
+        if (regionAndTargetAnswers.deploymentTarget === 'realtime-inference' ||
             regionAndTargetAnswers.deploymentTarget === 'async-inference' ||
             regionAndTargetAnswers.deploymentTarget === 'batch-transform' ||
             regionAndTargetAnswers.deploymentTarget === 'hyperpod-eks') {
@@ -521,7 +521,18 @@ export default class PromptRunner {
         // First, add any existing config values to previousAnswers so they're available for defaults
         const allPreviousAnswers = { ...existingConfig, ...previousAnswers };
         
-        return await this._runPrompts(promptablePrompts.map(prompt => ({
+        // Collect explicit values for prompts that will be skipped.
+        // When a prompt is skipped because its value is in explicitConfig,
+        // the prompt library won't include it in the returned answers.
+        // Downstream code expects the value to be present, so we inject it.
+        const skippedValues = {};
+        for (const prompt of promptablePrompts) {
+            if (explicitConfig[prompt.name] !== undefined && explicitConfig[prompt.name] !== null) {
+                skippedValues[prompt.name] = explicitConfig[prompt.name];
+            }
+        }
+
+        const promptedAnswers = await this._runPrompts(promptablePrompts.map(prompt => ({
             ...prompt,
             // Wrap message to inject previousAnswers so prompts can access _mcpInstanceChoices etc.
             message: typeof prompt.message === 'function' ? (answers) => {
@@ -541,14 +552,34 @@ export default class PromptRunner {
             } : (existingConfig[prompt.name] !== undefined && existingConfig[prompt.name] !== null) ? 
                 existingConfig[prompt.name] : undefined,
             // Skip prompt ONLY if we have explicit config (not defaults)
+            // In auto-prompt mode, also skip optional prompts (not required in parameter matrix)
             when: prompt.when ? (answers) => {
                 // Skip if we have the value from explicit config (CLI, env vars, config files)
                 if (explicitConfig[prompt.name] !== undefined && explicitConfig[prompt.name] !== null) {
                     return false;
                 }
+                // In auto-prompt mode, skip optional/non-matrix parameters entirely
+                if (this.configManager?.isAutoPrompt()) {
+                    const paramConfig = this.configManager.parameterMatrix[prompt.name];
+                    // Skip if not in matrix (supplementary prompt) or if optional
+                    if (!paramConfig || !paramConfig.required) {
+                        return false;
+                    }
+                }
                 return prompt.when({...allPreviousAnswers, ...answers});
-            } : (explicitConfig[prompt.name] !== undefined && explicitConfig[prompt.name] !== null) ? 
-                () => false : undefined,
+            } : (_answers) => {
+                // No original when condition — skip if explicit or if auto-prompt + optional/non-matrix
+                if (explicitConfig[prompt.name] !== undefined && explicitConfig[prompt.name] !== null) {
+                    return false;
+                }
+                if (this.configManager?.isAutoPrompt()) {
+                    const paramConfig = this.configManager.parameterMatrix[prompt.name];
+                    if (!paramConfig || !paramConfig.required) {
+                        return false;
+                    }
+                }
+                return true;
+            },
             // Provide access to previous answers for conditional logic
             // For unbounded parameters, inject MCP-provided choices if available
             choices: prompt.choices ? (answers) => {
@@ -563,6 +594,9 @@ export default class PromptRunner {
                 return prompt.choices;
             } : undefined
         })));
+
+        // Merge skipped explicit values into the answers so downstream code sees them
+        return { ...skippedValues, ...promptedAnswers };
     }
 
     /**
@@ -638,7 +672,7 @@ export default class PromptRunner {
 
     /**
      * Query MCP instance-recommender server after deployment target is known.
-     * Only runs when deploymentTarget is managed-inference.
+     * Only runs when deploymentTarget is realtime-inference.
      * Populates configManager.mcpChoices so _runPhase injects them into list prompts.
      * @private
      */
@@ -1314,12 +1348,22 @@ export default class PromptRunner {
             return inferenceAmiVersion ? { cudaVersion, inferenceAmiVersion } : null;
         }
 
-        // Multiple options — let the user choose
+        // Multiple options — let the user choose (or auto-select in auto-prompt mode)
         const defaultVersion = frameworkAccel?.version
             && compatibleVersions.includes(frameworkAccel.version)
             ? frameworkAccel.version
             : instanceInfo.accelerator.default || compatibleVersions[compatibleVersions.length - 1];
 
+        // In auto-prompt mode, auto-select the default without prompting
+        if (this.configManager?.isAutoPrompt()) {
+            const inferenceAmiVersion = PromptRunner.CUDA_AMI_MAP[defaultVersion];
+            if (inferenceAmiVersion) {
+                console.log(`\n🔧 CUDA ${defaultVersion} auto-selected (auto-prompt mode)`);
+                console.log(`   AMI: ${inferenceAmiVersion}`);
+            }
+            return inferenceAmiVersion ? { cudaVersion: defaultVersion, inferenceAmiVersion } : null;
+        }
+
         const choices = compatibleVersions.map(v => {
             const ami = PromptRunner.CUDA_AMI_MAP[v] || 'unknown';
             const isDefault = v === defaultVersion ? ' (recommended)' : '';

package/src/lib/prompts.js

@@ -649,7 +649,7 @@ const modulePrompts = [
 /**
  * Infrastructure prompts split into sub-phases so the prompt runner can
  * interleave MCP queries between them (e.g. query instance-recommender
- * only after we know the deployment target is managed-inference).
+ * only after we know the deployment target is realtime-inference).
  *
  * Ordering: Region → Deployment Target → Instance/HyperPod → Build Target → Role
  */
@@ -683,21 +683,21 @@ const infraRegionAndTargetPrompts = [
         name: 'deploymentTarget',
         message: 'Deployment target?',
         choices: [
-            { name: 'SageMaker Managed Inference - Real Time', value: 'managed-inference' },
-            { name: 'SageMaker Managed Inference - Async', value: 'async-inference' },
-            { name: 'SageMaker Managed Inference - Batch', value: 'batch-transform' },
+            { name: 'SageMaker Real-Time Inference', value: 'realtime-inference' },
+            { name: 'SageMaker Async Inference', value: 'async-inference' },
+            { name: 'SageMaker Batch Transform', value: 'batch-transform' },
             { name: 'SageMaker HyperPod - EKS', value: 'hyperpod-eks' }
         ],
-        default: 'managed-inference'
+        default: 'realtime-inference'
     }
 ];
 
-// Sub-phase B: Instance type (only when deploymentTarget === 'managed-inference')
+// Sub-phase B: Instance type (only when deploymentTarget === 'realtime-inference')
 const infraInstancePrompts = [
     {
         type: 'list',
         name: 'instanceType',
-        when: answers => answers.deploymentTarget === 'managed-inference' || answers.deploymentTarget === 'async-inference' || answers.deploymentTarget === 'batch-transform' || answers.deploymentTarget === 'hyperpod-eks',
+        when: answers => answers.deploymentTarget === 'realtime-inference' || answers.deploymentTarget === 'async-inference' || answers.deploymentTarget === 'batch-transform' || answers.deploymentTarget === 'hyperpod-eks',
         message: (answers) => {
             const framework = answers.framework || answers.deploymentConfig?.split('-')[0];
             

package/src/lib/template-manager.js

@@ -64,7 +64,7 @@ export default class TemplateManager {
                 'diffusors-vllm-omni'
             ],
             buildTargets: ['codebuild'],
-            deploymentTargets: ['managed-inference', 'async-inference', 'batch-transform', 'hyperpod-eks'],
+            deploymentTargets: ['realtime-inference', 'async-inference', 'batch-transform', 'hyperpod-eks'],
             testTypes: ['local-model-cli', 'local-model-server', 'hosted-model-endpoint'],
             awsRegions: [
                 'us-east-1', 'us-east-2', 'us-west-1', 'us-west-2',
@@ -135,7 +135,7 @@ export default class TemplateManager {
         // Validate batch transform specific fields
         this._validateBatchTransformConfig();
         
-        // Validate instance type format (ml.*.*) - only for managed-inference
+        // Validate instance type format (ml.*.*) - only for realtime-inference
         if (this.answers.instanceType && this.answers.instanceType !== 'custom') {
             const instancePattern = /^ml\.[a-z0-9]+\.(nano|micro|small|medium|large|xlarge|[0-9]+xlarge)$/;
             if (!instancePattern.test(this.answers.instanceType)) {

package/templates/do/clean

@@ -24,7 +24,7 @@ done
 
 # Function to display usage
 show_usage() {
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
     echo "Usage: ./do/clean [local|ecr|endpoint|codebuild|all]"
 <% } else if (deploymentTarget === 'async-inference') { %>
     echo "Usage: ./do/clean [local|ecr|endpoint|codebuild|all]"
@@ -37,7 +37,7 @@ show_usage() {
     echo "Cleanup targets:"
     echo "  local     - Remove local Docker images"
     echo "  ecr       - Remove images from Amazon ECR"
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
     echo "  endpoint  - Delete SageMaker endpoint, configuration, and model"
 <% } else if (deploymentTarget === 'async-inference') { %>
     echo "  endpoint  - Delete SageMaker async endpoint, configuration, and inference component"
@@ -51,7 +51,7 @@ show_usage() {
     echo ""
     echo "Examples:"
     echo "  ./do/clean local      # Remove local Docker images only"
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
     echo "  ./do/clean endpoint   # Delete SageMaker resources only"
 <% } else if (deploymentTarget === 'async-inference') { %>
     echo "  ./do/clean endpoint   # Delete SageMaker async resources only"
@@ -198,7 +198,7 @@ clean_ecr() {
     fi
 }
 
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 # Function to clean SageMaker endpoint and inference components
 clean_endpoint() {
     echo "🧹 Cleaning SageMaker resources"
@@ -720,7 +720,7 @@ case "${CLEANUP_TARGET}" in
     ecr)
         clean_ecr
         ;;
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
     endpoint)
         clean_endpoint
         ;;
@@ -761,7 +761,7 @@ case "${CLEANUP_TARGET}" in
         
         echo ""
         
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
         # Clean SageMaker resources
         if clean_endpoint; then
             CLEANED_ITEMS+=("SageMaker resources")

package/templates/do/config

@@ -24,8 +24,8 @@ export CODEBUILD_PROJECT_NAME="${PROJECT_NAME}-build-$(date +%Y%m%d)"
 # Deployment configuration — WHERE the model runs
 export DEPLOYMENT_TARGET="<%= deploymentTarget %>"
 
-<% if (deploymentTarget === 'managed-inference') { %>
-# SageMaker Managed Inference configuration
+<% if (deploymentTarget === 'realtime-inference') { %>
+# SageMaker Real-Time Inference configuration
 export INSTANCE_TYPE="<%= instanceType %>"
 <% if (inferenceAmiVersion) { %>
 export INFERENCE_AMI_VERSION="<%= inferenceAmiVersion %>"
@@ -33,7 +33,7 @@ export INFERENCE_AMI_VERSION="<%= inferenceAmiVersion %>"
 <% } %>
 
 <% if (deploymentTarget === 'async-inference') { %>
-# SageMaker Managed Inference - Async configuration
+# SageMaker Async Inference configuration
 export INSTANCE_TYPE="<%= instanceType %>"
 <% if (inferenceAmiVersion) { %>
 export INFERENCE_AMI_VERSION="<%= inferenceAmiVersion %>"
@@ -77,7 +77,7 @@ export FSX_VOLUME_HANDLE="<%= fsxVolumeHandle %>"
 <% } %>
 
 <% if (deploymentTarget === 'batch-transform') { %>
-# SageMaker Managed Inference - Batch configuration
+# SageMaker Batch Transform configuration
 export INSTANCE_TYPE="<%= instanceType %>"
 
 # Resolve AWS account ID at runtime for default resource names
@@ -187,7 +187,7 @@ export BASE_IMAGE=${BASE_IMAGE:-<%= baseImage %>}
 
 # Allow environment variable overrides
 export AWS_REGION=${AWS_REGION:-<%= awsRegion %>}
-<% if (deploymentTarget === 'managed-inference' || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
+<% if (deploymentTarget === 'realtime-inference' || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
 export INSTANCE_TYPE=${INSTANCE_TYPE:-<%= instanceType %>}
 <% } %>
 export ECR_REPOSITORY_NAME=${ECR_REPOSITORY_NAME:-ml-container-creator}
@@ -241,7 +241,7 @@ echo "   Model env vars: <%= Object.keys(modelEnvVars).length %>"
 <% if (typeof serverEnvVars !== 'undefined' && serverEnvVars && Object.keys(serverEnvVars).length > 0) { %>
 echo "   Server env vars: <%= Object.keys(serverEnvVars).length %>"
 <% } %>
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 echo "   Instance: ${INSTANCE_TYPE}"
 <% } else if (deploymentTarget === 'async-inference') { %>
 echo "   Instance: ${INSTANCE_TYPE}"

package/templates/do/deploy

@@ -36,7 +36,7 @@ echo "   Deployment config: ${DEPLOYMENT_CONFIG}"
 echo "   Region: ${AWS_REGION}"
 echo "   Build target: ${BUILD_TARGET}"
 echo "   Deployment target: ${DEPLOYMENT_TARGET}"
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 echo "   Instance type: ${INSTANCE_TYPE}"
 <% } else if (deploymentTarget === 'async-inference') { %>
 echo "   Instance type: ${INSTANCE_TYPE}"
@@ -95,9 +95,9 @@ fi
 echo "✅ ECR image found: ${ECR_REPOSITORY}:${PROJECT_NAME}-latest"
 IMAGE_TAG="${PROJECT_NAME}-latest"
 
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 # ============================================================
-# SageMaker Managed Inference Deployment (Inference Components)
+# SageMaker Real-Time Inference Deployment (Inference Components)
 # ============================================================
 
 # Validate execution role ARN
@@ -520,7 +520,7 @@ echo "   ./do/clean endpoint"
 
 <% } else if (deploymentTarget === 'async-inference') { %>
 # ============================================================
-# SageMaker Managed Inference - Async Deployment (Model-Based)
+# SageMaker Async Inference Deployment (Model-Based)
 # SageMaker async inference does NOT support Inference Components.
 # Flow: create-model → create-endpoint-config (with AsyncInferenceConfig) → create-endpoint
 # ============================================================
@@ -1151,7 +1151,7 @@ _update_config_var "KUBECONFIG" "${KUBECONFIG_PATH}"
 
 <% } else if (deploymentTarget === 'batch-transform') { %>
 # ============================================================
-# SageMaker Managed Inference - Batch Deployment
+# SageMaker Batch Transform Deployment
 # Flow: create-model → create-transform-job → poll until completion
 # ============================================================
 

package/templates/do/export

@@ -42,8 +42,8 @@ if [ "${1:-}" = "--json" ]; then
     # Deployment target
     JSON="${JSON},\"deploymentTarget\":\"${DEPLOYMENT_TARGET}\""
 
-<% if (deploymentTarget === 'managed-inference') { %>
-    # SageMaker Managed Inference
+<% if (deploymentTarget === 'realtime-inference') { %>
+    # SageMaker Real-Time Inference
     JSON="${JSON},\"instanceType\":\"${INSTANCE_TYPE}\""
 <% } else if (deploymentTarget === 'async-inference') { %>
     # SageMaker Async Inference
@@ -206,11 +206,11 @@ fi
 # Deployment target
 CMD="${CMD} --deployment-target=${DEPLOYMENT_TARGET}"
 
-<% if (deploymentTarget === 'managed-inference') { %>
-# SageMaker Managed Inference
+<% if (deploymentTarget === 'realtime-inference') { %>
+# SageMaker Real-Time Inference
 CMD="${CMD} --instance-type=${INSTANCE_TYPE}"
 <% } else if (deploymentTarget === 'batch-transform') { %>
-# SageMaker Managed Inference - Batch
+# SageMaker Batch Transform
 CMD="${CMD} --instance-type=${INSTANCE_TYPE}"
 CMD="${CMD} --batch-input-path=${BATCH_INPUT_PATH}"
 CMD="${CMD} --batch-output-path=${BATCH_OUTPUT_PATH}"

package/templates/do/logs

@@ -10,9 +10,9 @@ set -o pipefail
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 source "${SCRIPT_DIR}/config"
 
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 # ============================================================
-# SageMaker Managed Inference Logs (CloudWatch)
+# SageMaker Real-Time Inference Logs (CloudWatch)
 # ============================================================
 
 # Allow inference component name as argument or from config
@@ -95,7 +95,7 @@ aws logs tail "${LOG_GROUP}" \
 
 <% } else if (deploymentTarget === 'async-inference') { %>
 # ============================================================
-# SageMaker Managed Inference - Async Logs (CloudWatch)
+# SageMaker Async Inference Logs (CloudWatch)
 # ============================================================
 
 ENDPOINT="${1:-${ENDPOINT_NAME:-}}"
@@ -166,7 +166,7 @@ aws logs tail "${LOG_GROUP}" \
 
 <% } else if (deploymentTarget === 'batch-transform') { %>
 # ============================================================
-# SageMaker Managed Inference - Batch Logs (CloudWatch)
+# SageMaker Batch Transform Logs (CloudWatch)
 # ============================================================
 
 # Allow transform job name as argument or from config

package/templates/do/register

@@ -266,7 +266,7 @@ echo "   Backend:           ${BACKEND}"
 <% if (framework === 'transformers') { %>
 echo "   Model name:        ${MODEL_NAME:-N/A}"
 <% } %>
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 echo "   Instance type:     ${INSTANCE_TYPE}"
 <% } else if (deploymentTarget === 'batch-transform') { %>
 echo "   Instance:          ${INSTANCE_TYPE} x ${BATCH_INSTANCE_COUNT}"
@@ -498,7 +498,7 @@ DJEOF
     if [ "${CI_MODE}" = true ]; then
         echo ""
         echo "⚠️  CI Integration is experimental and currently only tested for"
-        echo "   SageMaker Managed Inference — Real Time endpoints."
+        echo "   SageMaker Real-Time Inference endpoints."
         echo ""
 
         # Compute configId
@@ -542,7 +542,7 @@ if [ -n "${MODEL_NAME:-}" ]; then
 fi
 <% } %>
 
-<% if (deploymentTarget === 'managed-inference' || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
+<% if (deploymentTarget === 'realtime-inference' || deploymentTarget === 'async-inference' || deploymentTarget === 'batch-transform') { %>
 CMD_ARGS+=("--instance-type" "${INSTANCE_TYPE}")
 <% } %>
 

package/templates/do/test

@@ -10,9 +10,9 @@ set -o pipefail
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 source "${SCRIPT_DIR}/config"
 
-<% if (deploymentTarget === 'managed-inference') { %>
+<% if (deploymentTarget === 'realtime-inference') { %>
 # ============================================================
-# SageMaker Managed Inference Testing
+# SageMaker Real-Time Inference Testing
 # ============================================================
 
 # Parse arguments
@@ -295,7 +295,7 @@ fi
 
 <% } else if (deploymentTarget === 'async-inference') { %>
 # ============================================================
-# SageMaker Managed Inference - Async Testing
+# SageMaker Async Inference Testing
 # ============================================================
 
 # Parse arguments
@@ -875,7 +875,7 @@ fi
 
 <% } else if (deploymentTarget === 'batch-transform') { %>
 # ============================================================
-# SageMaker Managed Inference - Batch Testing
+# SageMaker Batch Transform Testing
 # ============================================================
 
 # Parse arguments: local or batch test mode