@aws/ml-container-creator 0.2.0

package/servers/instance-recommender/index.js

@@ -0,0 +1,284 @@
+#!/usr/bin/env node
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+/**
+ * Instance Recommender MCP Server
+ *
+ * A bundled MCP server that recommends SageMaker instance types and
+ * IAM role ARNs based on the current ML framework and model configuration.
+ *
+ * Supports two modes:
+ *   - Static (default): Returns hardcoded instance lists by framework category
+ *   - Smart (--smart flag or BEDROCK_SMART=true): Queries Amazon Bedrock for
+ *     context-aware recommendations, falling back to static on failure
+ *
+ * Tool: get_instance_types
+ *   Accepts: { parameters: string[], limit: number, context: object }
+ *   Returns: { values: Record<string, string>, choices: Record<string, string[]> }
+ */
+
+import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
+import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'
+import { z } from 'zod'
+import { readFileSync } from 'node:fs'
+import { fileURLToPath } from 'node:url'
+import { resolve, dirname } from 'node:path'
+import { queryBedrock } from '../lib/bedrock-client.js'
+
+// ── Catalog loader ───────────────────────────────────────────────────────────
+
+const __filename = fileURLToPath(import.meta.url)
+const __dirname = dirname(__filename)
+
+/**
+ * Load and parse a JSON catalog file relative to the server directory.
+ * Throws on missing file or invalid JSON with the file path in the message.
+ *
+ * @param {string} relativePath - Path relative to server dir (e.g. './catalogs/instances.json')
+ * @returns {any} Parsed JSON content
+ */
+function loadCatalog(relativePath) {
+    const fullPath = resolve(__dirname, relativePath)
+    let raw
+    try {
+        raw = readFileSync(fullPath, 'utf8')
+    } catch (err) {
+        throw new Error(`Catalog file not found: ${fullPath}`)
+    }
+    try {
+        return JSON.parse(raw)
+    } catch (err) {
+        throw new Error(`Failed to parse catalog ${fullPath}: ${err.message}`)
+    }
+}
+
+// ── Load catalogs from JSON files ─────────────────────────────────────────────
+
+let INSTANCE_CATALOG
+let INSTANCE_RECOMMENDATIONS
+
+try {
+    const data = loadCatalog('./catalogs/instances.json')
+    INSTANCE_CATALOG = data.catalog
+    INSTANCE_RECOMMENDATIONS = data.recommendations
+} catch (err) {
+    process.stderr.write(`[instance-recommender] Fatal: ${err.message}\n`)
+    process.exit(1)
+}
+
+const GPU_FRAMEWORKS = new Set(['transformers'])
+
+// Bedrock configuration
+const SMART_MODE = process.env.BEDROCK_SMART === 'true'
+const BEDROCK_MODEL = process.env.BEDROCK_MODEL || 'global.anthropic.claude-sonnet-4-20250514-v1:0'
+const BEDROCK_REGION = process.env.BEDROCK_REGION || process.env.AWS_REGION || 'us-east-1'
+
+/**
+ * Per-server configuration passed to the shared Bedrock client.
+ */
+const SERVER_CONFIG = {
+    serverName: 'instance-recommender',
+    systemPromptTemplate: `You are an AWS SageMaker instance type advisor. Given the following ML deployment context, recommend the best SageMaker instance types.
+
+Current configuration: {context}
+Requested parameters: {parameters}
+Maximum recommendations per parameter: {limit}
+
+Respond with ONLY a JSON object in this exact format, no other text:
+{
+  "values": {
+    "instanceType": "the single best instance type as a string",
+    "awsRoleArn": "a recommended role ARN pattern if applicable"
+  }
+}
+
+Rules:
+- Only include parameters that were requested
+- For instanceType: recommend real SageMaker instance types (ml.* prefix) appropriate for the framework and model
+- For awsRoleArn: skip this field, do not recommend ARNs
+- The first value in any list should be your top recommendation
+- Consider GPU vs CPU needs based on the framework
+- Consider model size and memory requirements if model info is available
+- Return valid JSON only`,
+    temperature: 0.3,
+    maxTokens: 1024,
+    modelId: BEDROCK_MODEL,
+    region: BEDROCK_REGION
+}
+
+/**
+ * Determine which instance list to use based on framework context and optional search term.
+ * When instanceSearch is provided, filters the catalog by keyword matching.
+ */
+function getStaticInstances(context) {
+    const framework = context?.framework
+    const search = context?.instanceSearch
+
+    // Start with framework-based category filter
+    const isGpu = framework && GPU_FRAMEWORKS.has(framework)
+    if (!search) {
+        // No search term — return the legacy category-based list
+        return isGpu ? INSTANCE_RECOMMENDATIONS.gpu : INSTANCE_RECOMMENDATIONS.cpu
+    }
+
+    // Search mode: use the full catalog
+    let candidates = Object.entries(INSTANCE_CATALOG)
+
+    // Tokenize search into lowercase keywords
+    const tokens = search.toLowerCase().split(/[\s,\-_]+/).filter(Boolean)
+
+    // Detect compound terms before tokenization
+    const rawLower = search.toLowerCase()
+    const wantsMultiGpu = rawLower.includes('multi gpu') || rawLower.includes('multi-gpu') || rawLower.includes('multigpu')
+
+    // Detect CUDA version requests: "cuda 12", "cuda 11.8", "cuda-12.1"
+    const cudaMatch = rawLower.match(/cuda[\s\-_]*(\d+(?:\.\d+)?)/)
+    const wantsCudaVersion = cudaMatch ? cudaMatch[1] : null
+
+    // Score each instance by how many tokens match its tags, accelerator, or instance name
+    const scored = candidates.map(([name, meta]) => {
+        let score = 0
+        const cudaStr = meta.cudaVersions ? meta.cudaVersions.join(' ') : ''
+        const haystack = [...meta.tags, meta.accelerator.toLowerCase(), name, meta.category, cudaStr].join(' ')
+
+        // Compound term: multi-gpu — only match instances with >1 GPU
+        if (wantsMultiGpu) {
+            if (meta.gpus > 1) {
+                score += 5
+            } else {
+                return { name, meta, score: 0 }
+            }
+        }
+
+        // Compound term: cuda version — only match instances supporting that version
+        if (wantsCudaVersion) {
+            if (!meta.cudaVersions) return { name, meta, score: 0 }
+            const hasExact = meta.cudaVersions.includes(wantsCudaVersion)
+            const hasMajor = meta.cudaVersions.some(v => v.startsWith(wantsCudaVersion))
+            if (hasExact) {
+                score += 4
+            } else if (hasMajor) {
+                score += 3
+            } else {
+                return { name, meta, score: 0 }
+            }
+        }
+
+        for (const token of tokens) {
+            // Skip tokens already handled by compound term detection
+            if (wantsMultiGpu && (token === 'multi' || token === 'gpu')) continue
+            if (wantsCudaVersion && (token === 'cuda' || token === wantsCudaVersion)) continue
+
+            if (haystack.includes(token)) score += 1
+            if (meta.gpus > 1 && (token === 'parallel')) score += 2
+            if (token === 'gpu' && meta.gpus > 0) score += 1
+            if (token === 'cpu' && meta.gpus === 0) score += 1
+            if (token === 'cheap' || token === 'budget' || token === 'cost') {
+                if (meta.tags.includes('budget') || meta.tags.includes('cost-effective')) score += 1
+            }
+            if (token === 'memory' || token === 'high-memory') {
+                if (meta.memGb >= 32) score += 1
+            }
+            if (token === 'large' && meta.vcpus >= 16) score += 1
+            // Match specific CUDA versions (e.g. "11.8", "12.1")
+            if (meta.cudaVersions && meta.cudaVersions.includes(token)) score += 2
+        }
+        return { name, meta, score }
+    })
+
+    // Keep only instances with a positive score, sorted descending
+    const matched = scored.filter(s => s.score > 0).sort((a, b) => b.score - a.score)
+
+    if (matched.length === 0) {
+        // No matches — fall back to legacy category-based list
+        return isGpu ? INSTANCE_RECOMMENDATIONS.gpu : INSTANCE_RECOMMENDATIONS.cpu
+    }
+
+    return matched.map(s => s.name)
+}
+
+/**
+ * Log to stderr so it doesn't interfere with MCP stdio protocol on stdout.
+ */
+function log(message) {
+    process.stderr.write(`[instance-recommender] ${message}\n`)
+}
+
+// Create MCP server
+const server = new McpServer({
+    name: 'instance-recommender',
+    version: '1.0.0'
+})
+
+// Register the get_instance_types tool
+server.tool(
+    'get_instance_types',
+    'Returns recommended SageMaker instance types and configuration values for ML Container Creator',
+    {
+        parameters: z.array(z.string()).describe('List of parameter names to provide values for'),
+        limit: z.number().int().positive().default(10).describe('Maximum number of choices per parameter'),
+        context: z.record(z.string(), z.any()).optional().describe('Current configuration context (framework, modelServer, etc.)')
+    },
+    async ({ parameters, limit, context }) => {
+        const values = {}
+        const choices = {}
+        let usedSmart = false
+
+        // Smart mode: try Bedrock first
+        if (SMART_MODE && parameters.includes('instanceType')) {
+            log('[smart] Smart mode enabled, querying Amazon Bedrock...')
+            const bedrockResult = await queryBedrock(SERVER_CONFIG, parameters, limit, context || {})
+
+            if (bedrockResult?.values?.instanceType) {
+                values.instanceType = bedrockResult.values.instanceType
+                // Use the Bedrock recommendation as the top choice, pad with static list
+                const staticInstances = getStaticInstances(context || {})
+                const bedrockValue = bedrockResult.values.instanceType
+                const combined = [bedrockValue, ...staticInstances.filter(i => i !== bedrockValue)]
+                choices.instanceType = combined.slice(0, limit)
+                usedSmart = true
+                log(`[smart] Using Bedrock recommendation: ${bedrockValue}`)
+            } else {
+                log('[smart] Bedrock did not return usable results, falling back to static recommendations')
+            }
+        }
+
+        // Static fallback (or non-smart mode)
+        if (!usedSmart) {
+            for (const param of parameters) {
+                if (param === 'instanceType') {
+                    const instances = getStaticInstances(context || {})
+                    const limited = instances.slice(0, limit)
+                    values.instanceType = limited[0]
+                    choices.instanceType = limited
+                }
+                // awsRoleArn is left to the user — no default recommendations
+            }
+        }
+
+        return {
+            content: [{
+                type: 'text',
+                text: JSON.stringify({ values, choices })
+            }]
+        }
+    }
+)
+
+// Export for standalone testing
+export { loadCatalog, getStaticInstances, INSTANCE_CATALOG, INSTANCE_RECOMMENDATIONS, GPU_FRAMEWORKS }
+
+// Guard MCP transport — only connect when run as main module
+const isMain = process.argv[1] && resolve(process.argv[1]) === __filename
+
+if (isMain) {
+    if (SMART_MODE) {
+        log(`Smart mode enabled (model: ${BEDROCK_MODEL}, region: ${BEDROCK_REGION})`)
+    } else {
+        log('Static mode (set BEDROCK_SMART=true to enable Bedrock-powered recommendations)')
+    }
+
+    const transport = new StdioServerTransport()
+    await server.connect(transport)
+}

package/servers/instance-recommender/manifest.json

@@ -0,0 +1,16 @@
+{
+    "name": "@amzn/ml-container-creator-instance-recommender",
+    "version": "1.0.0",
+    "description": "MCP server that recommends SageMaker instance types for ML Container Creator.",
+    "modes": {
+        "static": true,
+        "smart": true,
+        "discover": false
+    },
+    "catalogs": {
+        "instances": "./catalogs/instances.json"
+    },
+    "tool": {
+        "name": "get_instance_types"
+    }
+}

package/servers/instance-recommender/package.json

@@ -0,0 +1,15 @@
+{
+  "name": "@amzn/ml-container-creator-instance-recommender",
+  "private": true,
+  "version": "1.0.0",
+  "description": "MCP server that recommends SageMaker instance types and IAM role ARNs based on ML framework and model configuration. Supports Bedrock-powered smart recommendations.",
+  "type": "module",
+  "main": "index.js",
+  "license": "Apache-2.0",
+  "scripts": {
+    "test": "node test.js"
+  },
+  "dependencies": {
+    "@modelcontextprotocol/sdk": "^1.0.0"
+  }
+}

package/servers/lib/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

package/servers/lib/bedrock-client.js

@@ -0,0 +1,160 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+/**
+ * Shared Bedrock Client
+ *
+ * Reusable module that encapsulates Amazon Bedrock LLM invocation,
+ * JSON extraction, and fail-fast error handling for bundled MCP servers.
+ *
+ * Each server passes its own SERVER_CONFIG to queryBedrock() so that
+ * system prompts, hyperparameters, and model selection are per-server.
+ */
+
+/**
+ * Extract a JSON object from text that may be raw JSON or wrapped
+ * in a markdown-fenced code block (```json ... ```).
+ *
+ * @param {string} text - Raw LLM response text
+ * @returns {object|null} Parsed JSON object, or null if extraction fails
+ */
+export function extractJson(text) {
+    if (!text || typeof text !== 'string') return null
+
+    // Try markdown-fenced code block first (```json ... ``` or ``` ... ```)
+    const fencedMatch = text.match(/```(?:json)?\s*\n?([\s\S]*?)```/)
+    if (fencedMatch) {
+        try {
+            return JSON.parse(fencedMatch[1].trim())
+        } catch {
+            // Fall through to raw extraction
+        }
+    }
+
+    // Try extracting raw JSON object
+    const jsonMatch = text.match(/\{[\s\S]*\}/)
+    if (jsonMatch) {
+        try {
+            return JSON.parse(jsonMatch[0])
+        } catch {
+            return null
+        }
+    }
+
+    return null
+}
+
+/**
+ * Query Amazon Bedrock for context-aware recommendations.
+ * Returns parsed JSON object on success, null on any failure.
+ *
+ * @param {object} serverConfig - Per-server configuration
+ * @param {string} serverConfig.systemPromptTemplate - Prompt template with {context}, {parameters}, {limit} placeholders
+ * @param {number} serverConfig.temperature - LLM temperature (0-1)
+ * @param {number} serverConfig.maxTokens - Max response tokens
+ * @param {string} serverConfig.modelId - Bedrock model ID
+ * @param {string} serverConfig.region - AWS region for Bedrock API
+ * @param {string} serverConfig.serverName - Server name for log prefixes
+ * @param {string[]} parameters - Requested parameter names
+ * @param {number} limit - Max choices per parameter
+ * @param {object} context - Current configuration context
+ * @returns {Promise<{values: object} | null>}
+ */
+export async function queryBedrock(serverConfig, parameters, limit, context) {
+    const prefix = `[${serverConfig.serverName}]`
+
+    // Dynamic import with 1s timeout
+    let BedrockRuntimeClient, InvokeModelCommand
+    try {
+        const mod = await Promise.race([
+            import('@aws-sdk/client-bedrock-runtime'),
+            new Promise((_, reject) =>
+                setTimeout(() => reject(new Error('Import timed out')), 1000)
+            )
+        ])
+        BedrockRuntimeClient = mod.BedrockRuntimeClient
+        InvokeModelCommand = mod.InvokeModelCommand
+    } catch {
+        log(prefix, 'Failed to load @aws-sdk/client-bedrock-runtime. Run "npm install" in the servers/lib/ directory')
+        return null
+    }
+
+    const client = new BedrockRuntimeClient({
+        region: serverConfig.region,
+        requestHandler: {
+            requestTimeout: 10000
+        }
+    })
+
+    // Build prompt from template
+    const contextStr = context && Object.keys(context).length > 0
+        ? JSON.stringify(context)
+        : 'No specific configuration context provided.'
+
+    const prompt = serverConfig.systemPromptTemplate
+        .replace('{context}', contextStr)
+        .replace('{parameters}', parameters.join(', '))
+        .replace('{limit}', String(limit))
+
+    try {
+        log(prefix, `Querying Bedrock model ${serverConfig.modelId} in ${serverConfig.region}...`)
+
+        const body = JSON.stringify({
+            anthropic_version: 'bedrock-2023-05-31',
+            max_tokens: serverConfig.maxTokens,
+            temperature: serverConfig.temperature,
+            messages: [{
+                role: 'user',
+                content: prompt
+            }]
+        })
+
+        const command = new InvokeModelCommand({
+            modelId: serverConfig.modelId,
+            contentType: 'application/json',
+            accept: 'application/json',
+            body
+        })
+
+        const response = await client.send(command)
+        const responseBody = JSON.parse(new TextDecoder().decode(response.body))
+
+        const text = responseBody.content?.[0]?.text
+        if (!text) {
+            log(prefix, 'Bedrock response contained no text content')
+            return null
+        }
+
+        const parsed = extractJson(text)
+        if (!parsed) {
+            log(prefix, 'Could not extract JSON from Bedrock response')
+            return null
+        }
+
+        if (!parsed.values || typeof parsed.values !== 'object') {
+            log(prefix, 'Bedrock response missing "values" object')
+            return null
+        }
+
+        log(prefix, `Bedrock returned recommendations: ${JSON.stringify(parsed.values)}`)
+        return parsed
+    } catch (err) {
+        if (err.name === 'AccessDeniedException') {
+            log(prefix, `Access denied. Ensure bedrock:InvokeModel permission for arn:aws:bedrock:${serverConfig.region}:*:inference-profile/${serverConfig.modelId}`)
+        } else if (err.name === 'ResourceNotFoundException') {
+            log(prefix, `Model "${serverConfig.modelId}" not found. Set BEDROCK_MODEL env var. Example: BEDROCK_MODEL=global.anthropic.claude-sonnet-4-20250514-v1:0`)
+        } else if (err.name === 'ThrottlingException') {
+            log(prefix, 'Bedrock rate limit hit. Falling back to static recommendations')
+        } else {
+            log(prefix, `Bedrock query failed: ${err.name}: ${err.message}`)
+        }
+        return null
+    }
+}
+
+/**
+ * Log to stderr so it doesn't interfere with MCP stdio protocol on stdout.
+ */
+function log(prefix, message) {
+    process.stderr.write(`${prefix} ${message}\n`)
+}