converse-mcp-server 2.22.3 → 2.22.5

package/docs/ALTERNATIVE_PROVIDERS.md

@@ -1,450 +1,450 @@
-# Alternative Cloud Provider Integrations for Converse MCP Server
-
-This document outlines alternative API providers for each AI provider currently supported in the Converse MCP Server, along with implementation requirements for adding support.
-
-## Table of Contents
-- [Anthropic](#anthropic)
-  - [AWS Bedrock](#aws-bedrock)
-  - [Google Vertex AI](#google-vertex-ai)
-- [OpenAI](#openai)
-  - [Azure OpenAI Service](#azure-openai-service)
-- [Google](#google)
-  - [Vertex AI Platform](#vertex-ai-platform)
-- [DeepSeek](#deepseek)
-  - [AWS Bedrock](#aws-bedrock-1)
-  - [Azure AI Foundry](#azure-ai-foundry)
-- [Mistral AI](#mistral-ai)
-  - [AWS Bedrock](#aws-bedrock-2)
-  - [Azure AI](#azure-ai)
-  - [Google Vertex AI](#google-vertex-ai-1)
-- [XAI (Grok)](#xai-grok)
-  - [Azure AI Foundry](#azure-ai-foundry-1)
-- [Implementation Strategy](#implementation-strategy)
-
-## Anthropic
-
-### AWS Bedrock
-
-**Status**: Available and fully managed
-
-**Key Features**:
-- Anthropic's client SDKs support Bedrock directly
-- Models available: Claude Opus 4, Claude Sonnet 4, and earlier versions
-- 200,000 token context window
-- 60-minute timeout for inference calls
-- Supports both Messages API and Text Completions API
-
-**Implementation Requirements**:
-```javascript
-// Using Anthropic Bedrock SDK
-import AnthropicBedrock from '@anthropic-ai/bedrock-sdk';
-
-const client = new AnthropicBedrock({
-  awsAccessKey: process.env.AWS_ACCESS_KEY,
-  awsSecretKey: process.env.AWS_SECRET_KEY,
-  awsRegion: 'us-west-2',
-});
-
-// Alternative: Using AWS SDK directly
-import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
-```
-
-**Configuration Needed**:
-- AWS credentials (access key, secret key)
-- AWS region selection
-- Model ID mapping (e.g., `anthropic.claude-opus-4-1-20250805-v1:0`)
-- Timeout adjustments (default AWS SDK timeout is 1 minute, needs to be increased to 60 minutes)
-
-### Google Vertex AI
-
-**Status**: Available with FedRAMP High and IL-2 authorization (as of June 2025)
-
-**Key Features**:
-- Fully managed, serverless infrastructure
-- Supports Claude Opus 4, Claude Sonnet 3.7, and other models
-- Integration with Google Cloud services
-- Claude Code integration for agent development
-- Pay-as-you-go or provisioned throughput pricing
-
-**Implementation Requirements**:
-```javascript
-// Using Anthropic Vertex SDK
-import { AnthropicVertex } from '@anthropic-ai/vertex-sdk';
-
-const client = new AnthropicVertex({
-  region: 'us-central1',
-  projectId: 'my-project-id',
-});
-
-// Alternative: With custom GoogleAuth
-import { GoogleAuth } from 'google-auth-library';
-
-const client = new AnthropicVertex({
-  googleAuth: new GoogleAuth({
-    scopes: 'https://www.googleapis.com/auth/cloud-platform',
-    keyFile: '/path/to/service-account.json',
-  }),
-  region: 'us-central1',
-  projectId: 'my-project-id',
-});
-```
-
-**Configuration Needed**:
-- Google Cloud project ID
-- Location/region selection
-- Google Cloud authentication (service account or ADC)
-- Model ID mapping for Vertex AI (e.g., `claude-3-5-sonnet-v2@20241022`)
-
-## OpenAI
-
-### Azure OpenAI Service
-
-**Status**: Available with enterprise features
-
-**Key Features**:
-- Latest models: o4-mini, o3, gpt-4.1, GPT-4o, and others
-- New Responses API combining chat completions and assistants
-- Realtime API with WebRTC support
-- Private networking and regional availability
-- Microsoft Entra ID or API key authentication
-
-**Implementation Requirements**:
-```javascript
-// Option 1: Using OpenAI client with Azure (Recommended for 2025)
-import { AzureOpenAI } from 'openai';
-
-const client = new AzureOpenAI({
-  apiKey: process.env.AZURE_OPENAI_API_KEY,
-  apiVersion: '2025-04-01-preview',
-  endpoint: process.env.AZURE_OPENAI_ENDPOINT,
-  deployment: 'gpt-4o', // Your deployment name
-});
-
-// Option 2: With Azure AD Authentication (Recommended for production)
-import { DefaultAzureCredential, getBearerTokenProvider } from '@azure/identity';
-import { AzureOpenAI } from 'openai';
-import '@azure/openai/types';
-
-const credential = new DefaultAzureCredential();
-const scope = 'https://cognitiveservices.azure.com/.default';
-const azureADTokenProvider = getBearerTokenProvider(credential, scope);
-
-const client = new AzureOpenAI({ 
-  azureADTokenProvider, 
-  deployment: 'gpt-4o',
-  apiVersion: '2025-04-01-preview'
-});
-```
-
-**Configuration Needed**:
-- Azure OpenAI resource name
-- API key or Microsoft Entra ID credentials
-- API version selection
-- Deployment names for models
-
-## Google
-
-### Vertex AI Platform
-
-**Status**: Native platform (not an alternative, but enhanced deployment options)
-
-**Key Features**:
-- Regional and global endpoints
-- Managed APIs (no deployment required for Gemini models)
-- Support for tuned models
-- Enterprise-grade security and data residency
-- Live API with native audio (30 HD voices in 24 languages)
-
-**Implementation Requirements**:
-```javascript
-// Using Google Cloud Vertex AI SDK
-import { VertexAI } from '@google-cloud/vertexai';
-
-const vertexAI = new VertexAI({
-  project: 'your-cloud-project',
-  location: 'us-central1'
-});
-
-const generativeModel = vertexAI.getGenerativeModel({
-  model: 'gemini-2-5-pro-001',
-  generationConfig: { maxOutputTokens: 256 },
-  systemInstruction: {
-    role: 'system',
-    parts: [{ text: 'You are a helpful assistant' }]
-  }
-});
-
-// Alternative: Using Google Gen AI SDK (Unified SDK for 2025)
-import { GoogleGenAI } from '@google/genai';
-
-const ai = new GoogleGenAI({
-  project: 'your-cloud-project',
-  location: 'us-central1',
-  // Uses Application Default Credentials
-});
-```
-
-**Configuration Needed**:
-- Endpoint type selection (regional vs global)
-- Support for custom/tuned models
-- Live API configuration
-- Google Cloud authentication setup
-
-## DeepSeek
-
-### AWS Bedrock
-
-**Status**: Available as of March 2025 (first cloud provider to offer fully managed)
-
-**Key Features**:
-- DeepSeek-R1 available as serverless model
-- Available in US East (N. Virginia), US East (Ohio), and US West (Oregon)
-- Cross-region inference support
-- Enterprise-grade security and monitoring
-- Integration with Amazon Bedrock guardrails recommended
-
-**Implementation Requirements**:
-```javascript
-// Using AWS SDK for Bedrock
-import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
-
-const client = new BedrockRuntimeClient({ region: 'us-east-1' });
-
-const command = new InvokeModelCommand({
-  modelId: 'deepseek-r1', // Exact model ID from AWS Bedrock
-  body: JSON.stringify({
-    messages: [{ role: 'user', content: 'Hello' }],
-    max_tokens: 1024,
-  }),
-  contentType: 'application/json',
-  accept: 'application/json',
-});
-
-const response = await client.send(command);
-```
-
-### Azure AI Foundry
-
-**Status**: Available as of 2025
-
-**Key Features**:
-- Part of Azure AI Foundry platform
-- Built-in model evaluation tools
-- Red teaming and safety evaluations completed
-- SLA, security, and responsible AI commitments
-
-**Implementation Requirements**:
-```javascript
-// Azure AI Foundry API integration
-// Similar pattern to Azure OpenAI
-import { AzureAI } from '@azure/ai-inference';
-
-const client = new AzureAI({
-  endpoint: process.env.AZURE_AI_ENDPOINT,
-  apiKey: process.env.AZURE_AI_API_KEY,
-});
-
-// Model deployment through Azure AI Foundry
-const response = await client.completions.create({
-  model: 'deepseek-r1',
-  messages: [{ role: 'user', content: 'Hello' }],
-});
-```
-
-## Mistral AI
-
-### AWS Bedrock
-
-**Status**: Available with multiple models
-
-**Key Features**:
-- Models: Mistral 7B, Mixtral 8x7B, Pixtral Large (multimodal)
-- First major cloud to offer Pixtral Large
-- Serverless, fully managed
-- White-box solutions (weights and code available)
-- 50% lower batch inference pricing (as of April 2025)
-
-### Azure AI
-
-**Status**: Available
-
-**Key Features**:
-- Part of Azure AI ecosystem
-- Integration with Azure services
-
-### Google Vertex AI
-
-**Status**: Available in Model Garden
-
-**Key Features**:
-- Available through Vertex AI Model Garden
-- Standard Vertex AI features apply
-
-**Implementation Requirements**:
-```javascript
-// AWS Bedrock integration for Mistral
-import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
-
-const client = new BedrockRuntimeClient({ region: 'us-east-1' });
-
-const command = new InvokeModelCommand({
-  modelId: 'mistral.mixtral-8x7b-instruct-v0:1', // or 'mistral.mistral-7b-instruct-v0:2'
-  body: JSON.stringify({
-    prompt: '<s>[INST] Hello [/INST]',
-    max_tokens: 512,
-    temperature: 0.7,
-  }),
-  contentType: 'application/json',
-  accept: 'application/json',
-});
-
-// For Azure/Vertex, Mistral provides their own SDK which works across platforms
-import { Mistral } from '@mistralai/mistralai';
-
-const mistral = new Mistral({
-  apiKey: process.env.MISTRAL_API_KEY,
-  // For cloud platforms, configure the appropriate endpoint
-});
-```
-
-## XAI (Grok)
-
-### Azure AI Foundry
-
-**Status**: Available as of May 2025
-
-**Key Features**:
-- Models: Grok 3 and Grok 3 Mini
-- Full Azure SLA support
-- Structured outputs with JSON schema support
-- Functions and tools for agentic workflows
-- Direct Microsoft billing
-- OpenAI API format compatibility
-
-**Implementation Requirements**:
-```javascript
-// Azure AI Foundry integration for Grok
-// Uses OpenAI-compatible API format
-import { OpenAI } from 'openai';
-
-const client = new OpenAI({
-  apiKey: process.env.AZURE_API_KEY,
-  baseURL: 'https://YOUR-RESOURCE.openai.azure.com/openai',
-  defaultQuery: {
-    'api-version': '2025-05-01-preview'
-  },
-  defaultHeaders: {
-    'api-key': process.env.AZURE_API_KEY
-  }
-});
-
-// Use with deployment name
-const response = await client.chat.completions.create({
-  model: 'grok-3', // or 'grok-3-mini'
-  messages: [{ role: 'user', content: 'Hello' }],
-});
-```
-
-**Note**: No AWS integration announced as of 2025, but xAI's infrastructure supports potential multi-cloud deployment.
-
-## Implementation Strategy
-
-### 1. Create Platform Abstraction Layer
-
-```javascript
-// src/providers/platforms/index.js
-class PlatformProvider {
-  constructor(provider, platform, config) {
-    this.provider = provider; // 'anthropic', 'openai', etc.
-    this.platform = platform; // 'native', 'bedrock', 'azure', 'vertex'
-    this.config = config;
-  }
-
-  async createClient() {
-    // Platform-specific client initialization
-  }
-
-  async sendRequest(messages, options) {
-    // Platform-specific request handling
-  }
-}
-```
-
-### 2. Extend Provider Configuration
-
-```javascript
-// Example configuration
-{
-  "providers": {
-    "anthropic": {
-      "platform": "bedrock", // or "native", "vertex"
-      "bedrock": {
-        "awsRegion": "us-west-2",
-        "awsAccessKey": "...",
-        "awsSecretKey": "..."
-      },
-      "vertex": {
-        "projectId": "...",
-        "location": "us-central1"
-      }
-    }
-  }
-}
-```
-
-### 3. Update Environment Variables
-
-```bash
-# Platform selection
-ANTHROPIC_PLATFORM=bedrock
-OPENAI_PLATFORM=azure
-DEEPSEEK_PLATFORM=bedrock
-MISTRAL_PLATFORM=bedrock
-XAI_PLATFORM=azure
-
-# Platform-specific credentials
-AWS_ACCESS_KEY_ID=...
-AWS_SECRET_ACCESS_KEY=...
-AWS_REGION=us-west-2
-
-AZURE_OPENAI_ENDPOINT=...
-AZURE_OPENAI_API_KEY=...
-
-GOOGLE_CLOUD_PROJECT=...
-GOOGLE_CLOUD_LOCATION=...
-```
-
-### 4. Model Mapping
-
-Create a mapping system for model names across platforms:
-
-```javascript
-const MODEL_MAPPINGS = {
-  anthropic: {
-    bedrock: {
-      'claude-opus-4': 'anthropic.claude-opus-4-1-20250805-v1:0',
-      'claude-sonnet-4': 'anthropic.claude-sonnet-4-20250514-v1:0'
-    },
-    vertex: {
-      'claude-opus-4': 'claude-opus-4@001',
-      'claude-sonnet-4': 'claude-sonnet-4@001'
-    }
-  }
-};
-```
-
-### 5. Testing Strategy
-
-- Create integration tests for each platform
-- Mock platform-specific responses
-- Test failover between platforms
-- Verify feature parity across platforms
-
-### 6. Documentation Updates
-
-- Update README with platform configuration
-- Add platform-specific examples
-- Document limitations and differences
-- Provide migration guides
-
+# Alternative Cloud Provider Integrations for Converse MCP Server
+
+This document outlines alternative API providers for each AI provider currently supported in the Converse MCP Server, along with implementation requirements for adding support.
+
+## Table of Contents
+- [Anthropic](#anthropic)
+  - [AWS Bedrock](#aws-bedrock)
+  - [Google Vertex AI](#google-vertex-ai)
+- [OpenAI](#openai)
+  - [Azure OpenAI Service](#azure-openai-service)
+- [Google](#google)
+  - [Vertex AI Platform](#vertex-ai-platform)
+- [DeepSeek](#deepseek)
+  - [AWS Bedrock](#aws-bedrock-1)
+  - [Azure AI Foundry](#azure-ai-foundry)
+- [Mistral AI](#mistral-ai)
+  - [AWS Bedrock](#aws-bedrock-2)
+  - [Azure AI](#azure-ai)
+  - [Google Vertex AI](#google-vertex-ai-1)
+- [XAI (Grok)](#xai-grok)
+  - [Azure AI Foundry](#azure-ai-foundry-1)
+- [Implementation Strategy](#implementation-strategy)
+
+## Anthropic
+
+### AWS Bedrock
+
+**Status**: Available and fully managed
+
+**Key Features**:
+- Anthropic's client SDKs support Bedrock directly
+- Models available: Claude Opus 4, Claude Sonnet 4, and earlier versions
+- 200,000 token context window
+- 60-minute timeout for inference calls
+- Supports both Messages API and Text Completions API
+
+**Implementation Requirements**:
+```javascript
+// Using Anthropic Bedrock SDK
+import AnthropicBedrock from '@anthropic-ai/bedrock-sdk';
+
+const client = new AnthropicBedrock({
+  awsAccessKey: process.env.AWS_ACCESS_KEY,
+  awsSecretKey: process.env.AWS_SECRET_KEY,
+  awsRegion: 'us-west-2',
+});
+
+// Alternative: Using AWS SDK directly
+import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
+```
+
+**Configuration Needed**:
+- AWS credentials (access key, secret key)
+- AWS region selection
+- Model ID mapping (e.g., `anthropic.claude-opus-4-1-20250805-v1:0`)
+- Timeout adjustments (default AWS SDK timeout is 1 minute, needs to be increased to 60 minutes)
+
+### Google Vertex AI
+
+**Status**: Available with FedRAMP High and IL-2 authorization (as of June 2025)
+
+**Key Features**:
+- Fully managed, serverless infrastructure
+- Supports Claude Opus 4, Claude Sonnet 3.7, and other models
+- Integration with Google Cloud services
+- Claude Code integration for agent development
+- Pay-as-you-go or provisioned throughput pricing
+
+**Implementation Requirements**:
+```javascript
+// Using Anthropic Vertex SDK
+import { AnthropicVertex } from '@anthropic-ai/vertex-sdk';
+
+const client = new AnthropicVertex({
+  region: 'us-central1',
+  projectId: 'my-project-id',
+});
+
+// Alternative: With custom GoogleAuth
+import { GoogleAuth } from 'google-auth-library';
+
+const client = new AnthropicVertex({
+  googleAuth: new GoogleAuth({
+    scopes: 'https://www.googleapis.com/auth/cloud-platform',
+    keyFile: '/path/to/service-account.json',
+  }),
+  region: 'us-central1',
+  projectId: 'my-project-id',
+});
+```
+
+**Configuration Needed**:
+- Google Cloud project ID
+- Location/region selection
+- Google Cloud authentication (service account or ADC)
+- Model ID mapping for Vertex AI (e.g., `claude-3-5-sonnet-v2@20241022`)
+
+## OpenAI
+
+### Azure OpenAI Service
+
+**Status**: Available with enterprise features
+
+**Key Features**:
+- Latest models: o4-mini, o3, gpt-4.1, GPT-4o, and others
+- New Responses API combining chat completions and assistants
+- Realtime API with WebRTC support
+- Private networking and regional availability
+- Microsoft Entra ID or API key authentication
+
+**Implementation Requirements**:
+```javascript
+// Option 1: Using OpenAI client with Azure (Recommended for 2025)
+import { AzureOpenAI } from 'openai';
+
+const client = new AzureOpenAI({
+  apiKey: process.env.AZURE_OPENAI_API_KEY,
+  apiVersion: '2025-04-01-preview',
+  endpoint: process.env.AZURE_OPENAI_ENDPOINT,
+  deployment: 'gpt-4o', // Your deployment name
+});
+
+// Option 2: With Azure AD Authentication (Recommended for production)
+import { DefaultAzureCredential, getBearerTokenProvider } from '@azure/identity';
+import { AzureOpenAI } from 'openai';
+import '@azure/openai/types';
+
+const credential = new DefaultAzureCredential();
+const scope = 'https://cognitiveservices.azure.com/.default';
+const azureADTokenProvider = getBearerTokenProvider(credential, scope);
+
+const client = new AzureOpenAI({ 
+  azureADTokenProvider, 
+  deployment: 'gpt-4o',
+  apiVersion: '2025-04-01-preview'
+});
+```
+
+**Configuration Needed**:
+- Azure OpenAI resource name
+- API key or Microsoft Entra ID credentials
+- API version selection
+- Deployment names for models
+
+## Google
+
+### Vertex AI Platform
+
+**Status**: Native platform (not an alternative, but enhanced deployment options)
+
+**Key Features**:
+- Regional and global endpoints
+- Managed APIs (no deployment required for Gemini models)
+- Support for tuned models
+- Enterprise-grade security and data residency
+- Live API with native audio (30 HD voices in 24 languages)
+
+**Implementation Requirements**:
+```javascript
+// Using Google Cloud Vertex AI SDK
+import { VertexAI } from '@google-cloud/vertexai';
+
+const vertexAI = new VertexAI({
+  project: 'your-cloud-project',
+  location: 'us-central1'
+});
+
+const generativeModel = vertexAI.getGenerativeModel({
+  model: 'gemini-2-5-pro-001',
+  generationConfig: { maxOutputTokens: 256 },
+  systemInstruction: {
+    role: 'system',
+    parts: [{ text: 'You are a helpful assistant' }]
+  }
+});
+
+// Alternative: Using Google Gen AI SDK (Unified SDK for 2025)
+import { GoogleGenAI } from '@google/genai';
+
+const ai = new GoogleGenAI({
+  project: 'your-cloud-project',
+  location: 'us-central1',
+  // Uses Application Default Credentials
+});
+```
+
+**Configuration Needed**:
+- Endpoint type selection (regional vs global)
+- Support for custom/tuned models
+- Live API configuration
+- Google Cloud authentication setup
+
+## DeepSeek
+
+### AWS Bedrock
+
+**Status**: Available as of March 2025 (first cloud provider to offer fully managed)
+
+**Key Features**:
+- DeepSeek-R1 available as serverless model
+- Available in US East (N. Virginia), US East (Ohio), and US West (Oregon)
+- Cross-region inference support
+- Enterprise-grade security and monitoring
+- Integration with Amazon Bedrock guardrails recommended
+
+**Implementation Requirements**:
+```javascript
+// Using AWS SDK for Bedrock
+import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
+
+const client = new BedrockRuntimeClient({ region: 'us-east-1' });
+
+const command = new InvokeModelCommand({
+  modelId: 'deepseek-r1', // Exact model ID from AWS Bedrock
+  body: JSON.stringify({
+    messages: [{ role: 'user', content: 'Hello' }],
+    max_tokens: 1024,
+  }),
+  contentType: 'application/json',
+  accept: 'application/json',
+});
+
+const response = await client.send(command);
+```
+
+### Azure AI Foundry
+
+**Status**: Available as of 2025
+
+**Key Features**:
+- Part of Azure AI Foundry platform
+- Built-in model evaluation tools
+- Red teaming and safety evaluations completed
+- SLA, security, and responsible AI commitments
+
+**Implementation Requirements**:
+```javascript
+// Azure AI Foundry API integration
+// Similar pattern to Azure OpenAI
+import { AzureAI } from '@azure/ai-inference';
+
+const client = new AzureAI({
+  endpoint: process.env.AZURE_AI_ENDPOINT,
+  apiKey: process.env.AZURE_AI_API_KEY,
+});
+
+// Model deployment through Azure AI Foundry
+const response = await client.completions.create({
+  model: 'deepseek-r1',
+  messages: [{ role: 'user', content: 'Hello' }],
+});
+```
+
+## Mistral AI
+
+### AWS Bedrock
+
+**Status**: Available with multiple models
+
+**Key Features**:
+- Models: Mistral 7B, Mixtral 8x7B, Pixtral Large (multimodal)
+- First major cloud to offer Pixtral Large
+- Serverless, fully managed
+- White-box solutions (weights and code available)
+- 50% lower batch inference pricing (as of April 2025)
+
+### Azure AI
+
+**Status**: Available
+
+**Key Features**:
+- Part of Azure AI ecosystem
+- Integration with Azure services
+
+### Google Vertex AI
+
+**Status**: Available in Model Garden
+
+**Key Features**:
+- Available through Vertex AI Model Garden
+- Standard Vertex AI features apply
+
+**Implementation Requirements**:
+```javascript
+// AWS Bedrock integration for Mistral
+import { BedrockRuntimeClient, InvokeModelCommand } from '@aws-sdk/client-bedrock-runtime';
+
+const client = new BedrockRuntimeClient({ region: 'us-east-1' });
+
+const command = new InvokeModelCommand({
+  modelId: 'mistral.mixtral-8x7b-instruct-v0:1', // or 'mistral.mistral-7b-instruct-v0:2'
+  body: JSON.stringify({
+    prompt: '<s>[INST] Hello [/INST]',
+    max_tokens: 512,
+    temperature: 0.7,
+  }),
+  contentType: 'application/json',
+  accept: 'application/json',
+});
+
+// For Azure/Vertex, Mistral provides their own SDK which works across platforms
+import { Mistral } from '@mistralai/mistralai';
+
+const mistral = new Mistral({
+  apiKey: process.env.MISTRAL_API_KEY,
+  // For cloud platforms, configure the appropriate endpoint
+});
+```
+
+## XAI (Grok)
+
+### Azure AI Foundry
+
+**Status**: Available as of May 2025
+
+**Key Features**:
+- Models: Grok 3 and Grok 3 Mini
+- Full Azure SLA support
+- Structured outputs with JSON schema support
+- Functions and tools for agentic workflows
+- Direct Microsoft billing
+- OpenAI API format compatibility
+
+**Implementation Requirements**:
+```javascript
+// Azure AI Foundry integration for Grok
+// Uses OpenAI-compatible API format
+import { OpenAI } from 'openai';
+
+const client = new OpenAI({
+  apiKey: process.env.AZURE_API_KEY,
+  baseURL: 'https://YOUR-RESOURCE.openai.azure.com/openai',
+  defaultQuery: {
+    'api-version': '2025-05-01-preview'
+  },
+  defaultHeaders: {
+    'api-key': process.env.AZURE_API_KEY
+  }
+});
+
+// Use with deployment name
+const response = await client.chat.completions.create({
+  model: 'grok-3', // or 'grok-3-mini'
+  messages: [{ role: 'user', content: 'Hello' }],
+});
+```
+
+**Note**: No AWS integration announced as of 2025, but xAI's infrastructure supports potential multi-cloud deployment.
+
+## Implementation Strategy
+
+### 1. Create Platform Abstraction Layer
+
+```javascript
+// src/providers/platforms/index.js
+class PlatformProvider {
+  constructor(provider, platform, config) {
+    this.provider = provider; // 'anthropic', 'openai', etc.
+    this.platform = platform; // 'native', 'bedrock', 'azure', 'vertex'
+    this.config = config;
+  }
+
+  async createClient() {
+    // Platform-specific client initialization
+  }
+
+  async sendRequest(messages, options) {
+    // Platform-specific request handling
+  }
+}
+```
+
+### 2. Extend Provider Configuration
+
+```javascript
+// Example configuration
+{
+  "providers": {
+    "anthropic": {
+      "platform": "bedrock", // or "native", "vertex"
+      "bedrock": {
+        "awsRegion": "us-west-2",
+        "awsAccessKey": "...",
+        "awsSecretKey": "..."
+      },
+      "vertex": {
+        "projectId": "...",
+        "location": "us-central1"
+      }
+    }
+  }
+}
+```
+
+### 3. Update Environment Variables
+
+```bash
+# Platform selection
+ANTHROPIC_PLATFORM=bedrock
+OPENAI_PLATFORM=azure
+DEEPSEEK_PLATFORM=bedrock
+MISTRAL_PLATFORM=bedrock
+XAI_PLATFORM=azure
+
+# Platform-specific credentials
+AWS_ACCESS_KEY_ID=...
+AWS_SECRET_ACCESS_KEY=...
+AWS_REGION=us-west-2
+
+AZURE_OPENAI_ENDPOINT=...
+AZURE_OPENAI_API_KEY=...
+
+GOOGLE_CLOUD_PROJECT=...
+GOOGLE_CLOUD_LOCATION=...
+```
+
+### 4. Model Mapping
+
+Create a mapping system for model names across platforms:
+
+```javascript
+const MODEL_MAPPINGS = {
+  anthropic: {
+    bedrock: {
+      'claude-opus-4': 'anthropic.claude-opus-4-1-20250805-v1:0',
+      'claude-sonnet-4': 'anthropic.claude-sonnet-4-20250514-v1:0'
+    },
+    vertex: {
+      'claude-opus-4': 'claude-opus-4@001',
+      'claude-sonnet-4': 'claude-sonnet-4@001'
+    }
+  }
+};
+```
+
+### 5. Testing Strategy
+
+- Create integration tests for each platform
+- Mock platform-specific responses
+- Test failover between platforms
+- Verify feature parity across platforms
+
+### 6. Documentation Updates
+
+- Update README with platform configuration
+- Add platform-specific examples
+- Document limitations and differences
+- Provide migration guides
+
 This implementation would allow users to choose their preferred cloud platform for each AI provider, enabling better cost optimization, compliance requirements, and geographic distribution.