azure-image-generation-server.jsā¢12.1 kB
#!/usr/bin/env node
import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
import { CallToolRequestSchema, ListToolsRequestSchema } from '@modelcontextprotocol/sdk/types.js';
class AzureImageGenerationMCPServer {
constructor() {
this.server = new Server({
name: 'Image Generation',
version: '1.0.0',
}, {
capabilities: {
tools: {},
},
});
// Azure configuration
this.azureApiKey = process.env.AZURE_IMAGE_API_KEY;
this.azureBaseUrl = process.env.AZURE_IMAGE_BASE_URL;
if (!this.azureApiKey) {
throw new Error('AZURE_IMAGE_API_KEY environment variable is required. Please set it in your .env file or LibreChat configuration.');
}
if (!this.azureBaseUrl) {
throw new Error('AZURE_IMAGE_BASE_URL environment variable is required. Please set it to your Azure endpoint (e.g., https://your-resource.openai.azure.com/openai/deployments)');
}
this.setupToolHandlers();
}
setupToolHandlers() {
// List available tools
this.server.setRequestHandler(ListToolsRequestSchema, async () => ({
tools: [
{
name: 'generate_image',
description: 'šØ Create stunning AI-generated images using Azure DALL-E 3 or FLUX models with intelligent model selection',
inputSchema: {
type: 'object',
properties: {
prompt: {
type: 'string',
description: 'Describe the image you want to create in natural language. Be detailed for best results. Examples: "A serene mountain landscape at sunset", "Modern minimalist logo design", "Cute cartoon mascot for a coffee shop"'
},
model: {
type: 'string',
description: 'Choose AI model: "dall-e-3" (photorealistic, artistic), "flux" (creative, flexible), or "auto" (smart selection based on prompt)',
enum: ['dall-e-3', 'flux', 'auto'],
default: 'auto'
},
size: {
type: 'string',
description: 'Image dimensions: Square (1024x1024) for social media, Wide (1792x1024) for banners, Tall (1024x1792) for posters',
enum: ['1024x1024', '1792x1024', '1024x1792'],
default: '1024x1024'
},
style: {
type: 'string',
description: 'Visual style (DALL-E only): "vivid" for dramatic/artistic, "natural" for realistic/subdued',
enum: ['vivid', 'natural'],
default: 'vivid'
},
quality: {
type: 'string',
description: 'Image quality (DALL-E only): "standard" for faster generation, "hd" for higher detail',
enum: ['standard', 'hd'],
default: 'standard'
}
},
required: ['prompt']
}
}
]
}));
// Handle tool calls
this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
const { name, arguments: args } = request.params;
if (name === 'generate_image') {
return await this.generateImage(args);
}
throw new Error(`Unknown tool: ${name}`);
});
}
// Simple model selection: FLUX by default, DALL-E when explicitly requested
selectModel(prompt, requestedModel) {
if (requestedModel !== 'auto') {
return requestedModel;
}
// Check if user explicitly mentions DALL-E
const lowerPrompt = prompt.toLowerCase();
const isDalleRequest = ['dall-e', 'dalle', 'dall e'].some(keyword => lowerPrompt.includes(keyword));
return isDalleRequest ? 'dall-e-3' : 'flux';
}
async generateImage(args) {
const startTime = Date.now();
// Starting image generation
const { prompt, model: requestedModel = 'auto', size = '1024x1024', style = 'vivid', quality = 'standard' } = args;
if (!prompt) {
throw new Error('Prompt is required for image generation');
}
try {
const modelSelectionStart = Date.now();
const selectedModel = this.selectModel(prompt, requestedModel);
const modelSelectionTime = Date.now() - modelSelectionStart;
// Model selection and generation setup complete
const apiCallStart = Date.now();
let result;
if (selectedModel === 'dall-e-3') {
result = await this.generateWithDallE(prompt, size, style, quality);
} else {
result = await this.generateWithFlux(prompt, size);
}
const apiCallTime = Date.now() - apiCallStart;
const totalTime = Date.now() - startTime;
// Image generation completed successfully
return {
content: [
{
type: 'text',
text: `šØ **Your AI-Generated Image is Ready!**\n\nāØ **Created from:** "${prompt}"\nš¤ **AI Model:** ${selectedModel.toUpperCase()} ${selectedModel === 'dall-e-3' ? '(Photorealistic AI)' : '(Creative AI)'}\nš **Size:** ${size}\nā±ļø **Generation Time:** ${(totalTime / 1000).toFixed(1)}s\n${requestedModel === 'auto' ? `š§ **Smart Selection:** Chose ${selectedModel.toUpperCase()} based on your prompt\n` : ''}\nš¼ļø Your custom image has been generated and is displayed below. Feel free to save, share, or use it however you'd like!`
},
{
type: 'image',
mimeType: 'image/png',
data: result.imageData
}
]
};
} catch (error) {
const totalTime = Date.now() - startTime;
console.error(`ā [MCP] Error after ${totalTime}ms:`, error);
throw new Error(`Failed to generate image: ${error.message}`);
}
}
async generateWithDallE(prompt, size, style, quality) {
const dalleStartTime = Date.now();
const endpoint = `${this.azureBaseUrl}/dall-e-3/images/generations?api-version=2024-02-01`;
const requestBody = {
model: "dall-e-3",
prompt: prompt,
size: size,
style: style,
quality: quality,
n: 1
};
// Making request to Azure DALL-E 3 API
// DALL-E endpoint configured
// DALL-E request body prepared
const fetchStart = Date.now();
const response = await fetch(endpoint, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Authorization': `Bearer ${this.azureApiKey}`
},
body: JSON.stringify(requestBody)
});
const fetchTime = Date.now() - fetchStart;
// DALL-E response received
// DALL-E response status received
if (!response.ok) {
const errorText = await response.text();
console.error(`šµ [DALL-E] Azure API error response: ${errorText}`);
throw new Error(`Azure DALL-E API error: ${response.status} - ${errorText}`);
}
const jsonParseStart = Date.now();
const result = await response.json();
const jsonParseTime = Date.now() - jsonParseStart;
// DALL-E JSON parsing completed
// DALL-E response structure validated
// DALL-E returns URL, we need to fetch the image and convert to base64
if (result.data && result.data[0] && result.data[0].url) {
const imageUrl = result.data[0].url;
// Fetching image from DALL-E URL
const imageFetchStart = Date.now();
const imageResponse = await fetch(imageUrl, {
headers: {
'Accept': 'image/png, image/jpeg, image/*',
'Accept-Encoding': 'gzip, deflate, br'
},
timeout: 60000 // 60 second timeout
});
if (!imageResponse.ok) {
throw new Error(`Failed to fetch image: ${imageResponse.status}`);
}
const bufferStart = Date.now();
// Use arrayBuffer for Node.js fetch compatibility
const arrayBuffer = await imageResponse.arrayBuffer();
const imageBuffer = Buffer.from(arrayBuffer);
const bufferTime = Date.now() - bufferStart;
const base64Start = Date.now();
const imageData = imageBuffer.toString('base64');
const base64Time = Date.now() - base64Start;
const imageFetchTime = Date.now() - imageFetchStart;
// Image processing completed
const totalDalleTime = Date.now() - dalleStartTime;
return { imageData };
} else {
console.error('Could not find image URL in DALL-E response:', JSON.stringify(result, null, 2));
throw new Error('No image URL returned from Azure DALL-E API');
}
}
async generateWithFlux(prompt, size) {
const fluxStartTime = Date.now();
const endpoint = `${this.azureBaseUrl}/FLUX.1-Kontext-pro/images/generations?api-version=2025-04-01-preview`;
const [width, height] = size.split('x').map(Number);
const requestBody = {
prompt: prompt,
size: `${width}x${height}`,
n: 1,
model: "flux.1-kontext-pro"
};
// Making request to Azure FLUX API
// FLUX endpoint configured
// FLUX request body prepared
const fetchStart = Date.now();
const response = await fetch(endpoint, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Authorization': `Bearer ${this.azureApiKey}`
},
body: JSON.stringify(requestBody)
});
const fetchTime = Date.now() - fetchStart;
// FLUX response received
// FLUX response status received
if (!response.ok) {
const errorText = await response.text();
console.error('Azure FLUX API error response:', errorText);
throw new Error(`Azure FLUX API error: ${response.status} - ${errorText}`);
}
const jsonParseStart = Date.now();
const result = await response.json();
const jsonParseTime = Date.now() - jsonParseStart;
if (result.data && result.data[0] && result.data[0].b64_json) {
const imageData = result.data[0].b64_json;
const totalFluxTime = Date.now() - fluxStartTime;
// Base64 image data found
return { imageData };
} else {
console.error('Could not find b64_json in FLUX response:', JSON.stringify(result, null, 2));
throw new Error('No image data returned from Azure FLUX API');
}
}
async run() {
const transport = new StdioServerTransport();
await this.server.connect(transport);
console.error('Azure Image Generation MCP server running on stdio');
}
}
const server = new AzureImageGenerationMCPServer();
server.run().catch(console.error);