MCP Elicitations Demo Server

import { z } from "zod"; import { zodToJsonSchema } from "zod-to-json-schema"; import { Server } from "@modelcontextprotocol/sdk/server/index.js"; import { requestSampling, MessageContent, Message } from "../lib/sampling.js"; const TextContentSchema = z.object({ type: z.literal("text"), text: z.string().describe("Text content"), }); const ImageContentSchema = z.object({ type: z.literal("image"), data: z.string().describe("Base64-encoded image data"), mimeType: z .string() .describe("Image MIME type (e.g., 'image/jpeg', 'image/png')"), }); const AudioContentSchema = z.object({ type: z.literal("audio"), data: z.string().describe("Base64-encoded audio data"), mimeType: z .string() .describe("Audio MIME type (e.g., 'audio/wav', 'audio/mp3')"), }); const MessageContentSchema = z.discriminatedUnion("type", [ TextContentSchema, ImageContentSchema, AudioContentSchema, ]); const MessageSchema = z.object({ role: z.enum(["user", "assistant"]).describe("Message role"), content: MessageContentSchema.describe("Message content"), }); const MultimodalSampleSchema = z.object({ messages: z .array(MessageSchema) .min(1) .describe("Array of messages with various content types"), systemPrompt: z .string() .optional() .describe("System prompt to guide the LLM's behavior"), maxTokens: z .number() .default(100) .describe("Maximum number of tokens to generate"), }); export const sampleMultimodalTool = { name: "sampleMultimodal", description: "Sample from an LLM with multimodal content (text, images, audio)", inputSchema: zodToJsonSchema(MultimodalSampleSchema), handler: async (args: any, request: any, server: Server) => { const validatedArgs = MultimodalSampleSchema.parse(args); const { messages, systemPrompt, maxTokens } = validatedArgs; // Convert validated messages to our internal format const typedMessages: Message[] = messages.map((msg: any) => ({ role: msg.role, content: msg.content as MessageContent, })); const result = await requestSampling( { messages: typedMessages, systemPrompt, maxTokens, }, server ); // Provide detailed response information const contentSummary = messages .map((msg: any, index: number) => { const contentType = msg.content.type; if (contentType === "text") { return `Message ${index + 1} (${ msg.role }): Text - "${msg.content.text.substring(0, 50)}${ msg.content.text.length > 50 ? "..." : "" }"`; } else if (contentType === "image") { return `Message ${index + 1} (${msg.role}): Image - ${ msg.content.mimeType }`; } else if (contentType === "audio") { return `Message ${index + 1} (${msg.role}): Audio - ${ msg.content.mimeType }`; } return `Message ${index + 1} (${msg.role}): ${contentType}`; }) .join("\n"); return { content: [ { type: "text" as const, text: `Multimodal Sampling Result: Input Summary: ${contentSummary} Model: ${result.model || "unknown"} Stop Reason: ${result.stopReason || "unknown"} Response: ${result.content.text}`, }, ], }; }, };