/**
 * Client Sampling Utility
 *
 * This module provides utilities for MCP clients that don't natively support sampling.
 * It offers a simple configuration-based API to add sampling capability to any MCP client.
 *
 * @see https://modelcontextprotocol.io/specification/2025-06-18/client/sampling
 */ import type { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { type CreateMessageRequest, type CreateMessageResultWithTools } from "@modelcontextprotocol/sdk/types.js";
import { type AISDKMessage } from "./utils.js";
/**
 * AI SDK handler function type
 */ export type AISDKHandler = (params: {
  systemPrompt?: string;
  messages: AISDKMessage[];
  modelPreferences?: CreateMessageRequest["params"]["modelPreferences"];
  tools?: CreateMessageRequest["params"]["tools"];
  helpers?: {
    tool: (...args: any[]) => any;
    jsonSchema: (...args: any[]) => any;
  };
}) => Promise<CreateMessageResultWithTools>;
/**
 * Client sampling configuration
 */ export interface ClientSamplingConfig {
  /**
   * AI SDK handler function that generates completions
   */ handler: AISDKHandler;
  /**
   * AI SDK helpers for tool conversion (tool and jsonSchema from "ai" package)
   * Required for proper tool support when MCP server sends tools in createMessage request
   */ helpers?: {
    tool: (...args: any[]) => any;
    jsonSchema: (...args: any[]) => any;
  };
  /**
   * Optional model selection map for preference-based routing
   */ modelMap?: {
    hints?: Record<string, string>;
    priorities?: {
      speed?: string;
      intelligence?: string;
      cost?: string;
    };
    default?: string;
  };
}
/**
 * Create a configured client sampling handler
 *
 * This is the main entry point for adding sampling support to MCP clients.
 * It takes a configuration object and returns a ready-to-use sampling handler.
 *
 * @param config - Client sampling configuration
 * @returns MCP sampling handler function
 *
 * @example
 * ```typescript
 * import { createClientSampling } from "@mcpc/mcp-sampling-ai-provider";
 * import { generateText } from "ai";
 *
 * const samplingHandler = createClientSampling({
 *   handler: async (params) => {
 *     const result = await generateText({
 *       model: "openai/gpt-5-mini",
 *       messages: params.messages,
 *     });
 *     return {
 *       model: "openai/gpt-5-mini",
 *       role: "assistant",
 *       content: { type: "text", text: result.text },
 *       stopReason: result.finishReason === "stop" ? "endTurn" : "maxTokens",
 *     };
 *   },
 * });
 *
 * server.setRequestHandler(CreateMessageRequestSchema, samplingHandler);
 * ```
 */ export declare function createClientSampling(config: ClientSamplingConfig): (request: CreateMessageRequest) => Promise<CreateMessageResultWithTools>;
/**
 * Helper to select model based on MCP model preferences
 *
 * @param preferences - MCP model preferences
 * @param modelMap - Map of hint names/priorities to your model identifiers
 * @returns Selected model identifier
 *
 * @example
 * ```typescript
 * const model = selectModelFromPreferences(modelPreferences, {
 *   hints: {
 *     "gpt-4o": "openai/gpt-4o",
 *     "claude": "anthropic/claude-3-5-sonnet-20241022",
 *   },
 *   priorities: {
 *     speed: "openai/gpt-4o-mini",
 *     intelligence: "openai/gpt-4o",
 *     cost: "openai/gpt-4o-mini",
 *   },
 *   default: "openai/gpt-5-mini",
 * });
 * ```
 */ export declare function selectModelFromPreferences(preferences: CreateMessageRequest["params"]["modelPreferences"] | undefined, modelMap: {
  /**
     * Map hint names to model identifiers
     */ hints?: Record<string, string>;
  /**
     * Map priority types to model identifiers
     */ priorities?: {
    speed?: string;
    intelligence?: string;
    cost?: string;
  };
  /**
     * Default model if no match found
     */ default: string;
}): string;
/**
 * Add sampling support to an existing MCP client
 *
 * This function configures an existing MCP Client instance to handle sampling requests
 * from the server. It sets up the necessary request handler for createMessage requests.
 *
 * @param client - Existing MCP Client instance
 * @param config - Sampling handler configuration
 *
 * @example
 * ```typescript
 * import { Client } from "@modelcontextprotocol/sdk/client/index.js";
 * import { setupClientSampling } from "@mcpc/mcp-sampling-ai-provider";
 * import { generateText } from "ai";
 *
 * const client = new Client(
 *   { name: "my-client", version: "1.0.0" },
 *   { capabilities: { sampling: {} } }
 * );
 *
 * setupClientSampling(client, {
 *   handler: async (params) => {
 *     const result = await generateText({
 *       model: "openai/gpt-5-mini",
 *       messages: params.messages,
 *     });
 *     return {
 *       model: "openai/gpt-5-mini",
 *       role: "assistant",
 *       content: { type: "text", text: result.text },
 *       stopReason: result.finishReason === "stop" ? "endTurn" : "maxTokens",
 *     };
 *   },
 * });
 *
 * await client.connect(transport);
 * ```
 */ export declare function setupClientSampling(client: Client, config: ClientSamplingConfig): void;
//# sourceMappingURL=client-sampling.d.ts.map