Module sampling 

Client-side sampling support for handling server-initiated requests

This module provides production-grade LLM backend integration for processing sampling requests from servers, enabling bidirectional LLM interactions in the MCP protocol.

Features

• Multi-provider support: OpenAI, Anthropic, and extensible architecture
• MCP protocol compliance: Full CreateMessageRequest → CreateMessageResult flow
• Production-grade error handling: Comprehensive error types and recovery
• Conversation context: Proper message history management
• Configuration: Flexible backend selection and parameter tuning
• Async-first: Send + Sync throughout with proper async patterns

Example

use turbomcp_client::sampling::{LLMBackendConfig, LLMProvider, ProductionSamplingHandler};

let config = LLMBackendConfig {
    provider: LLMProvider::OpenAI {
        api_key: std::env::var("OPENAI_API_KEY").unwrap(),
        base_url: None,
        organization: None,
    },
    default_model: Some("gpt-4".to_string()),
    timeout_seconds: 30,
    max_retries: 3,
};

let handler = ProductionSamplingHandler::new(config)?;

Structs

DefaultSamplingHandler

Default sampling handler - provides echo functionality for testing/development

LLMBackendConfig

Comprehensive backend configuration

MockLLMHandler

Mock LLM handler for testing

ProductionSamplingHandler

Production-grade sampling handler with real LLM backend integration

Enums

LLMBackendError

Comprehensive error types for LLM backend operations

LLMProvider

Supported LLM providers

Traits

SamplingHandler

Handler for server-initiated sampling requests