Http Transport client with sampling

[kaniosm]

Pre-submission Checklist

• I have verified that this discussion would not be more appropriate as an issue in a specific repository
• I have searched existing discussions to avoid duplicates

Discussion Topic

Anyone can provide a sample with a client that supports sampling?
I tried connecting a Python client to the MCP Sse Server sample, but the callback method is never reached in my Python code.
As for .NET I did not find any sample demonstrating sampling.

[kaniosm]

Sample Python code:

The communication to get the tools and execute the tools works as expected, but when the server responds with the sampling request the callback will not execute at the client and the operation will cancel at the server.

from mcp.client.streamable_http import streamablehttp_client
from mcp import ClientSession
from mcp import ClientSession, StdioServerParameters, types
from mcp.client.stdio import stdio_client


async def handle_sampling_message(message: types.CreateMessageRequestParams,) -> types.CreateMessageResult:
    m = message
    return types.CreateMessageResult(
        role="assistant",
        content=types.TextContent(
            type="text",
            text="Hello, world! from model",
        ),
        model="gpt-3.5-turbo",
        stopReason="endTurn",
    )


async def main():
    # Connect to a streamable HTTP server
    async with streamablehttp_client("http://localhost:3001") as (
        read_stream,
        write_stream,
        _,
    ):
        # Create a session using the client streams
        async with ClientSession(read_stream, write_stream, sampling_callback=handle_sampling_message) as session:
            # Initialize the connection
            await session.initialize()
            tools = await session.list_tools()
            # Call a tool
            tool_result = await session.call_tool("sampleLLM", {"prompt": "hello", "maxTokens": 100})
            print(tool_result)


if __name__ == "__main__":
    import asyncio
    asyncio.run(main())

[kaniosm]

Got a working sample in C#, but my Python code still will not work:

using Microsoft.Extensions.Configuration;
using Microsoft.Extensions.Hosting;
using ModelContextProtocol.Client;
using ModelContextProtocol.Protocol.Transport;
using ModelContextProtocol.Protocol.Types;

var builder = Host.CreateApplicationBuilder(args);

builder.Configuration
    .AddEnvironmentVariables()
    .AddUserSecrets<Program>();

var clientTransport = new ModelContextProtocol.Protocol.Transport.SseClientTransport(
    new SseClientTransportOptions
    {
        Endpoint = new Uri("http://localhost:3001"),
        UseStreamableHttp = true
    });
await using var mcpClient = await McpClientFactory.CreateAsync(clientTransport, new McpClientOptions
{
    Capabilities = new()
    {
        Sampling = new()
        {
            SamplingHandler = async (_, _, _) =>
            {
                return new CreateMessageResult
                {
                    Model = "test-model",
                    Role = Role.Assistant,
                    Content = new Content
                    {
                        Type = "text",
                        Text = "Test response"
                    }
                };
            },
        },
    },
});


// Call the server's sampleLLM tool which should trigger our sampling handler
var result = await mcpClient.CallToolAsync(
    "sampleLLM",
    new Dictionary<string, object?>
    {
        ["prompt"] = "Test prompt",
        ["maxTokens"] = 100
    });
Console.WriteLine($"Result: {result}");
Console.ReadKey();

[kaniosm]

Solved, this was a python issue, missing from the documentation.
From those interested the callback signature from the samples is wrong (it's missing the context parameter)

async def handle_sampling_message(ctx,
                                  message: types.CreateMessageRequestParams,
                                  ) -> types.CreateMessageResult:
    return types.CreateMessageResult(
        role="assistant",
        content=types.TextContent(
            type="text",
            text="Hello, world! from model",
        ),
        model="gpt-3.5-turbo",
        stopReason="endTurn",
    )