Add ModelContextProtocol.Core package (#428)

src/ModelContextProtocol.Core/Protocol/SamplingCapability.cs

@@ -0,0 +1,43 @@
+using Microsoft.Extensions.AI;
+using ModelContextProtocol.Client;
+using System.Text.Json.Serialization;
+
+namespace ModelContextProtocol.Protocol;
+
+/// <summary>
+/// Represents the capability for a client to generate text or other content using an AI model.
+/// </summary>
+/// <remarks>
+/// <para>
+/// This capability enables the MCP client to respond to sampling requests from an MCP server.
+/// </para>
+/// <para>
+/// When this capability is enabled, an MCP server can request the client to generate content
+/// using an AI model. The client must set a <see cref="SamplingHandler"/> to process these requests.
+/// </para>
+/// </remarks>
+public class SamplingCapability
+{
+    // Currently empty in the spec, but may be extended in the future
+
+    /// <summary>
+    /// Gets or sets the handler for processing <see cref="RequestMethods.SamplingCreateMessage"/> requests.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// This handler function is called when an MCP server requests the client to generate content
+    /// using an AI model. The client must set this property for the sampling capability to work.
+    /// </para>
+    /// <para>
+    /// The handler receives message parameters, a progress reporter for updates, and a 
+    /// cancellation token. It should return a <see cref="CreateMessageResult"/> containing the 
+    /// generated content.
+    /// </para>
+    /// <para>
+    /// You can create a handler using the <see cref="McpClientExtensions.CreateSamplingHandler"/> extension
+    /// method with any implementation of <see cref="IChatClient"/>.
+    /// </para>
+    /// </remarks>
+    [JsonIgnore]
+    public Func<CreateMessageRequestParams?, IProgress<ProgressNotificationValue>, CancellationToken, ValueTask<CreateMessageResult>>? SamplingHandler { get; set; }
+}