Add support for GPT-5 minimal reasoning effort

See announcement at https://cookbook.openai.com/examples/gpt-5/gpt-5_new_params_and_tools

Author: kzu

src/AI.Tests/OpenAITests.cs

@@ -82,6 +82,43 @@ public async Task OpenAIThinks()
         });
     }
 
+    [SecretsFact("OPENAI_API_KEY")]
+    public async Task GPT5_ThinksFast()
+    {
+        var messages = new Chat()
+        {
+            { "system", "You are an intelligent AI assistant that's an expert on financial matters." },
+            { "user", "If you have a debt of 100k and accumulate a compounding 5% debt on top of it every year, how long before you are a negative millonaire? (round up to full integer value)" },
+        };
+
+        var requests = new List<JsonNode>();
+
+        var chat = new OpenAIChatClient(Configuration["OPENAI_API_KEY"]!, "gpt-5-nano",
+            OpenAIClientOptions.Observable(requests.Add).WriteTo(output));
+
+        var options = new ChatOptions
+        {
+            ModelId = "gpt-5",
+            ReasoningEffort = ReasoningEffort.Minimal
+        };
+
+        var response = await chat.GetResponseAsync(messages, options);
+
+        var text = response.Text;
+
+        Assert.Contains("48 years", text);
+        // NOTE: the chat client was requested as grok-3 but the chat options wanted a 
+        // different model and the grok client honors that choice.
+        Assert.StartsWith("gpt-5", response.ModelId);
+        Assert.DoesNotContain("nano", response.ModelId);
+
+        // Reasoning should have been set to medium
+        Assert.All(requests, x =>
+        {
+            var search = Assert.IsType<JsonObject>(x["reasoning"]);
+            Assert.Equal("minimal", search["effort"]?.GetValue<string>());
+        });
+    }
     [SecretsFact("OPENAI_API_KEY")]
     public async Task WebSearchCountryHighContext()
     {

src/AI/OpenAI/OpenAIChatClient.cs

@@ -1,6 +1,7 @@
 using System.ClientModel;
 using System.ClientModel.Primitives;
 using System.Collections.Concurrent;
+using System.Text.Json;
 using Microsoft.Extensions.AI;
 using OpenAI;
 using OpenAI.Responses;
@@ -50,15 +51,18 @@ IChatClient GetChatClient(string modelId) => clients.GetOrAdd(modelId, model
         {
             options.RawRepresentationFactory = _ => new ResponseCreationOptions
             {
-                ReasoningOptions = new ResponseReasoningOptions()
-                {
-                    ReasoningEffortLevel = effort switch
-                    {
-                        ReasoningEffort.High => ResponseReasoningEffortLevel.High,
-                        ReasoningEffort.Medium => ResponseReasoningEffortLevel.Medium,
-                        _ => ResponseReasoningEffortLevel.Low
-                    },
-                }
+                ReasoningOptions = new ReasoningEffortOptions(effort)
+                //ReasoningOptions = new ResponseReasoningOptions()
+                //{
+                //    ReasoningEffortLevel = effort switch
+                //    {
+                //        ReasoningEffort.High => ResponseReasoningEffortLevel.High,
+                //        ReasoningEffort.Medium => ResponseReasoningEffortLevel.Medium,
+                //        // TODO: not exposed yet in the OpenAI package
+                //        // ReasoningEffort.Minimal => ResponseReasoningEffortLevel.Minimal,
+                //        _ => ResponseReasoningEffortLevel.Low
+                //    },
+                //}
             };
         }
 
@@ -71,4 +75,14 @@ IChatClient GetChatClient(string modelId) => clients.GetOrAdd(modelId, model
 
     // Allows creating the base OpenAIClient with a pre-created pipeline.
     class PipelineClient(ClientPipeline pipeline, OpenAIClientOptions? options) : OpenAIClient(pipeline, options) { }
+
+    class ReasoningEffortOptions(ReasoningEffort effort) : ResponseReasoningOptions
+    {
+        protected override void JsonModelWriteCore(Utf8JsonWriter writer, ModelReaderWriterOptions options)
+        {
+            writer.WritePropertyName("effort"u8);
+            writer.WriteStringValue(effort.ToString().ToLowerInvariant());
+            base.JsonModelWriteCore(writer, options);
+        }
+    }
 }

src/AI/ReasoningEffort.cs

@@ -5,6 +5,10 @@
 /// </summary>
 public enum ReasoningEffort
 {
+    /// <summary>
+    /// Minimal reasoning effort, which may result in faster responses. Support depends on the model.
+    /// </summary>
+    Minimal,
     /// <summary>
     /// Low effort reasoning, which may result in faster responses but less detailed or accurate answers.
     /// </summary>