Add support for max_completion_tokens property (#168)

Author: marcominerva

README.md

@@ -36,6 +36,7 @@ builder.Services.AddChatGpt(options =>
     options.DefaultParameters = new ChatGptParameters
     {
         MaxTokens = 800,
+        //MaxCompletionTokens = 800,  // o1 series models support this property instead of MaxTokens
         Temperature = 0.7
     };
 });
@@ -85,6 +86,8 @@ Currently available models are:
 - gpt-4-turbo
 - gpt-4o
 - gpt-4o-mini
+- o1-preview
+- o1-mini
 
 They have fixed names, available in the [OpenAIChatGptModels.cs file](https://github.com/marcominerva/ChatGptNet/blob/master/src/ChatGptNet/Models/OpenAIChatGptModels.cs).
 
@@ -163,6 +166,7 @@ The configuration can be automatically read from [IConfiguration](https://learn.
     //    "Temperature": 0.8,
     //    "TopP": 1,
     //    "MaxTokens": 500,
+    //    "MaxCompletionTokens": null,  // o1 series models support this property instead of MaxTokens
     //    "PresencePenalty": 0,
     //    "FrequencyPenalty": 0,
     //    "ResponseFormat": { "Type": "text" }, // Allowed values for Type: text (default) or json_object

samples/ChatGptApi/ChatGptApi.csproj

@@ -9,8 +9,8 @@
     <ItemGroup>
         <PackageReference Include="Microsoft.AspNetCore.OpenApi" Version="8.0.10" />
         <PackageReference Include="Microsoft.Extensions.Http.Resilience" Version="8.10.0" />
-        <PackageReference Include="Swashbuckle.AspNetCore" Version="6.8.1" />
-        <PackageReference Include="TinyHelpers.AspNetCore" Version="3.1.17" />
+        <PackageReference Include="Swashbuckle.AspNetCore" Version="6.9.0" />
+        <PackageReference Include="TinyHelpers.AspNetCore" Version="3.1.19" />
     </ItemGroup>
 
     <ItemGroup>

samples/ChatGptApi/appsettings.json

@@ -17,6 +17,7 @@
         //    "Temperature": 0.8,
         //    "TopP": 1,
         //    "MaxTokens": 500,
+        //    "MaxCompletionTokens": null,  // o1 series models supports this property instead of MaxTokens
         //    "PresencePenalty": 0,
         //    "FrequencyPenalty": 0,
         //    "ResponseFormat": { "Type": "text" }, // Allowed values for Type: text (default) or json_object

samples/ChatGptBlazor.Wasm/ChatGptBlazor.Wasm.csproj

@@ -8,7 +8,7 @@
     </PropertyGroup>
 
     <ItemGroup>
-        <PackageReference Include="Markdig" Version="0.37.0" />
+        <PackageReference Include="Markdig" Version="0.38.0" />
         <PackageReference Include="Microsoft.AspNetCore.Components.WebAssembly" Version="8.0.10" />
         <PackageReference Include="Microsoft.AspNetCore.Components.WebAssembly.DevServer" Version="8.0.10" PrivateAssets="all" />
         <PackageReference Include="Microsoft.Extensions.Http.Resilience" Version="8.10.0" />

samples/ChatGptConsole/appsettings.json

@@ -17,6 +17,7 @@
         //    "Temperature": 0.8,
         //    "TopP": 1,
         //    "MaxTokens": 500,
+        //    "MaxCompletionTokens": null,  // o1 series models support this property instead of MaxTokens
         //    "PresencePenalty": 0,
         //    "FrequencyPenalty": 0,
         //    "ResponseFormat": { "Type": "text" }, // Allowed values for Type: text (default) or json_object

samples/ChatGptFunctionCallingConsole/appsettings.json

@@ -17,6 +17,7 @@
         //    "Temperature": 0.8,
         //    "TopP": 1,
         //    "MaxTokens": 500,
+        //    "MaxCompletionTokens": null,  // o1 series models support this property instead of MaxTokens
         //    "PresencePenalty": 0,
         //    "FrequencyPenalty": 0,
         //    "ResponseFormat": { "Type": "text" }, // Allowed values for Type: text (default) or json_object

samples/ChatGptStreamConsole/appsettings.json

@@ -17,6 +17,7 @@
         //    "Temperature": 0.8,
         //    "TopP": 1,
         //    "MaxTokens": 500,
+        //    "MaxCompletionTokens": null,  // o1 series models support this property instead of MaxTokens
         //    "PresencePenalty": 0,
         //    "FrequencyPenalty": 0,
         //    "ResponseFormat": { "Type": "text" }, // Allowed values for Type: text (default) or json_object

src/ChatGptNet/ChatGptClient.cs

@@ -332,7 +332,8 @@ private async Task<IList<ChatGptMessage>> CreateMessageListAsync(Guid conversati
     }
 
     private ChatGptRequest CreateChatGptRequest(IEnumerable<ChatGptMessage> messages, ChatGptToolParameters? toolParameters, bool stream, ChatGptParameters? parameters, string? model)
-        => new()
+    {
+        var request = new ChatGptRequest()
         {
             Model = model ?? options.DefaultModel,
             Messages = messages,
@@ -362,6 +363,7 @@ private ChatGptRequest CreateChatGptRequest(IEnumerable<ChatGptMessage> messages
             Temperature = parameters?.Temperature ?? options.DefaultParameters.Temperature,
             TopP = parameters?.TopP ?? options.DefaultParameters.TopP,
             MaxTokens = parameters?.MaxTokens ?? options.DefaultParameters.MaxTokens,
+            MaxCompletionTokens = parameters?.MaxCompletionTokens ?? options.DefaultParameters.MaxCompletionTokens,
             PresencePenalty = parameters?.PresencePenalty ?? options.DefaultParameters.PresencePenalty,
             FrequencyPenalty = parameters?.FrequencyPenalty ?? options.DefaultParameters.FrequencyPenalty,
             ResponseFormat = parameters?.ResponseFormat ?? options.DefaultParameters.ResponseFormat,
@@ -370,6 +372,19 @@ private ChatGptRequest CreateChatGptRequest(IEnumerable<ChatGptMessage> messages
             User = options.User
         };
 
+        /*
+         * As of 2024-09-01-preview, Azure OpenAI conditionally supports the use of the new max_completion_tokens property:
+         *   - The o1-mini and o1-preview models accept max_completion_tokens and reject max_tokens
+         *   - All other models reject max_completion_tokens and accept max_tokens
+         */
+        if (request.MaxCompletionTokens is not null)
+        {
+            request.MaxTokens = null;
+        }
+
+        return request;
+    }
+
     private EmbeddingRequest CreateEmbeddingRequest(IEnumerable<string> messages, EmbeddingParameters? parameters, string? model)
         => new()
         {

src/ChatGptNet/Models/ChatGptParameters.cs

@@ -40,8 +40,21 @@ public class ChatGptParameters
     /// <summary>
     /// Gets or sets the maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length.
     /// </summary>
+    /// <remarks>
+    /// This value is now deprecated in favor of <see cref="MaxCompletionTokens"/>, and is not compatible with <see href="https://platform.openai.com/docs/guides/reasoning">o1 series models</see>.
+    /// </remarks>
+    /// <seealso cref="MaxCompletionTokens"/>
+    [JsonPropertyName("max_tokens")]
     public int? MaxTokens { get; set; }
 
+    /// <summary>
+    /// An upper bound for the number of tokens that can be generated for a completion, including visible output tokens and <see href="https://platform.openai.com/docs/guides/reasoning">reasoning tokens</see>.
+    /// </summary>
+    /// <remarks>o1 series models must use this property instead of <see cref="MaxTokens"/>.</remarks>
+    /// <seealso cref="MaxTokens"/>
+    [JsonPropertyName("max_completion_tokens")]
+    public int? MaxCompletionTokens { get; set; }
+
     /// <summary>
     /// Gets or sets the presence penalties for chat completion. Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics (default: 0).
     /// </summary>

src/ChatGptNet/Models/ChatGptRequest.cs

@@ -126,9 +126,21 @@ internal class ChatGptRequest
     /// <summary>
     /// Gets or sets the maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length.
     /// </summary>
+    /// <remarks>
+    /// This value is now deprecated in favor of <see cref="MaxCompletionTokens"/>, and is not compatible with <see href="https://platform.openai.com/docs/guides/reasoning">o1 series models</see>.
+    /// </remarks>
+    /// <seealso cref="MaxCompletionTokens"/>
     [JsonPropertyName("max_tokens")]
     public int? MaxTokens { get; set; }
 
+    /// <summary>
+    /// An upper bound for the number of tokens that can be generated for a completion, including visible output tokens and <see href="https://platform.openai.com/docs/guides/reasoning">reasoning tokens</see>.
+    /// </summary>
+    /// <remarks>o1 series models must use this property instead of <see cref="MaxTokens"/>.</remarks>
+    /// <seealso cref="MaxTokens"/>
+    [JsonPropertyName("max_completion_tokens")]
+    public int? MaxCompletionTokens { get; set; }
+
     /// <summary>
     /// Gets or sets the presence penalties for chat completion. A number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics (default: 0).
     /// </summary>