diff --git a/README.md b/README.md index f02f0f3..bc5a4c7 100644 --- a/README.md +++ b/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 diff --git a/samples/ChatGptApi/ChatGptApi.csproj b/samples/ChatGptApi/ChatGptApi.csproj index 982e462..22e23df 100644 --- a/samples/ChatGptApi/ChatGptApi.csproj +++ b/samples/ChatGptApi/ChatGptApi.csproj @@ -9,8 +9,8 @@ - - + + diff --git a/samples/ChatGptApi/appsettings.json b/samples/ChatGptApi/appsettings.json index 855cd8e..7446548 100644 --- a/samples/ChatGptApi/appsettings.json +++ b/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 diff --git a/samples/ChatGptBlazor.Wasm/ChatGptBlazor.Wasm.csproj b/samples/ChatGptBlazor.Wasm/ChatGptBlazor.Wasm.csproj index c062d86..c99c52a 100644 --- a/samples/ChatGptBlazor.Wasm/ChatGptBlazor.Wasm.csproj +++ b/samples/ChatGptBlazor.Wasm/ChatGptBlazor.Wasm.csproj @@ -8,7 +8,7 @@ - + diff --git a/samples/ChatGptConsole/appsettings.json b/samples/ChatGptConsole/appsettings.json index e7a2f19..bf70f22 100644 --- a/samples/ChatGptConsole/appsettings.json +++ b/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 diff --git a/samples/ChatGptFunctionCallingConsole/appsettings.json b/samples/ChatGptFunctionCallingConsole/appsettings.json index e7a2f19..bf70f22 100644 --- a/samples/ChatGptFunctionCallingConsole/appsettings.json +++ b/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 diff --git a/samples/ChatGptStreamConsole/appsettings.json b/samples/ChatGptStreamConsole/appsettings.json index c58b74d..7988164 100644 --- a/samples/ChatGptStreamConsole/appsettings.json +++ b/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 diff --git a/src/ChatGptNet/ChatGptClient.cs b/src/ChatGptNet/ChatGptClient.cs index c9f73e6..c99ef5e 100644 --- a/src/ChatGptNet/ChatGptClient.cs +++ b/src/ChatGptNet/ChatGptClient.cs @@ -332,7 +332,8 @@ private async Task> CreateMessageListAsync(Guid conversati } private ChatGptRequest CreateChatGptRequest(IEnumerable 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 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 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 messages, EmbeddingParameters? parameters, string? model) => new() { diff --git a/src/ChatGptNet/Models/ChatGptParameters.cs b/src/ChatGptNet/Models/ChatGptParameters.cs index 6d831e1..2ab6d9a 100644 --- a/src/ChatGptNet/Models/ChatGptParameters.cs +++ b/src/ChatGptNet/Models/ChatGptParameters.cs @@ -40,8 +40,21 @@ public class ChatGptParameters /// /// 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. /// + /// + /// This value is now deprecated in favor of , and is not compatible with o1 series models. + /// + /// + [JsonPropertyName("max_tokens")] public int? MaxTokens { get; set; } + /// + /// An upper bound for the number of tokens that can be generated for a completion, including visible output tokens and reasoning tokens. + /// + /// o1 series models must use this property instead of . + /// + [JsonPropertyName("max_completion_tokens")] + public int? MaxCompletionTokens { get; set; } + /// /// 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). /// diff --git a/src/ChatGptNet/Models/ChatGptRequest.cs b/src/ChatGptNet/Models/ChatGptRequest.cs index 1bbc4ff..0483c40 100644 --- a/src/ChatGptNet/Models/ChatGptRequest.cs +++ b/src/ChatGptNet/Models/ChatGptRequest.cs @@ -126,9 +126,21 @@ internal class ChatGptRequest /// /// 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. /// + /// + /// This value is now deprecated in favor of , and is not compatible with o1 series models. + /// + /// [JsonPropertyName("max_tokens")] public int? MaxTokens { get; set; } + /// + /// An upper bound for the number of tokens that can be generated for a completion, including visible output tokens and reasoning tokens. + /// + /// o1 series models must use this property instead of . + /// + [JsonPropertyName("max_completion_tokens")] + public int? MaxCompletionTokens { get; set; } + /// /// 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). /// diff --git a/src/ChatGptNet/Models/OpenAIChatGptModels.cs b/src/ChatGptNet/Models/OpenAIChatGptModels.cs index 0bc2878..94856b7 100644 --- a/src/ChatGptNet/Models/OpenAIChatGptModels.cs +++ b/src/ChatGptNet/Models/OpenAIChatGptModels.cs @@ -69,4 +69,22 @@ public static class OpenAIChatGptModels /// See GPT-4 for more information. /// public const string Gpt4_o_mini = "gpt-4o-mini"; + + /// + /// Reasoning model designed to solve hard problems across domains. + /// + /// + /// This model supports 128.000 tokens and returns a maximum of 32.768 outpout tokens. + /// See o1-preview and o1-mini for more information. + /// + public const string O1_preview = "o1-preview"; + + /// + /// Faster and cheaper reasoning model particularly good at coding, math, and science. + /// + /// + /// This model supports 128.000 tokens and returns a maximum of 32.768 outpout tokens. + /// See o1-preview and o1-mini for more information. + /// + public const string O1_mini = "o1-mini"; } \ No newline at end of file diff --git a/src/Directory.Build.props b/src/Directory.Build.props index 60aa7b1..04b73b8 100644 --- a/src/Directory.Build.props +++ b/src/Directory.Build.props @@ -9,7 +9,7 @@ - +