OA2023/OASystem/OASystem.Api/OAMethodLib/HunYuanAPI/HunyuanService.cs

using TencentCloud.Hunyuan.V20230901;
using TencentCloud.Hunyuan.V20230901.Models;

namespace OASystem.API.OAMethodLib.HunYuanAPI
{
    public class HunyuanService : IHunyuanService
    {
        private readonly HunyuanClient _hunyuanClient;

        /// <summary>
        /// 构造函数注入配置好的HunyuanClient
        /// </summary>
        /// <param name="hunyuanClient"></param>
        public HunyuanService(HunyuanClient hunyuanClient)
        {
            _hunyuanClient = hunyuanClient;
        }

        /// <inheritdoc />
        public async Task<string> ChatCompletionsHunyuan_t1_latestAsync(string question)
        {
            var request = new ChatCompletionsRequest
            {
                Model = "hunyuan-t1-latest",
                Messages = new Message[]
                {
                    new Message
                    {
                        Role = "user",
                        Content = question
                    }
                },
                Stream = false,
                Temperature = 0.5f,
                TopP = 1.0f,
                // 其他参数根据需要设置
            };

            // 直接调用SDK方法
            var response = await _hunyuanClient.ChatCompletions(request);

            // 提取并返回模型生成的回答
            // 注意：响应结构可能包含多个Choice，这里取第一个。
            if (response.Choices != null && response.Choices.Length > 0)
            {
                return response.Choices[0].Message.Content?.Trim() ?? "模型未返回内容。";
            }

            return "模型未生成有效回答。";
        }

        public async Task<ChatCompletionsResponse> ChatCompletionsAsync(ChatCompletionsRequest request)
        {
            // 直接调用SDK方法
            return await _hunyuanClient.ChatCompletions(request);
        }

        /// <inheritdoc />
        public async Task<string> AskWithFileContextAsync(string fileContent, string question, string model = "hunyuan-lite")
        {
            // 1. 构建提示词：将文件内容作为上下文，与用户问题结合。
            // 这是一个简单示例，实际可根据需求设计更复杂的Prompt。
            string prompt = $"请根据以下文本内容回答问题。\n文本内容：{fileContent}\n问题：{question}";

            // 2. 使用SDK自带实体构建请求
            var request = new ChatCompletionsRequest
            {
                Model = model,
                Messages = new Message[]
                {
                    new Message
                    {
                        Role = "user",
                        Content = prompt
                    }
                },
                // 可根据需要设置其他参数，如Stream, Temperature, TopP等
                // Stream = false,
                // Temperature = 0.5f,
                // TopP = 1.0f,
            };

            // 3. 调用SDK方法
            var response = await ChatCompletionsAsync(request);

            // 4. 提取并返回模型生成的回答
            // 注意：响应结构可能包含多个Choice，这里取第一个。
            if (response.Choices != null && response.Choices.Length > 0)
            {
                return response.Choices[0].Message.Content?.Trim() ?? "模型未返回内容。";
            }

            return "模型未生成有效回答。";
        }
    }
}