OA2023/OASystem/OASystem.Api/OAMethodLib/DoubaoAPI/DoubaoService.cs

using Aspose.Words;
using OASystem.API.OAMethodLib.File;
using System.Net.Http.Headers;
using System.Text;

namespace OASystem.API.OAMethodLib.DoubaoAPI
{
    public class DoubaoService : IDoubaoService
    {
        private readonly IHttpClientFactory _httpClientFactory;
        private readonly DoubaoSetting _doubaoSetting;
        private readonly ILogger<DoubaoService> _logger;

        public DoubaoService(IHttpClientFactory httpClientFactory, DoubaoSetting doubaoSetting, ILogger<DoubaoService> logger)
        {
            _httpClientFactory = httpClientFactory;
            _doubaoSetting = doubaoSetting;
            _logger = logger;
        }

        public async Task<string> CompleteChatAsync(List<DouBaoChatMessage> messages, CompleteChatOptions? options)
        {

            if (messages == null || !messages.Any())
            {
                _logger.LogError("消息不能为空  " + DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss"));
                return null;
            }

            if (!messages.Any(x => x.Role == DouBaoRole.system))
            {
                messages.Insert(0,
                new DouBaoChatMessage()
                {
                    Role = DouBaoRole.system,
                    Content = "你是一个专业的AI助手，请根据用户的问题给出回答"
                });
            }

            options ??= new CompleteChatOptions();
            var httpClient = _httpClientFactory.CreateClient("Doubao");
            httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _doubaoSetting.ApiKey);
            httpClient.Timeout = TimeSpan.FromMinutes(10);

            var body = new Dictionary<string, object>
            {
                ["model"] = _doubaoSetting.EndpointId,
                ["messages"] = messages.Select(x => new { role = x.Role.ToString(), content = x.Content }).ToArray()
            };

            if (options.ThinkingOptions.IsThinking)
            {
                body["reasoning_effort"] = options.ThinkingOptions.ReasoningEffort.ToString().ToLower();
                body["thinking"] = new
                {
                    type = "enabled",
                };
            }

            var json = JsonConvert.SerializeObject(body);
            var response = await httpClient.PostAsync(
                "chat/completions",
                new StringContent(json, Encoding.UTF8, "application/json")
            );

            response.EnsureSuccessStatusCode();
            var responseContent = await response.Content.ReadAsStringAsync();
            var doubaoResponse = JsonConvert.DeserializeObject<DoubaoResponse>(responseContent);
            return doubaoResponse.choices[0].message.content;
        }

        /// <summary>
        /// 多模态实现
        /// </summary>
        /// <param name="messages"></param>
        /// <param name="options"></param>
        /// <returns></returns>
        public async Task<string> CompleteMultimodalChatAsync(List<DoubaoMultimodalChatMessage> messages, CompleteMultimodalChatOptions? options)
        {
            // 1. 入参校验
            if (messages == null || !messages.Any())
            {
                _logger.LogError("多模态消息不能为空");
                return null;
            }
            if (string.IsNullOrWhiteSpace(_doubaoSetting.ApiKey) || string.IsNullOrWhiteSpace(_doubaoSetting.EndpointId) || string.IsNullOrWhiteSpace(_doubaoSetting.BaseAddress))
            {
                _logger.LogError("Doubao多模态配置不完整（ApiKey/EndpointId/BaseAddress）");
                return null;
            }

            options ??= new CompleteMultimodalChatOptions();
            var httpClient = _httpClientFactory.CreateClient("Doubao");
            httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _doubaoSetting.ApiKey);
            httpClient.Timeout = TimeSpan.FromMinutes(10); // 超时控制

            // 2. 构建请求体（核心：修正type值）
            var requestBody = new
            {
                model = _doubaoSetting.EndpointId,
                input = messages.Select(msg => new
                {
                    role = msg.Role.ToLower(), // 接口要求小写
                    type = "message", // 接口必传的消息类型（固定值）
                    content = msg.Content.Select(contentItem =>
                    {
                        // 自动修正type值，避免传入text/image_url
                        var correctedType = contentItem.Type.ToLower() switch
                        {
                            "text" => "input_text", // 把text修正为input_text
                            "image_url" => "input_image", // 把image_url修正为input_image
                            "file" => "input_file", // 文件类型
                            _ => contentItem.Type // 其他合法值直接使用
                        };

                        // 构建内容项（根据类型补充字段）
                        var contentObj = new Dictionary<string, object>
                        {
                            ["type"] = correctedType
                        };

                        if (correctedType == "input_text" && !string.IsNullOrWhiteSpace(contentItem.Text))
                        {
                            contentObj["text"] = contentItem.Text;
                        }
                        else if (correctedType == "input_image" && contentItem.ImageUrl != null && !string.IsNullOrWhiteSpace(contentItem.ImageUrl.Url))
                        {
                            // Responses API 要求 image_url 为字符串，而不是对象 { url: ... }
                            contentObj["image_url"] = contentItem.ImageUrl.Url;
                        }
                        else if (correctedType == "input_file" && !string.IsNullOrWhiteSpace(contentItem.FileId))
                        {
                            contentObj["file_id"] = contentItem.FileId;
                        }

                        return contentObj;
                    }).ToArray()
                }).ToArray(),
                temperature = options.Temperature,
                thinking = options.ThinkingOptions.IsThinking ? new { type = "enabled" } : null,
                reasoning = options.ThinkingOptions.IsThinking ? new { effort = options.ThinkingOptions.ReasoningEffort.ToLower() } : null,
                max_output_tokens = options.MaxOutputTokens,
                previous_response_id = options.PreviousResponseId,
                expire_at = options.ExpireAt
            };

            try
            {
                // 3. 发送请求
                var json = JsonConvert.SerializeObject(requestBody, new JsonSerializerSettings { NullValueHandling = NullValueHandling.Ignore });
                _logger.LogDebug($"Doubao多模态请求体：{json}");

                var response = await httpClient.PostAsync(
                    "responses",
                    new StringContent(json, Encoding.UTF8, "application/json")
                );

                var responseContent = await response.Content.ReadAsStringAsync();
                _logger.LogDebug($"Doubao多模态响应：{response.StatusCode} {responseContent}");

                if (!response.IsSuccessStatusCode)
                {
                    _logger.LogError($"多模态对话调用失败: {response.StatusCode}, {responseContent}");
                    return null;
                }

                // 4. 解析响应（优先兼容 Responses API 的 output 结构）
                var parsed = JObject.Parse(responseContent);

                // 新结构：output -> type=message -> content -> type=output_text -> text
                var outputText = parsed["output"]?
                    .FirstOrDefault(x => string.Equals(x?["type"]?.ToString(), "message", StringComparison.OrdinalIgnoreCase))?["content"]?
                    .FirstOrDefault(x => string.Equals(x?["type"]?.ToString(), "output_text", StringComparison.OrdinalIgnoreCase))?["text"]?
                    .ToString();

                if (!string.IsNullOrWhiteSpace(outputText))
                {
                    return outputText;
                }

                var doubaoResponse = JsonConvert.DeserializeObject<DoubaoMultimodalResponse>(responseContent);
                var textContent = doubaoResponse?.Choices?.FirstOrDefault()
                    ?.Message?.Content?
                    .FirstOrDefault(c =>
                        string.Equals(c.Type, "input_text", StringComparison.OrdinalIgnoreCase) ||
                        string.Equals(c.Type, "output_text", StringComparison.OrdinalIgnoreCase))?
                    .Text;

                if (string.IsNullOrWhiteSpace(textContent))
                {
                    _logger.LogError("Doubao多模态响应无有效结果");
                    return null;
                }

                return textContent;
            }
            catch (TaskCanceledException ex)
            {
                _logger.LogError(ex, "Doubao多模态接口调用超时");
                return null;
            }
            catch (Exception ex)
            {
                _logger.LogError(ex, "Doubao多模态对话调用异常");
                return null;
            }
        }

        public async Task<DoubaoFileResponse> UploadFileAsync(Stream fileStream, string fileName, string purpose = "user_data")
        {
            var httpClient = _httpClientFactory.CreateClient("Doubao");
            httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _doubaoSetting.ApiKey);

            using var content = new MultipartFormDataContent();

            // 添加 purpose 参数
            content.Add(new StringContent(purpose), "purpose");

            // 添加文件流
            var fileContent = new StreamContent(fileStream);
            fileContent.Headers.ContentType = new MediaTypeHeaderValue("application/octet-stream");
            content.Add(fileContent, "file", fileName);

            var response = await httpClient.PostAsync("files", content);

            if (!response.IsSuccessStatusCode)
            {
                var errorContent = await response.Content.ReadAsStringAsync();
                _logger.LogError($"上传文件失败: {response.StatusCode}, {errorContent}");
                response.EnsureSuccessStatusCode();
            }

            var responseContent = await response.Content.ReadAsStringAsync();
            return JsonConvert.DeserializeObject<DoubaoFileResponse>(responseContent);
        }

        public async Task<DoubaoFileListResponse> ListFilesAsync()
        {
            var httpClient = _httpClientFactory.CreateClient("Doubao");
            httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _doubaoSetting.ApiKey);

            var response = await httpClient.GetAsync("files");
            response.EnsureSuccessStatusCode();

            var responseContent = await response.Content.ReadAsStringAsync();
            return JsonConvert.DeserializeObject<DoubaoFileListResponse>(responseContent);
        }

        public async Task<bool> DeleteFileAsync(string fileId)
        {
            var httpClient = _httpClientFactory.CreateClient("Doubao");
            httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _doubaoSetting.ApiKey);

            var response = await httpClient.DeleteAsync($"files/{fileId}");
            return response.IsSuccessStatusCode;
        }

        /// <summary>
        /// Word(.docx)二进制流转PDF二进制流。
        /// </summary>
        /// <param name="docxStream">Word二进制流（仅支持.docx）</param>
        /// <param name="saveDirectory">PDF保存目录，空则使用程序目录下 temp/pdf</param>
        /// <param name="outputFileName">输出PDF文件名（可不带后缀）</param>
        /// <returns>PDF二进制流（MemoryStream）</returns>
        public static MemoryStream ConvertDocxStreamToPdfStream(
            Stream docxStream,
            string? saveDirectory = null,
            string? outputFileName = null)
        {
            if (docxStream == null || !docxStream.CanRead)
            {
                throw new ArgumentException("Word二进制流不可读或为空", nameof(docxStream));
            }

            if (docxStream.CanSeek)
            {
                docxStream.Position = 0;
            }

            try
            {
                Encoding.RegisterProvider(CodePagesEncodingProvider.Instance);

                AsposeHelper.removewatermark_v2180();

                var wordDoc = new Document(docxStream);
                using var pdfStream = new MemoryStream();
                wordDoc.Save(pdfStream, SaveFormat.Pdf);
                pdfStream.Position = 0;

                var pdfBytes = pdfStream.ToArray();
                // var finalSaveDirectory = string.IsNullOrWhiteSpace(saveDirectory)
                //     ? global::System.IO.Path.Combine(AppContext.BaseDirectory, "temp", "pdf")
                //     : saveDirectory;
                //
                // global::System.IO.Directory.CreateDirectory(finalSaveDirectory);
                //
                // var finalFileName = string.IsNullOrWhiteSpace(outputFileName)
                //     ? $"docx_to_pdf_{DateTime.Now:yyyyMMdd_HHmmss_fff}.pdf"
                //     : (outputFileName.EndsWith(".pdf", StringComparison.OrdinalIgnoreCase) ? outputFileName : $"{outputFileName}.pdf");
                //
                // var savePath = global::System.IO.Path.Combine(finalSaveDirectory, finalFileName);
                // global::System.IO.File.WriteAllBytes(savePath, pdfBytes);

                return new MemoryStream(pdfBytes);
            }
            catch (Exception ex)
            {
                var rootMessage = ex.InnerException?.Message ?? ex.Message;
                throw new InvalidOperationException($"Aspose文档转换失败: {rootMessage}", ex);
            }
        }
    }
}