dramaling-vocab-learning/backend/DramaLing.Api/Controllers/SpeechController.cs

using Microsoft.AspNetCore.Mvc;
using Microsoft.AspNetCore.Authorization;
using DramaLing.Api.Models.DTOs;
using DramaLing.Api.Contracts.Services.Speech;
using DramaLing.Api.Services;

namespace DramaLing.Api.Controllers;

[Route("api/speech")]
[AllowAnonymous] // 暫時開放測試，之後可以加上認證
[ApiExplorerSettings(IgnoreApi = true)] // 暫時從 Swagger 排除，避免 IFormFile 相關問題
public class SpeechController : BaseController
{
    private readonly IPronunciationAssessmentService _assessmentService;

    public SpeechController(
        IPronunciationAssessmentService assessmentService,
        IAuthService authService,
        ILogger<SpeechController> logger) : base(logger, authService)
    {
        _assessmentService = assessmentService;
    }

    /// <summary>
    /// 發音評估 - 上傳音頻檔案並獲得 AI 發音評估結果
    /// </summary>
    /// <param name="audio">音頻檔案 (WAV/WebM/MP3 格式，最大 10MB)</param>
    /// <param name="referenceText">參考文本 - 用戶應該說出的目標句子</param>
    /// <param name="flashcardId">詞卡 ID</param>
    /// <param name="language">語言代碼 (預設: en-US)</param>
    /// <returns>包含準確度、流暢度等多維度評分的評估結果</returns>
    [HttpPost("pronunciation-assessment")]
    [Consumes("multipart/form-data")]
    [ProducesResponseType(typeof(PronunciationResult), 200)]
    [ProducesResponseType(400)]
    [ProducesResponseType(500)]
    [DisableRequestSizeLimit] // 允許大檔案上傳
    public async Task<IActionResult> EvaluatePronunciation(
        [FromForm] IFormFile audio,
        [FromForm] string referenceText,
        [FromForm] string flashcardId,
        [FromForm] string language = "en-US")
    {
        _logger.LogInformation("✅ Controller Action 開始執行 - FlashcardId: {FlashcardId}, ReferenceText: {ReferenceText}",
            flashcardId ?? "NULL", referenceText?.Substring(0, Math.Min(50, referenceText?.Length ?? 0)) ?? "NULL");

        // 檢查 ModelState 是否有效
        if (!ModelState.IsValid)
        {
            _logger.LogWarning("ModelState 驗證失敗:");
            foreach (var modelError in ModelState.Where(m => m.Value.Errors.Count > 0))
            {
                foreach (var error in modelError.Value.Errors)
                {
                    _logger.LogWarning("  {Key}: {Error}", modelError.Key, error.ErrorMessage);
                }
            }
            return ErrorResponse("MODEL_VALIDATION_ERROR", "請求參數驗證失敗", ModelState, 400);
        }

        try
        {
            // 1. 驗證請求
            if (audio == null || audio.Length == 0)
            {
                return ErrorResponse("AUDIO_REQUIRED", "音頻檔案不能為空", null, 400);
            }

            if (audio.Length > 10 * 1024 * 1024) // 10MB 限制
            {
                return ErrorResponse("AUDIO_TOO_LARGE", "音頻檔案過大，請限制在 10MB 以內",
                    new { maxSize = "10MB", actualSize = $"{audio.Length / 1024 / 1024}MB" }, 400);
            }

            if (string.IsNullOrWhiteSpace(referenceText))
            {
                return ErrorResponse("REFERENCE_TEXT_REQUIRED", "參考文本不能為空", null, 400);
            }

            if (string.IsNullOrWhiteSpace(flashcardId))
            {
                return ErrorResponse("FLASHCARD_ID_REQUIRED", "詞卡 ID 不能為空", null, 400);
            }

            // 2. 驗證音頻格式 - 支援更多格式
            var contentType = audio.ContentType?.ToLowerInvariant();
            var allowedTypes = new[] {
                "audio/wav", "audio/webm", "audio/mp3", "audio/mpeg",
                "audio/ogg", "audio/mp4", "audio/x-wav", "audio/wave"
            };

            _logger.LogInformation("接收到音頻檔案: ContentType={ContentType}, Size={Size}bytes, FileName={FileName}",
                contentType, audio.Length, audio.FileName);

            // 如果沒有 Content-Type 或者不在允許列表中，記錄但不立即拒絕
            if (string.IsNullOrEmpty(contentType) || !allowedTypes.Contains(contentType))
            {
                _logger.LogWarning("音頻格式可能不支援: ContentType={ContentType}, 將嘗試處理", contentType);
                // 註解掉嚴格驗證，讓 Azure Speech Services 自己處理
                // return ErrorResponse("INVALID_AUDIO_FORMAT", "不支援的音頻格式",
                //     new { supportedFormats = allowedTypes }, 400);
            }

            // 3. 驗證音頻時長 (簡單檢查檔案大小作為時長估算)
            if (audio.Length < 100) // 降低到 100 bytes，允許短小的測試檔案
            {
                return ErrorResponse("AUDIO_TOO_SHORT", "錄音時間太短或檔案損壞",
                    new {
                        minSize = "100 bytes",
                        actualSize = $"{audio.Length} bytes",
                        fileName = audio.FileName,
                        contentType = contentType
                    }, 400);
            }

            _logger.LogInformation("開始處理發音評估: FlashcardId={FlashcardId}, Size={Size}MB",
                flashcardId, audio.Length / 1024.0 / 1024.0);

            // 4. 處理音頻流並呼叫 Azure Speech Services
            using var audioStream = audio.OpenReadStream();
            var result = await _assessmentService.EvaluatePronunciationAsync(
                audioStream, referenceText, flashcardId, language);

            _logger.LogInformation("發音評估完成: Score={Score}, ProcessingTime={Time}ms",
                result.Scores.Overall, result.ProcessingTime);

            return SuccessResponse(result, "發音評估完成");
        }
        catch (InvalidOperationException ex)
        {
            _logger.LogWarning(ex, "發音評估業務邏輯錯誤: FlashcardId={FlashcardId}", flashcardId);
            return ErrorResponse("SPEECH_PROCESSING_ERROR", ex.Message, null, 400);
        }
        catch (Exception ex)
        {
            _logger.LogError(ex, "發音評估系統錯誤: FlashcardId={FlashcardId}", flashcardId);
            return ErrorResponse("INTERNAL_ERROR", "發音評估失敗，請稍後再試", null, 500);
        }
    }

    /// <summary>
    /// 測試用的簡化發音評估 endpoint - 用於除錯 model binding 問題
    /// </summary>
    [HttpPost("test-pronunciation")]
    [Consumes("multipart/form-data")]
    [ProducesResponseType(200)]
    [ProducesResponseType(400)]
    [DisableRequestSizeLimit]
    public async Task<IActionResult> TestPronunciation()
    {
        _logger.LogInformation("🔧 測試 endpoint 開始執行");

        try
        {
            // 直接使用 Request.Form 避開 model binding
            var form = await Request.ReadFormAsync();

            _logger.LogInformation("📝 Form 讀取成功，包含 {Count} 個欄位", form.Count);

            // 記錄所有 form fields
            foreach (var field in form)
            {
                _logger.LogInformation("  Field: {Key} = {Value}", field.Key, field.Value.ToString());
            }

            // 記錄所有 files
            if (form.Files.Count > 0)
            {
                _logger.LogInformation("📁 找到 {Count} 個檔案", form.Files.Count);
                foreach (var file in form.Files)
                {
                    _logger.LogInformation("  檔案: {Name}, 大小: {Size}bytes, 類型: {Type}",
                        file.Name, file.Length, file.ContentType);
                }
            }
            else
            {
                _logger.LogWarning("⚠️ 沒有找到檔案");
            }

            return SuccessResponse(new
            {
                FormFieldCount = form.Count,
                FileCount = form.Files.Count,
                Fields = form.ToDictionary(f => f.Key, f => f.Value.ToString()),
                Files = form.Files.Select(f => new { f.Name, f.Length, f.ContentType })
            }, "測試成功");

        }
        catch (Exception ex)
        {
            _logger.LogError(ex, "❌ 測試 endpoint 錯誤");
            return ErrorResponse("TEST_ERROR", ex.Message, null, 500);
        }
    }

    /// <summary>
    /// 檢查語音服務狀態
    /// </summary>
    /// <returns>Azure Speech Services 的可用性狀態</returns>
    [HttpGet("service-status")]
    [ProducesResponseType(typeof(object), 200)]
    [ProducesResponseType(500)]
    public async Task<IActionResult> GetServiceStatus()
    {
        try
        {
            var isAvailable = await _assessmentService.IsServiceAvailableAsync();

            var status = new
            {
                IsAvailable = isAvailable,
                ServiceName = "Azure Speech Services",
                CheckTime = DateTime.UtcNow,
                Message = isAvailable ? "服務正常運行" : "服務不可用"
            };

            return SuccessResponse(status);
        }
        catch (Exception ex)
        {
            _logger.LogError(ex, "檢查語音服務狀態時發生錯誤");
            return ErrorResponse("SERVICE_CHECK_ERROR", "無法檢查服務狀態", null, 500);
        }
    }
}