feat(background): add visual recognition task and refactor picture processing

This commit is contained in:
ShiYu
2025-06-08 15:40:08 +08:00
parent 7ad8b6c826
commit 39c40d2746
17 changed files with 473 additions and 269 deletions

View File

@@ -0,0 +1,218 @@
using Foxel.Models.DataBase;
using Foxel.Services.AI;
using Foxel.Services.Storage;
using Foxel.Services.VectorDB;
using Foxel.Utils;
using Microsoft.EntityFrameworkCore;
using System.Text.Json;
using Foxel.Services.Attributes;
namespace Foxel.Services.Background.Processors
{
public class VisualRecognitionTaskProcessor : ITaskProcessor
{
private readonly IDbContextFactory<MyDbContext> _contextFactory;
private readonly IServiceProvider _serviceProvider;
private readonly ILogger<VisualRecognitionTaskProcessor> _logger;
private readonly IWebHostEnvironment _environment;
public VisualRecognitionTaskProcessor(
IDbContextFactory<MyDbContext> contextFactory,
IServiceProvider serviceProvider,
ILogger<VisualRecognitionTaskProcessor> logger,
IWebHostEnvironment environment)
{
_contextFactory = contextFactory;
_serviceProvider = serviceProvider;
_logger = logger;
_environment = environment;
}
public async Task ProcessAsync(BackgroundTask backgroundTask)
{
if (backgroundTask.Payload == null)
{
await UpdateTaskStatusInDb(backgroundTask.Id, TaskExecutionStatus.Failed, 0, "任务 Payload 为空。");
_logger.LogError("视觉识别任务 Payload 为空: TaskId={TaskId}", backgroundTask.Id);
return;
}
VisualRecognitionPayload? payload;
try
{
payload = JsonSerializer.Deserialize<VisualRecognitionPayload>(backgroundTask.Payload);
}
catch (JsonException ex)
{
_logger.LogError(ex, "无法解析视觉识别任务的 Payload: TaskId={TaskId}", backgroundTask.Id);
await UpdateTaskStatusInDb(backgroundTask.Id, TaskExecutionStatus.Failed, 0, "Payload 解析失败。");
return;
}
if (payload == null || payload.PictureId == 0)
{
_logger.LogError("视觉识别任务的 Payload 无效或缺少 PictureId: TaskId={TaskId}", backgroundTask.Id);
await UpdateTaskStatusInDb(backgroundTask.Id, TaskExecutionStatus.Failed, 0, "Payload 无效或缺少 PictureId。");
return;
}
var pictureId = payload.PictureId;
string thumbnailForAIDownloadPath = string.Empty; // Path if thumbnail needs to be downloaded
bool isTempThumbnailFile = false;
await using var dbContext = await _contextFactory.CreateDbContextAsync();
var currentBackgroundTaskState = await dbContext.BackgroundTasks.FindAsync(backgroundTask.Id);
if (currentBackgroundTaskState == null)
{
_logger.LogError("在 VisualRecognitionTaskProcessor 中找不到后台任务: TaskId={TaskId}", backgroundTask.Id);
return;
}
var picture = await dbContext.Pictures.Include(p => p.User).ThenInclude(u => u.Tags).FirstOrDefaultAsync(p => p.Id == pictureId);
try
{
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 10, currentBackgroundTaskState: currentBackgroundTaskState);
if (picture == null)
{
throw new Exception($"找不到ID为{pictureId}的图片。");
}
if (string.IsNullOrEmpty(picture.ThumbnailPath))
{
throw new Exception($"图片ID {pictureId} 的缩略图路径为空无法进行AI分析。");
}
using var scope = _serviceProvider.CreateScope();
var aiService = scope.ServiceProvider.GetRequiredService<IAiService>();
var storageService = scope.ServiceProvider.GetRequiredService<IStorageService>();
string contentRootPath = _environment.ContentRootPath;
string actualThumbnailPathForAI;
if (picture.StorageType == StorageType.Local)
{
actualThumbnailPathForAI = Path.Combine(contentRootPath, picture.ThumbnailPath.TrimStart('/'));
}
else // Remote storage
{
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 15, currentBackgroundTaskState: currentBackgroundTaskState);
thumbnailForAIDownloadPath = await storageService.ExecuteAsync(picture.StorageType,
provider => provider.DownloadFileAsync(picture.ThumbnailPath));
actualThumbnailPathForAI = thumbnailForAIDownloadPath;
isTempThumbnailFile = true;
}
if (string.IsNullOrEmpty(actualThumbnailPathForAI) || !File.Exists(actualThumbnailPathForAI))
{
throw new Exception($"找不到用于AI分析的缩略图文件: {actualThumbnailPathForAI} (源路径: {picture.ThumbnailPath})");
}
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 20, currentBackgroundTaskState: currentBackgroundTaskState);
string base64Image = await ImageHelper.ConvertImageToBase64(actualThumbnailPathForAI);
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 40, currentBackgroundTaskState: currentBackgroundTaskState);
var (title, description) = await aiService.AnalyzeImageAsync(base64Image);
string finalTitle = !string.IsNullOrWhiteSpace(title) && title != "AI生成的标题" ? title : Path.GetFileNameWithoutExtension(picture.Name);
string finalDescription = !string.IsNullOrWhiteSpace(description) && description != "AI生成的描述" ? description : picture.Description;
picture.Name = finalTitle; // Potentially overwrites name set from filename
picture.Description = finalDescription;
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 60, currentBackgroundTaskState: currentBackgroundTaskState);
var combinedText = $"{finalTitle}. {finalDescription}";
var embedding = await aiService.GetEmbeddingAsync(combinedText);
picture.Embedding = embedding;
if (picture.UserId.HasValue && embedding != null && embedding.Length > 0)
{
var vectorDbService = scope.ServiceProvider.GetRequiredService<IVectorDbService>();
await vectorDbService.AddPictureToUserCollectionAsync(picture.UserId.Value, new Models.Vector.PictureVector
{
Id = (ulong)picture.Id,
Name = picture.Name,
Embedding = embedding
});
}
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 70, currentBackgroundTaskState: currentBackgroundTaskState);
var availableTagNames = await dbContext.Tags.Select(t => t.Name).ToListAsync();
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 80, currentBackgroundTaskState: currentBackgroundTaskState);
var matchedTagNames = await aiService.GenerateTagsFromImageAsync(base64Image, availableTagNames, true);
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Processing, 90, currentBackgroundTaskState: currentBackgroundTaskState);
if (picture.User != null && matchedTagNames.Any())
{
picture.Tags ??= new List<Tag>();
foreach (var tagName in matchedTagNames)
{
var existingTag = await dbContext.Tags.FirstOrDefaultAsync(t => t.Name.ToLower() == tagName.ToLower());
if (existingTag == null)
{
existingTag = new Tag { Name = tagName.Trim(), Description = tagName.Trim() };
dbContext.Tags.Add(existingTag);
}
if (!picture.Tags.Any(t => t.Id == existingTag.Id)) picture.Tags.Add(existingTag);
picture.User.Tags ??= new List<Tag>();
if (!picture.User.Tags.Any(t => t.Id == existingTag.Id)) picture.User.Tags.Add(existingTag);
}
}
await dbContext.SaveChangesAsync(); // Save all AI-related changes to Picture
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Completed, 100, completedAt: DateTime.UtcNow, currentBackgroundTaskState: currentBackgroundTaskState);
}
catch (Exception ex)
{
_logger.LogError(ex, "视觉识别任务失败: TaskId={TaskId}, PictureId={PictureId}", currentBackgroundTaskState.Id, pictureId);
await UpdateTaskStatusInDb(currentBackgroundTaskState.Id, TaskExecutionStatus.Failed, currentBackgroundTaskState.Progress, ex.Message, currentBackgroundTaskState: currentBackgroundTaskState);
// dbContext.SaveChangesAsync() might be called in UpdateTaskStatusInDb or here if picture state needs saving on error
}
finally
{
if (isTempThumbnailFile && File.Exists(thumbnailForAIDownloadPath))
{
try { File.Delete(thumbnailForAIDownloadPath); } catch (Exception ex) { _logger.LogWarning(ex, "删除临时AI缩略图文件失败: {FilePath}", thumbnailForAIDownloadPath); }
}
}
}
private async Task UpdateTaskStatusInDb(Guid taskId, TaskExecutionStatus status, int progress, string? error = null, DateTime? startedAt = null, DateTime? completedAt = null, BackgroundTask? currentBackgroundTaskState = null)
{
await using var dbContext = await _contextFactory.CreateDbContextAsync();
var taskToUpdate = currentBackgroundTaskState ?? await dbContext.BackgroundTasks.FindAsync(taskId);
if (taskToUpdate != null)
{
if (currentBackgroundTaskState != null && dbContext.Entry(currentBackgroundTaskState).State == EntityState.Detached)
{
dbContext.BackgroundTasks.Attach(currentBackgroundTaskState);
}
taskToUpdate.Status = status;
taskToUpdate.Progress = progress;
taskToUpdate.ErrorMessage = string.IsNullOrEmpty(error) ? taskToUpdate.ErrorMessage : error; // Keep existing error if new one is null/empty
if (startedAt.HasValue) taskToUpdate.StartedAt = startedAt;
if (completedAt.HasValue) taskToUpdate.CompletedAt = completedAt;
if ((status == TaskExecutionStatus.Completed || status == TaskExecutionStatus.Failed) && !taskToUpdate.StartedAt.HasValue)
{
taskToUpdate.StartedAt = taskToUpdate.CreatedAt; // Ensure StartedAt is set
}
if (status == TaskExecutionStatus.Completed || status == TaskExecutionStatus.Failed)
{
taskToUpdate.CompletedAt ??= DateTime.UtcNow; // Ensure CompletedAt is set
}
await dbContext.SaveChangesAsync();
_logger.LogInformation("任务状态更新 (VisualRecognitionProcessor): TaskId={TaskId}, Status={Status}, Progress={Progress}%", taskId, status, progress);
}
else
{
_logger.LogWarning("尝试在 VisualRecognitionProcessor 中更新不存在的任务状态: TaskId={TaskId}", taskId);
}
}
}
}