|
| 1 | +using BotSharp.Abstraction.Templating; |
| 2 | +using System.IO; |
| 3 | + |
| 4 | +namespace BotSharp.Plugin.FileHandler.Functions; |
| 5 | + |
| 6 | +public class EditImageFn : IFunctionCallback |
| 7 | +{ |
| 8 | + public string Name => "edit_image"; |
| 9 | + public string Indication => "Editing image"; |
| 10 | + |
| 11 | + private readonly IServiceProvider _services; |
| 12 | + private readonly ILogger<EditImageFn> _logger; |
| 13 | + private string _conversationId; |
| 14 | + private string _messageId; |
| 15 | + |
| 16 | + public EditImageFn( |
| 17 | + IServiceProvider services, |
| 18 | + ILogger<EditImageFn> logger) |
| 19 | + { |
| 20 | + _services = services; |
| 21 | + _logger = logger; |
| 22 | + } |
| 23 | + |
| 24 | + public async Task<bool> Execute(RoleDialogModel message) |
| 25 | + { |
| 26 | + var args = JsonSerializer.Deserialize<LlmContextIn>(message.FunctionArgs); |
| 27 | + var descrpition = args?.UserRequest ?? string.Empty; |
| 28 | + Init(message); |
| 29 | + SetImageOptions(); |
| 30 | + |
| 31 | + var image = await SelectConversationImage(); |
| 32 | + var response = await GetImageEditGeneration(message, descrpition, image); |
| 33 | + message.Content = response; |
| 34 | + return true; |
| 35 | + } |
| 36 | + |
| 37 | + private void Init(RoleDialogModel message) |
| 38 | + { |
| 39 | + var convService = _services.GetRequiredService<IConversationService>(); |
| 40 | + _conversationId = convService.ConversationId; |
| 41 | + _messageId = message.MessageId; |
| 42 | + } |
| 43 | + |
| 44 | + private void SetImageOptions() |
| 45 | + { |
| 46 | + var state = _services.GetRequiredService<IConversationStateService>(); |
| 47 | + state.SetState("image_format", "bytes"); |
| 48 | + state.SetState("image_count", "1"); |
| 49 | + } |
| 50 | + |
| 51 | + private async Task<MessageFileModel?> SelectConversationImage() |
| 52 | + { |
| 53 | + var convService = _services.GetRequiredService<IConversationService>(); |
| 54 | + var fileService = _services.GetRequiredService<IBotSharpFileService>(); |
| 55 | + var dialogs = convService.GetDialogHistory(fromBreakpoint: false); |
| 56 | + var messageIds = dialogs.Select(x => x.MessageId).Distinct().ToList(); |
| 57 | + var images = fileService.GetMessageFiles(_conversationId, messageIds, FileSourceType.User, imageOnly: true); |
| 58 | + return await SelectImage(images, dialogs); |
| 59 | + } |
| 60 | + |
| 61 | + private async Task<MessageFileModel?> SelectImage(IEnumerable<MessageFileModel> images, List<RoleDialogModel> dialogs) |
| 62 | + { |
| 63 | + if (images.IsNullOrEmpty()) return null; |
| 64 | + |
| 65 | + var llmProviderService = _services.GetRequiredService<ILlmProviderService>(); |
| 66 | + var render = _services.GetRequiredService<ITemplateRender>(); |
| 67 | + var db = _services.GetRequiredService<IBotSharpRepository>(); |
| 68 | + |
| 69 | + try |
| 70 | + { |
| 71 | + var promptImages = images.Where(x => x.ContentType == MediaTypeNames.Image.Png).Select((x, idx) => |
| 72 | + { |
| 73 | + return $"id: {idx + 1}, image_name: {x.FileName}.{x.FileType}"; |
| 74 | + }).ToList(); |
| 75 | + |
| 76 | + if (promptImages.IsNullOrEmpty()) return null; |
| 77 | + |
| 78 | + var prompt = db.GetAgentTemplate(BuiltInAgentId.UtilityAssistant, "select_edit_image_prompt"); |
| 79 | + prompt = render.Render(prompt, new Dictionary<string, object> |
| 80 | + { |
| 81 | + { "image_list", promptImages } |
| 82 | + }); |
| 83 | + |
| 84 | + var agent = new Agent |
| 85 | + { |
| 86 | + Id = BuiltInAgentId.UtilityAssistant, |
| 87 | + Name = "Utility Assistant", |
| 88 | + Instruction = prompt |
| 89 | + }; |
| 90 | + |
| 91 | + var provider = llmProviderService.GetProviders().FirstOrDefault(x => x == "openai"); |
| 92 | + var model = llmProviderService.GetProviderModel(provider: provider, id: "gpt-4"); |
| 93 | + var completion = CompletionProvider.GetChatCompletion(_services, provider: provider, model: model.Name); |
| 94 | + var response = await completion.GetChatCompletions(agent, dialogs); |
| 95 | + var content = response?.Content ?? string.Empty; |
| 96 | + var fid = JsonSerializer.Deserialize<int?>(content); |
| 97 | + return images.Where((x, idx) => idx == fid - 1).FirstOrDefault(); |
| 98 | + } |
| 99 | + catch (Exception ex) |
| 100 | + { |
| 101 | + _logger.LogWarning($"Error when getting the image edit response. {ex.Message}\r\n{ex.InnerException}"); |
| 102 | + return null; |
| 103 | + } |
| 104 | + } |
| 105 | + |
| 106 | + private async Task<string> GetImageEditGeneration(RoleDialogModel message, string description, MessageFileModel? image) |
| 107 | + { |
| 108 | + if (image == null) |
| 109 | + { |
| 110 | + return "Failed to find an image. Please provide an image."; |
| 111 | + } |
| 112 | + |
| 113 | + try |
| 114 | + { |
| 115 | + var completion = CompletionProvider.GetImageCompletion(_services, provider: "openai", model: "dall-e-2"); |
| 116 | + var text = !string.IsNullOrWhiteSpace(description) ? description : message.Content; |
| 117 | + var dialog = RoleDialogModel.From(message, AgentRole.User, text); |
| 118 | + var agent = new Agent |
| 119 | + { |
| 120 | + Id = BuiltInAgentId.UtilityAssistant, |
| 121 | + Name = "Utility Assistant" |
| 122 | + }; |
| 123 | + |
| 124 | + using var stream = File.OpenRead(image.FileStorageUrl); |
| 125 | + var result = await completion.GetImageEdits(agent, dialog, stream, image.FileName ?? string.Empty); |
| 126 | + stream.Close(); |
| 127 | + SaveGeneratedImage(result?.GeneratedImages?.FirstOrDefault()); |
| 128 | + |
| 129 | + return !string.IsNullOrWhiteSpace(result?.Content) ? result.Content : "Image edit is completed."; |
| 130 | + } |
| 131 | + catch (Exception ex) |
| 132 | + { |
| 133 | + var error = $"Error when getting image edit response. {ex.Message}"; |
| 134 | + _logger.LogWarning($"{error}\r\n{ex.InnerException}"); |
| 135 | + return error; |
| 136 | + } |
| 137 | + } |
| 138 | + |
| 139 | + private void SaveGeneratedImage(ImageGeneration? image) |
| 140 | + { |
| 141 | + if (image == null) return; |
| 142 | + |
| 143 | + var files = new List<BotSharpFile>() |
| 144 | + { |
| 145 | + new BotSharpFile |
| 146 | + { |
| 147 | + FileName = $"{Guid.NewGuid()}.png", |
| 148 | + FileData = $"data:{MediaTypeNames.Image.Png};base64,{image.ImageData}" |
| 149 | + } |
| 150 | + }; |
| 151 | + |
| 152 | + var fileService = _services.GetRequiredService<IBotSharpFileService>(); |
| 153 | + fileService.SaveMessageFiles(_conversationId, _messageId, FileSourceType.Bot, files); |
| 154 | + } |
| 155 | +} |
0 commit comments