| 1 | import { BadRequestException, Injectable, Logger, NotFoundException, Optional } from '@nestjs/common' |
| 2 | import { QueueService } from '@yikart/aitoearn-queue' |
| 3 | import { AssetsService } from '@yikart/assets' |
| 4 | import { AppException, getErrorMessage, getExtByMimeType, ImageType, ResponseCode, UserType } from '@yikart/common' |
| 5 | import { AiLogChannel, AiLogImageResult, AiLogRepository, AiLogStatus, AiLogType, AssetType, ImageAiLogResponse, Transactional } from '@yikart/mongodb' |
| 6 | import parseDataUri from 'data-urls' |
| 7 | import OpenAI from 'openai' |
| 8 | |
| 9 | import { runWithAiGenerationRetry } from '../ai-generation-retry.util' |
| 10 | import { GeminiService } from '../libs/gemini/gemini.service' |
| 11 | import { OpenaiService } from '../libs/openai' |
| 12 | import { ModelsConfigService } from '../models-config' |
| 13 | import { RelayMediaResolverService } from '../relay-media' |
| 14 | import { |
| 15 | GeminiImageGenerationDto, |
| 16 | ImageEditDto, |
| 17 | ImageEditModelsQueryDto, |
| 18 | ImageGenerationDto, |
| 19 | ImageGenerationModelsQueryDto, |
| 20 | UserGeminiImageGenerationDto, |
| 21 | UserImageEditDto, |
| 22 | UserImageGenerationDto, |
| 23 | } from './image.dto' |
| 24 | |
| 25 | type Uploadable = File | Response |
| 26 | |
| 27 | @Injectable() |
| 28 | export class ImageService { |
| 29 | private readonly logger = new Logger(ImageService.name) |
| 30 | |
| 31 | constructor( |
| 32 | private readonly assetsService: AssetsService, |
| 33 | private readonly openaiService: OpenaiService, |
| 34 | private readonly geminiService: GeminiService, |
| 35 | private readonly aiLogRepo: AiLogRepository, |
| 36 | private readonly modelsConfigService: ModelsConfigService, |
| 37 | private readonly queueService: QueueService, |
| 38 | @Optional() private readonly relayMediaResolver?: RelayMediaResolverService, |
| 39 | ) { } |
| 40 | |
| 41 | private resolveRuntimeImageModel(model: string, kind: 'generation' | 'edit'): string { |
| 42 | const modelConfig = kind === 'generation' |
| 43 | ? this.modelsConfigService.config.image.generation.find(item => item.name === model) |
| 44 | : this.modelsConfigService.config.image.edit.find(item => item.name === model) |
| 45 | |
| 46 | return modelConfig?.runtimeModel ?? model |
| 47 | } |
| 48 | |
| 49 | private getImageModelRetry(model: string, kind: 'generation' | 'edit'): number { |
| 50 | const modelConfig = kind === 'generation' |
| 51 | ? this.modelsConfigService.config.image.generation.find(item => item.name === model) |
| 52 | : this.modelsConfigService.config.image.edit.find(item => item.name === model) |
| 53 | |
| 54 | return modelConfig?.retry ?? 0 |
| 55 | } |
| 56 | |
| 57 | /** |
| 58 | * 将 data uri 转换为 Uploadable |
| 59 | */ |
| 60 | private getUploadableByDataUri(dataUri: string, filename = 'image'): Uploadable { |
| 61 | const file = parseDataUri(dataUri) |
| 62 | if (file == null) { |
| 63 | throw new BadRequestException('Invalid data URI') |
| 64 | } |
| 65 | const ext = getExtByMimeType(file.mimeType.essence as ImageType) |
| 66 | |
| 67 | return new File([file.body as Uint8Array<ArrayBuffer>], `${filename}.${ext}`, { type: file.mimeType.essence }) |
| 68 | } |
| 69 | |
| 70 | /** |
| 71 | * 将 URL 转换为 Uploadable |
| 72 | */ |
| 73 | private async getUploadableByUrl(url: string): Promise<Uploadable> { |
| 74 | return await fetch(url) |
| 75 | } |
| 76 | |
| 77 | /** |
| 78 | * 将 URL 或 Data URI 转换为 Uploadable |
| 79 | */ |
| 80 | private async getUploadableByUrlOrDataUri(urlOrDataUri: string, filename = 'image'): Promise<Uploadable> { |
| 81 | if (/^https?:\/\//.test(urlOrDataUri)) { |
| 82 | return await this.getUploadableByUrl(urlOrDataUri) |
| 83 | } |
| 84 | return this.getUploadableByDataUri(urlOrDataUri, filename) |
| 85 | } |
| 86 | |
| 87 | private async resolveRelayText(text: string): Promise<string> { |
| 88 | if (!this.relayMediaResolver) { |
| 89 | return text |
| 90 | } |
| 91 | return await this.relayMediaResolver.resolveText(text) |
| 92 | } |
| 93 | |
| 94 | /** |
| 95 | * 上传图片到S3并返回路径 |
| 96 | */ |
| 97 | private async uploadImageToS3(imageUrlOrResponse: string | Response, userId: string, subPath?: string): Promise<string> { |
| 98 | if (typeof imageUrlOrResponse === 'string') { |
| 99 | const result = await this.assetsService.uploadFromUrl(userId, { |
| 100 | url: imageUrlOrResponse, |
| 101 | type: AssetType.AiImage, |
| 102 | }, subPath) |
| 103 | return result.asset.path |
| 104 | } |
| 105 | else { |
| 106 | const contentType = imageUrlOrResponse.headers.get('content-type') || 'image/png' |
| 107 | const buffer = Buffer.from(await imageUrlOrResponse.arrayBuffer()) |
| 108 | const result = await this.assetsService.uploadFromBuffer(userId, buffer, { |
| 109 | type: AssetType.AiImage, |
| 110 | mimeType: contentType, |
| 111 | }, subPath) |
| 112 | return result.asset.path |
| 113 | } |
| 114 | } |
| 115 | |
| 116 | /** |
| 117 | * 图片生成 |
| 118 | */ |
| 119 | async generation(request: ImageGenerationDto) { |
| 120 | const { user, ...params } = request |
| 121 | const runtimeModel = this.resolveRuntimeImageModel(params.model, 'generation') |
| 122 | |
| 123 | if (!user) { |
| 124 | throw new BadRequestException('userId is required') |
| 125 | } |
| 126 | |
| 127 | if (runtimeModel === 'gpt-image-1') { |
| 128 | delete params.response_format |
| 129 | delete params.style |
| 130 | } |
| 131 | |
| 132 | const result = await this.openaiService.createImageGeneration({ |
| 133 | ...params, |
| 134 | model: runtimeModel, |
| 135 | } as Omit<OpenAI.Images.ImageGenerateParams, 'user'> & { apiKey?: string }) |
| 136 | |
| 137 | for (const image of result.data || []) { |
| 138 | if (image.url) { |
| 139 | image.url = await this.uploadImageToS3(image.url, user, `${request.model}`) |
| 140 | } |
| 141 | if (image.b64_json) { |
| 142 | const mimeType = `image/${result.output_format || 'png'}` |
| 143 | const buffer = Buffer.from(image.b64_json, 'base64') |
| 144 | const uploadResult = await this.assetsService.uploadFromBuffer(user, buffer, { |
| 145 | type: AssetType.AiImage, |
| 146 | mimeType, |
| 147 | }, `${request.model}`) |
| 148 | image.url = uploadResult.asset.path |
| 149 | delete image.b64_json |
| 150 | } |
| 151 | } |
| 152 | |
| 153 | return { |
| 154 | ...result, |
| 155 | list: result.data || [], |
| 156 | } |
| 157 | } |
| 158 | |
| 159 | /** |
| 160 | * 图片编辑 |
| 161 | */ |
| 162 | async edit(request: ImageEditDto) { |
| 163 | const { image, mask, user, ...params } = request |
| 164 | const runtimeModel = this.resolveRuntimeImageModel(params.model, 'edit') |
| 165 | |
| 166 | let imageFile: Uploadable | Uploadable[] |
| 167 | if (Array.isArray(image)) { |
| 168 | imageFile = await Promise.all(image.map(async (img, index) => |
| 169 | this.getUploadableByUrlOrDataUri(await this.resolveRelayText(img), `image-${index}`), |
| 170 | )) |
| 171 | } |
| 172 | else { |
| 173 | imageFile = await this.getUploadableByUrlOrDataUri(await this.resolveRelayText(image), 'image') |
| 174 | } |
| 175 | |
| 176 | const maskFile = mask ? await this.getUploadableByUrlOrDataUri(await this.resolveRelayText(mask), 'mask') : undefined |
| 177 | |
| 178 | if (runtimeModel === 'gpt-image-1') { |
| 179 | delete params.response_format |
| 180 | } |
| 181 | const imageResult = await this.openaiService.createImageEdit({ |
| 182 | ...params, |
| 183 | model: runtimeModel, |
| 184 | image: imageFile, |
| 185 | mask: maskFile, |
| 186 | size: params.size as 'auto', |
| 187 | }) |
| 188 | |
| 189 | for (const image of imageResult.data || []) { |
| 190 | if (image.url) { |
| 191 | image.url = await this.uploadImageToS3(image.url, user!, `${request.model}`) |
| 192 | } |
| 193 | if (image.b64_json) { |
| 194 | const mimeType = 'image/png' |
| 195 | const buffer = Buffer.from(image.b64_json, 'base64') |
| 196 | const uploadResult = await this.assetsService.uploadFromBuffer(user!, buffer, { |
| 197 | type: AssetType.AiImage, |
| 198 | mimeType, |
| 199 | }, `${request.model}`) |
| 200 | image.url = uploadResult.asset.path |
| 201 | delete image.b64_json |
| 202 | } |
| 203 | } |
| 204 | |
| 205 | return { |
| 206 | created: imageResult.created, |
| 207 | list: imageResult.data || [], |
| 208 | usage: imageResult.usage, |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | /** |
| 213 | * Gemini 图片生成 |
| 214 | */ |
| 215 | async geminiGeneration(userId: string, request: GeminiImageGenerationDto & { model: 'gemini-3.1-flash-image-preview' | 'gemini-3-pro-image-preview' }) { |
| 216 | const { model } = request |
| 217 | const result = await this.geminiService.generateImage({ |
| 218 | prompt: request.prompt, |
| 219 | imageUrls: request.imageUrls, |
| 220 | imageSize: request.imageSize, |
| 221 | aspectRatio: request.aspectRatio, |
| 222 | model, |
| 223 | }) |
| 224 | |
| 225 | const images: { url: string, data: string, mimeType: string }[] = [] |
| 226 | for (const image of result.images) { |
| 227 | const uploadResult = await this.assetsService.uploadFromBuffer(userId, image.imageData, { |
| 228 | type: AssetType.AiImage, |
| 229 | mimeType: image.mimeType, |
| 230 | }, `${model}`) |
| 231 | images.push({ url: uploadResult.asset.path, data: image.imageData.toString('base64'), mimeType: image.mimeType }) |
| 232 | } |
| 233 | |
| 234 | return { |
| 235 | images, |
| 236 | usage: result.usage, |
| 237 | } |
| 238 | } |
| 239 | |
| 240 | /** |
| 241 | * 用户 Gemini 图片生成 |
| 242 | */ |
| 243 | async userGeminiGeneration(request: UserGeminiImageGenerationDto) { |
| 244 | const { userId, userType, model: requestedModel, ...params } = request |
| 245 | const model = requestedModel || 'gemini-3.1-flash-image-preview' |
| 246 | |
| 247 | const modelConfig = await this.getGeminiImageModelConfig(model) |
| 248 | if (!modelConfig) { |
| 249 | throw new AppException(ResponseCode.InvalidModel) |
| 250 | } |
| 251 | |
| 252 | const startedAt = new Date() |
| 253 | const result = await this.geminiGeneration(userId, { ...params, model }) |
| 254 | |
| 255 | const usage = result.usage || { promptTokenCount: 0, candidatesTokenCount: 0, totalTokenCount: 0 } |
| 256 | |
| 257 | const duration = Date.now() - startedAt.getTime() |
| 258 | |
| 259 | await this.aiLogRepo.create({ |
| 260 | userId, |
| 261 | userType, |
| 262 | model, |
| 263 | channel: AiLogChannel.Gemini, |
| 264 | type: AiLogType.Image, |
| 265 | request: params, |
| 266 | response: { ...result, data: void 0 }, |
| 267 | status: AiLogStatus.Success, |
| 268 | startedAt, |
| 269 | duration, |
| 270 | }) |
| 271 | |
| 272 | return { |
| 273 | ...result, |
| 274 | usage: { |
| 275 | input_tokens: usage.promptTokenCount, |
| 276 | output_tokens: usage.candidatesTokenCount, |
| 277 | total_tokens: usage.totalTokenCount, |
| 278 | input_token_details: usage.inputTokenDetails, |
| 279 | output_token_details: usage.outputTokenDetails, |
| 280 | }, |
| 281 | } |
| 282 | } |
| 283 | |
| 284 | /** |
| 285 | * 获取 Gemini 图片模型配置 |
| 286 | */ |
| 287 | private async getGeminiImageModelConfig( |
| 288 | model: 'gemini-3.1-flash-image-preview' | 'gemini-3-pro-image-preview', |
| 289 | ) { |
| 290 | const chatModels = this.modelsConfigService.config.chat |
| 291 | const modelConfig = chatModels.find(m => m.name === model) |
| 292 | if (!modelConfig) { |
| 293 | return null |
| 294 | } |
| 295 | |
| 296 | return modelConfig |
| 297 | } |
| 298 | |
| 299 | /** |
| 300 | * 统一的用户请求处理：日志记录和状态更新 |
| 301 | */ |
| 302 | private async handleUserAiAction<T>(opts: { |
| 303 | userId: string |
| 304 | userType: UserType |
| 305 | model: string |
| 306 | channel?: AiLogChannel |
| 307 | type: AiLogType |
| 308 | request: Record<string, unknown> |
| 309 | retry?: number |
| 310 | run: () => Promise<T> |
| 311 | }): Promise<T> { |
| 312 | const { userId, userType, model, channel, type, request, retry, run } = opts |
| 313 | const startedAt = new Date() |
| 314 | |
| 315 | const log = await this.aiLogRepo.create({ |
| 316 | userId, |
| 317 | userType, |
| 318 | model, |
| 319 | channel: channel ?? AiLogChannel.NewApi, |
| 320 | type, |
| 321 | request, |
| 322 | status: AiLogStatus.Generating, |
| 323 | startedAt, |
| 324 | }) |
| 325 | |
| 326 | const result = await runWithAiGenerationRetry( |
| 327 | run, |
| 328 | retry, |
| 329 | (error, attempt, maxAttempts) => { |
| 330 | this.logger.warn( |
| 331 | { error, model, attempt, maxAttempts }, |
| 332 | 'AI image generation failed, retrying', |
| 333 | ) |
| 334 | }, |
| 335 | ).catch(async (e) => { |
| 336 | const duration = Date.now() - startedAt.getTime() |
| 337 | const errorMessage = getErrorMessage(e) |
| 338 | |
| 339 | await this.aiLogRepo.updateById(log.id, { |
| 340 | duration, |
| 341 | status: AiLogStatus.Failed, |
| 342 | errorMessage, |
| 343 | }) |
| 344 | throw e |
| 345 | }) |
| 346 | const duration = Date.now() - startedAt.getTime() |
| 347 | |
| 348 | await this.aiLogRepo.updateById(log.id, { |
| 349 | duration, |
| 350 | status: AiLogStatus.Success, |
| 351 | response: result as Record<string, unknown>, |
| 352 | }) |
| 353 | |
| 354 | return result |
| 355 | } |
| 356 | |
| 357 | /** |
| 358 | * 用户图片生成 |
| 359 | */ |
| 360 | async userGeneration(request: UserImageGenerationDto) { |
| 361 | const { userId, userType, ...params } = request |
| 362 | |
| 363 | return await this.handleUserAiAction({ |
| 364 | userId, |
| 365 | userType, |
| 366 | model: params.model, |
| 367 | type: AiLogType.Image, |
| 368 | request: params, |
| 369 | retry: this.getImageModelRetry(params.model, 'generation'), |
| 370 | run: () => this.generation({ ...params, user: userId }), |
| 371 | }) |
| 372 | } |
| 373 | |
| 374 | /** |
| 375 | * 用户图片编辑 |
| 376 | */ |
| 377 | async userEdit(request: UserImageEditDto) { |
| 378 | const { userId, userType, ...params } = request |
| 379 | |
| 380 | return await this.handleUserAiAction({ |
| 381 | userId, |
| 382 | userType, |
| 383 | model: params.model, |
| 384 | type: AiLogType.Image, |
| 385 | request: params, |
| 386 | retry: this.getImageModelRetry(params.model, 'edit'), |
| 387 | run: () => this.edit({ ...params, user: userId }), |
| 388 | }) |
| 389 | } |
| 390 | |
| 391 | /** |
| 392 | * 获取图片生成模型参数 |
| 393 | * @param data 查询参数，包含可选的 userId 和 userType，可用于后续个性化模型推荐 |
| 394 | */ |
| 395 | async generationModelConfig(_data: ImageGenerationModelsQueryDto) { |
| 396 | return this.modelsConfigService.config.image.generation |
| 397 | } |
| 398 | |
| 399 | /** |
| 400 | * 获取图片编辑模型参数 |
| 401 | * @param data 查询参数，包含可选的 userId 和 userType，可用于后续个性化模型推荐 |
| 402 | */ |
| 403 | async editModelConfig(_data: ImageEditModelsQueryDto) { |
| 404 | return this.modelsConfigService.config.image.edit |
| 405 | } |
| 406 | |
| 407 | /** |
| 408 | * 异步图片生成 |
| 409 | */ |
| 410 | @Transactional() |
| 411 | async userGenerationAsync(request: UserImageGenerationDto) { |
| 412 | const { userId, userType, ...params } = request |
| 413 | |
| 414 | // 创建 AiLog 记录 |
| 415 | const log = await this.aiLogRepo.create({ |
| 416 | userId, |
| 417 | userType, |
| 418 | model: params.model, |
| 419 | channel: AiLogChannel.NewApi, |
| 420 | type: AiLogType.Image, |
| 421 | request: params, |
| 422 | status: AiLogStatus.Generating, |
| 423 | startedAt: new Date(), |
| 424 | }) |
| 425 | |
| 426 | // 添加队列任务 |
| 427 | await this.queueService.addAiImageAsyncJob({ |
| 428 | logId: log.id, |
| 429 | userId, |
| 430 | userType, |
| 431 | model: params.model, |
| 432 | channel: AiLogChannel.NewApi, |
| 433 | type: AiLogType.Image, |
| 434 | retry: this.getImageModelRetry(params.model, 'generation'), |
| 435 | request: { ...params, user: userId }, |
| 436 | taskType: 'generation', |
| 437 | }) |
| 438 | |
| 439 | return { |
| 440 | logId: log.id, |
| 441 | status: AiLogStatus.Generating, |
| 442 | } |
| 443 | } |
| 444 | |
| 445 | /** |
| 446 | * 异步图片编辑 |
| 447 | */ |
| 448 | @Transactional() |
| 449 | async userEditAsync(request: UserImageEditDto) { |
| 450 | const { userId, userType, ...params } = request |
| 451 | |
| 452 | // 创建 AiLog 记录 |
| 453 | const log = await this.aiLogRepo.create({ |
| 454 | userId, |
| 455 | userType, |
| 456 | model: params.model, |
| 457 | channel: AiLogChannel.NewApi, |
| 458 | type: AiLogType.Image, |
| 459 | request: params, |
| 460 | status: AiLogStatus.Generating, |
| 461 | startedAt: new Date(), |
| 462 | }) |
| 463 | |
| 464 | // 添加队列任务 |
| 465 | await this.queueService.addAiImageAsyncJob({ |
| 466 | logId: log.id, |
| 467 | userId, |
| 468 | userType, |
| 469 | model: params.model, |
| 470 | channel: AiLogChannel.NewApi, |
| 471 | type: AiLogType.Image, |
| 472 | retry: this.getImageModelRetry(params.model, 'edit'), |
| 473 | request: { ...params, user: userId }, |
| 474 | taskType: 'edit', |
| 475 | }) |
| 476 | |
| 477 | return { |
| 478 | logId: log.id, |
| 479 | status: AiLogStatus.Generating, |
| 480 | } |
| 481 | } |
| 482 | |
| 483 | /** |
| 484 | * 查询任务状态 |
| 485 | */ |
| 486 | async getTaskStatus(logId: string) { |
| 487 | const log = await this.aiLogRepo.getById(logId) |
| 488 | if (!log || log.type !== AiLogType.Image) { |
| 489 | throw new NotFoundException('任务不存在') |
| 490 | } |
| 491 | const response = log.response as ImageAiLogResponse | undefined |
| 492 | |
| 493 | // 提取图片信息 |
| 494 | let images: AiLogImageResult[] | undefined |
| 495 | if (response?.list?.length) { |
| 496 | images = response.list |
| 497 | } |
| 498 | else if (response?.images?.length) { |
| 499 | images = response.images |
| 500 | } |
| 501 | else if (response?.image) { |
| 502 | images = [{ url: response.image }] |
| 503 | } |
| 504 | else if (response?.imageUrl) { |
| 505 | images = [{ url: response.imageUrl }] |
| 506 | } |
| 507 | |
| 508 | return { |
| 509 | logId: log.id, |
| 510 | status: log.status, |
| 511 | startedAt: log.startedAt, |
| 512 | duration: log.duration, |
| 513 | request: log.request, |
| 514 | response, |
| 515 | images, |
| 516 | errorMessage: log.errorMessage, |
| 517 | createdAt: log.createdAt, |
| 518 | updatedAt: log.updatedAt, |
| 519 | } |
| 520 | } |
| 521 | } |
| 522 |