ai-chatbot/lib/ai/box-cutout/service.ts

import sharp from "sharp";

import {
  executeRmbgJob,
  fetchImageBufferFromUrl,
} from "@/lib/ai/rmbg/service";

export type NormalizedCropRect = {
  x: number;
  y: number;
  width: number;
  height: number;
};

export interface ExecuteBoxCutoutJobParams {
  userId: string;
  imageUrl: string;
  cropRect: Partial<NormalizedCropRect>;
  paddingRatio?: number;
  prompt?: string;
  chatId?: string | null;
  messageId?: string | null;
  skipGallerySave?: boolean;
  sourceNodeId?: string;
}

export interface ExecuteBoxCutoutJobResult {
  success: true;
  original_url: string;
  thumbnail_url?: string | null;
  generatedImageId?: string | null;
  mediaAssetId?: string | null;
  cropRectApplied: NormalizedCropRect;
  paddedCropRectApplied: NormalizedCropRect;
  cropRectPixels: {
    left: number;
    top: number;
    width: number;
    height: number;
  };
  paddedCropRectPixels: {
    left: number;
    top: number;
    width: number;
    height: number;
  };
  sourceNodeId?: string | null;
}

export class BoxCutoutServiceError extends Error {
  status: number;
  detail?: string;

  constructor(message: string, status = 400, detail?: string) {
    super(message);
    this.name = "BoxCutoutServiceError";
    this.status = status;
    this.detail = detail;
  }
}

const MIN_CROP_EDGE_PX = 24;
const DEFAULT_PADDING_RATIO = 0.08;

function clamp(value: number, min: number, max: number) {
  return Math.min(max, Math.max(min, value));
}

export function sanitizeNormalizedCropRect(
  input: Partial<NormalizedCropRect> | undefined,
): NormalizedCropRect | null {
  if (!input) return null;

  const x = Number(input.x);
  const y = Number(input.y);
  const width = Number(input.width);
  const height = Number(input.height);

  if (![x, y, width, height].every((value) => Number.isFinite(value))) {
    return null;
  }

  const normalized: NormalizedCropRect = {
    x: clamp(x, 0, 1),
    y: clamp(y, 0, 1),
    width: clamp(width, 0, 1),
    height: clamp(height, 0, 1),
  };

  if (normalized.width <= 0 || normalized.height <= 0) {
    return null;
  }

  normalized.width = Math.min(normalized.width, 1 - normalized.x);
  normalized.height = Math.min(normalized.height, 1 - normalized.y);

  if (normalized.width <= 0 || normalized.height <= 0) {
    return null;
  }

  return normalized;
}

export function normalizeBoxCutoutPaddingRatio(value: unknown) {
  const numeric = Number(value);
  if (!Number.isFinite(numeric)) return DEFAULT_PADDING_RATIO;
  return clamp(numeric, 0, 0.25);
}

function toPixelRect(
  rect: NormalizedCropRect,
  sourceWidth: number,
  sourceHeight: number,
) {
  const left = clamp(Math.round(rect.x * sourceWidth), 0, Math.max(sourceWidth - 1, 0));
  const top = clamp(Math.round(rect.y * sourceHeight), 0, Math.max(sourceHeight - 1, 0));
  const width = clamp(Math.round(rect.width * sourceWidth), 1, sourceWidth - left);
  const height = clamp(Math.round(rect.height * sourceHeight), 1, sourceHeight - top);

  return { left, top, width, height };
}

function applyPadding(
  rect: NormalizedCropRect,
  paddingRatio: number,
): NormalizedCropRect {
  const padX = rect.width * paddingRatio;
  const padY = rect.height * paddingRatio;

  const x = clamp(rect.x - padX, 0, 1);
  const y = clamp(rect.y - padY, 0, 1);
  const right = clamp(rect.x + rect.width + padX, 0, 1);
  const bottom = clamp(rect.y + rect.height + padY, 0, 1);

  return {
    x,
    y,
    width: clamp(right - x, 0, 1),
    height: clamp(bottom - y, 0, 1),
  };
}

export async function executeBoxCutoutJob(
  params: ExecuteBoxCutoutJobParams,
): Promise<ExecuteBoxCutoutJobResult> {
  const imageUrl = String(params.imageUrl || "").trim();
  if (!imageUrl) {
    throw new BoxCutoutServiceError("缺少 imageUrl 参数", 400);
  }

  const cropRect = sanitizeNormalizedCropRect(params.cropRect);
  if (!cropRect) {
    throw new BoxCutoutServiceError("缺少有效的 cropRect 参数", 400);
  }

  const paddingRatio = normalizeBoxCutoutPaddingRatio(params.paddingRatio);
  const sourceBuffer = await sharp(await fetchImageBufferFromUrl(imageUrl), {
    failOn: "none",
  })
    .rotate()
    .toBuffer();
  const sourceImage = sharp(sourceBuffer, { failOn: "none" });
  const sourceMetadata = await sourceImage.metadata();
  const sourceWidth = Number(sourceMetadata.width || 0);
  const sourceHeight = Number(sourceMetadata.height || 0);

  if (sourceWidth <= 0 || sourceHeight <= 0) {
    throw new BoxCutoutServiceError("无法识别原图尺寸", 400);
  }

  const cropRectPixels = toPixelRect(cropRect, sourceWidth, sourceHeight);
  if (cropRectPixels.width < MIN_CROP_EDGE_PX || cropRectPixels.height < MIN_CROP_EDGE_PX) {
    throw new BoxCutoutServiceError("框选区域过小，请重新框选", 400);
  }

  const paddedCropRect = applyPadding(cropRect, paddingRatio);
  const paddedCropRectPixels = toPixelRect(paddedCropRect, sourceWidth, sourceHeight);
  const croppedBuffer = await sourceImage
    .extract({
      left: paddedCropRectPixels.left,
      top: paddedCropRectPixels.top,
      width: paddedCropRectPixels.width,
      height: paddedCropRectPixels.height,
    })
    .toBuffer();

  const result = await executeRmbgJob({
    userId: params.userId,
    imageBuffer: croppedBuffer,
    imageBase64: croppedBuffer.toString("base64"),
    sourceImageUrl: imageUrl,
    prompt: params.prompt || "框选抠图",
    chatId: params.chatId || null,
    messageId: params.messageId || null,
    skipGallerySave: params.skipGallerySave === true,
  });

  return {
    success: true,
    original_url: result.original_url,
    thumbnail_url: result.thumbnail_url,
    generatedImageId: result.generatedImageId,
    mediaAssetId: result.mediaAssetId,
    cropRectApplied: cropRect,
    paddedCropRectApplied: paddedCropRect,
    cropRectPixels,
    paddedCropRectPixels,
    sourceNodeId: params.sourceNodeId || null,
  };
}