import { NextResponse } from "next/server";
import { uploadFile } from "@/lib/storage/service";
import { callQwen3Vl } from "@/lib/ai/qwen3-vl";
import sharp from "sharp";

export const runtime = "nodejs";
export const maxDuration = 30;

// 位置词库
const POSITION_OPTIONS = [
  '左上区域', '中央顶部', '右上区域',
  '左侧中部', '中央区域', '右侧中部',
  '左下区域', '中央底部', '右下区域'
] as const;

interface PickObjectRequest {
  imageData: string;        // 带红色标记的图片 base64
  clickX: number;           // 点击位置 X (0-1)
  clickY: number;           // 点击位置 Y (0-1)
}

interface PickObjectResult {
  name: string;
  position: string;
  confidence: number;
  boundingBox: {
    x: number;
    y: number;
    width: number;
    height: number;
  };
}

const MAX_VISION_SIDE = 500;
const PICK_OBJECT_FETCH_TIMEOUT_MS = 15000;

async function downscaleImageForVision(buffer: Buffer) {
  const source = sharp(buffer, { failOn: "none" });
  const meta = await source.metadata();
  const hasAlpha = Boolean(meta.hasAlpha);
  const width = meta.width || 0;
  const height = meta.height || 0;

  const resized = source
    .rotate()
    .resize({
      width: MAX_VISION_SIDE,
      height: MAX_VISION_SIDE,
      fit: "inside",
      withoutEnlargement: true,
    });

  if (hasAlpha) {
    return {
      buffer: await resized.png({ compressionLevel: 9 }).toBuffer(),
      ext: "png",
      width,
      height,
    };
  }

  return {
    buffer: await resized.jpeg({ quality: 84, mozjpeg: true }).toBuffer(),
    ext: "jpg",
    width,
    height,
  };
}

async function uploadVisionAnalysisImage(buffer: Buffer, filenamePrefix: string) {
  const { buffer: processedBuffer, ext, width, height } = await downscaleImageForVision(buffer);
  const filename = `${filenamePrefix}-${Date.now()}.${ext}`;
  const uploaded = await uploadFile(filename, processedBuffer as any);
  console.log("[PickObject] 视觉分析图片已缩放:", {
    originalWidth: width,
    originalHeight: height,
    maxSide: MAX_VISION_SIDE,
    outputExt: ext,
  });
  return uploaded.url;
}

async function normalizeImageInput(imageData: string): Promise<string> {
  const raw = String(imageData || "").trim();
  if (!raw) throw new Error("缺少图片数据");

  if (/^https?:\/\//i.test(raw)) {
    const controller = new AbortController();
    const timeout = setTimeout(() => controller.abort(), PICK_OBJECT_FETCH_TIMEOUT_MS);
    try {
      const response = await fetch(raw, {
        method: "GET",
        signal: controller.signal,
      });
      if (!response.ok) {
        throw new Error(`拉取图片失败: HTTP ${response.status}`);
      }
      const arrayBuffer = await response.arrayBuffer();
      return uploadVisionAnalysisImage(Buffer.from(arrayBuffer), "pick-object-url");
    } finally {
      clearTimeout(timeout);
    }
  }

  const base64Match = raw.match(/^data:(image\/[a-zA-Z0-9.+-]+);base64,(.+)$/);
  if (!base64Match?.[1] || !base64Match?.[2]) {
    throw new Error("图片格式不支持");
  }

  const base64Body = base64Match[2];
  const buffer = Buffer.from(base64Body, "base64");
  return uploadVisionAnalysisImage(buffer, "pick-object-base64");
}

export async function POST(request: Request) {
  try {
    const body: PickObjectRequest = await request.json();
    const { imageData, clickX, clickY } = body;
    
    console.log('[PickObject] 收到请求:', { imageDataLength: imageData?.length, clickX, clickY });
    
    // 验证参数
    if (!imageData) {
      return NextResponse.json({
        success: false,
        error: '缺少图片数据',
      }, { status: 400 });
    }
    
    if (typeof clickX !== 'number' || typeof clickY !== 'number') {
      return NextResponse.json({
        success: false,
        error: '缺少点击坐标',
      }, { status: 400 });
    }
    
    const normalizedImage = await normalizeImageInput(imageData);

    console.log("[PickObject] 使用硬编码模型: qwen3-vl");
    
    // 构建 prompt - AI 通过红色标记来识别用户选择的物品
    const prompt = `
图片中有一个红色圆形标记（带十字线），请识别这个红色标记所指向的具体物品。

【重要】
- 只识别红色标记正中心位置所属的“完整产品/主体”
- 如果标记落在局部部件上（如椅子扶手、桌子腿、沙发靠背），也要返回完整物体名称（如“椅子/桌子/沙发”），不要返回部件名称
- 例如：标记在花瓶上 → 返回"花瓶"（不是"桌子"或"屏风"）；标记在椅子扶手 → 返回"椅子"（不是"扶手"）

【输出要求 - 严格遵守】
返回纯 JSON，无其他文字：
{
  "name": "物品名称",
  "position": "位置描述",
  "confidence": 0.95
}

【name 格式】≤6字，简洁具体
正确示例：橙色花瓶、白色杯子、木质椅子、吊灯、抱枕、花盆
错误示例：屏风（太大）、一个漂亮的花瓶（太长）

【position 格式】固定4字，从以下选择：
${POSITION_OPTIONS.join('、')}
`.trim();

    console.log('[PickObject] 调用视觉模型...');

    // 调用视觉模型，发送带红色标记的图片
    const { text: responseText, model } = await callQwen3Vl({
      imageUrl: normalizedImage,
      prompt,
      temperature: 0.1,
      maxTokens: 300,
    });
    
    console.log('[PickObject] 模型返回内容:', model, responseText?.substring(0, 200));
    
    if (!responseText) {
      return NextResponse.json({
        success: false,
        error: 'AI 未返回结果',
      }, { status: 500 });
    }
    
    // 解析 JSON 结果
    let result: PickObjectResult;
    try {
      // 尝试从返回内容中提取 JSON
      const jsonMatch = responseText.match(/\{[\s\S]*\}/);
      if (!jsonMatch) {
        throw new Error('未找到 JSON');
      }
      const parsed = JSON.parse(jsonMatch[0]);
      
      // 验证和修正名称
      let name = parsed.name || '物品';
      if (name.length > 8) {
        name = name.substring(0, 6);
      }
      
      // 验证位置是否在词库中
      let position = parsed.position;
      if (!POSITION_OPTIONS.includes(position as any)) {
        // 根据点击位置推算
        const posX = clickX < 0.33 ? '左' : clickX > 0.66 ? '右' : '中央';
        const posY = clickY < 0.33 ? '上' : clickY > 0.66 ? '下' : '中';
        if (posX === '中央') {
          position = posY === '中' ? '中央区域' : posY === '上' ? '中央顶部' : '中央底部';
        } else {
          position = posY === '中' ? `${posX}侧中部` : `${posX}${posY}区域`;
        }
      }
      
      // 确保 confidence 有效
      let confidence = parsed.confidence;
      if (typeof confidence !== 'number' || confidence < 0 || confidence > 1) {
        confidence = 0.85;
      }
      
      result = {
        name,
        position,
        confidence,
        boundingBox: { x: clickX, y: clickY, width: 0, height: 0 }, // 不再使用，保持兼容
      };
      
    } catch (parseError) {
      console.error('[PickObject] JSON parse error:', parseError, 'Content:', responseText);
      
      // 返回默认结果
      const posX = clickX < 0.33 ? '左' : clickX > 0.66 ? '右' : '中央';
      const posY = clickY < 0.33 ? '上' : clickY > 0.66 ? '下' : '中';
      const position = posX === '中央' 
        ? (posY === '中' ? '中央区域' : posY === '上' ? '中央顶部' : '中央底部')
        : (posY === '中' ? `${posX}侧中部` : `${posX}${posY}区域`);
      
      result = {
        name: '物品',
        position,
        confidence: 0.5,
        boundingBox: { x: clickX, y: clickY, width: 0, height: 0 },
      };
    }
    
    console.log('[PickObject] 识别结果:', result);
    
    return NextResponse.json({
      success: true,
      object: result,
      model: {
        id: model,
        name: model,
        provider: "alibaba-image-key",
      },
    });
    
  } catch (error) {
    console.error('[PickObject] Error:', error);
    return NextResponse.json({
      success: false,
      error: `识别失败: ${error instanceof Error ? error.message : '未知错误'}`,
    }, { status: 500 });
  }
}