"""
OCR service: Use vision model to extract medical report data from images.
Supports OpenAI-compatible APIs (e.g. SiliconFlow).
"""

import base64
import json
import logging
from pathlib import Path

import aiofiles
from openai import AsyncOpenAI

from app.config import settings

logger = logging.getLogger(__name__)

OCR_PROMPT = """请分析这张图片，首先判断它是否是孕期/产检相关的医疗检查报告。

**孕期相关报告包括但不限于：**
- 产检报告（B超/彩超、NT检查、四维彩超、大排畸等）
- 血液检查（血常规、血糖、血型、唐氏筛查、TORCH、甲功等）
- 尿液检查（尿常规等）
- 肝肾功能检查
- 凝血功能检查
- 心电图报告
- 白带常规
- 胎心监护报告
- 其他与孕期产检相关的检查报告

**不属于孕期报告的情况：**
- 非医疗报告（发票、收据、菜单、截图、照片等）
- 与孕期无关的医疗报告（如骨科、眼科、口腔科等非产检相关）
- 无法识别内容的模糊图片

**请严格按以下JSON格式返回（不要输出其他内容）：**

如果**不是**孕期相关报告，返回：
{
  "is_pregnancy_report": false,
  "reject_reason": "简短说明为什么不是孕期报告（如：这是一张超市购物小票）"
}

如果**是**孕期相关报告，提取所有检查指标并返回：
{
  "is_pregnancy_report": true,
  "patient_name": "患者姓名（如有）",
  "report_type": "报告类型（如血常规、尿常规、B超等）",
  "report_date": "报告日期（如有，格式 YYYY-MM-DD）",
  "data_items": [
    {
      "name": "测定项目",
      "value": "测定结果（只提取数字，不要单位）",
      "unit": "单位",
      "reference_range": "参考范围（如有）"
    }
  ]
}

**重要提示：**
1. 必须先判断是否为孕期报告，is_pregnancy_report 字段必填
2. value字段只填写纯数字，不要包含单位（如"23.5"而不是"23.5mm"）
3. 单位单独放在unit字段中
4. 如果是B超报告，务必确保双顶径(BPD)、腹围(AC)、股骨长(FL)这三个指标被正确提取"""


def _get_client() -> AsyncOpenAI:
    return AsyncOpenAI(api_key=settings.ai_api_key, base_url=settings.ai_base_url)


async def ocr_recognize(image_path: str) -> dict:
    """
    Use vision model to extract medical report data from an image.
    Returns structured OCR results.
    """
    # Resolve path relative to upload_dir
    if image_path.startswith("/uploads/"):
        relative = image_path[len("/uploads/"):]
        full_path = Path(settings.upload_dir) / relative
    else:
        full_path = Path(settings.upload_dir) / image_path.lstrip("/")

    if not full_path.exists():
        logger.error("OCR 文件不存在: %s", full_path)
        raise FileNotFoundError(f"报告图片不存在: {image_path}")

    # Async file read
    async with aiofiles.open(full_path, "rb") as f:
        raw = await f.read()
    image_data = base64.standard_b64encode(raw).decode("utf-8")

    # Determine media type
    suffix = full_path.suffix.lower()
    media_type_map = {
        ".jpg": "image/jpeg",
        ".jpeg": "image/jpeg",
        ".png": "image/png",
        ".webp": "image/webp",
    }
    media_type = media_type_map.get(suffix, "image/jpeg")

    client = _get_client()
    try:
        response = await client.chat.completions.create(
            model=settings.ai_vision_model,
            max_tokens=2048,
            messages=[
                {
                    "role": "user",
                    "content": [
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:{media_type};base64,{image_data}",
                            },
                        },
                        {
                            "type": "text",
                            "text": OCR_PROMPT,
                        },
                    ],
                }
            ],
        )
    except Exception as e:
        err_msg = str(e)
        logger.error("AI OCR 调用失败: %s: %s", type(e).__name__, err_msg[:200])
        if "auth" in err_msg.lower() or "api key" in err_msg.lower():
            raise RuntimeError("AI 服务认证失败，请检查 API Key 是否有效") from e
        if "<html" in err_msg.lower() or "cloudflare" in err_msg.lower():
            raise RuntimeError("AI 服务请求被拦截，请检查 API 配置") from e
        raise RuntimeError("AI 识别服务暂时不可用，请稍后重试") from e

    response_text = response.choices[0].message.content
    try:
        if "```json" in response_text:
            response_text = response_text.split("```json")[1].split("```")[0]
        elif "```" in response_text:
            response_text = response_text.split("```")[1].split("```")[0]
        return json.loads(response_text.strip())
    except json.JSONDecodeError:
        return {"raw_text": response_text, "data_items": []}