kol-insight/backend/app/services/yuntu_api.py

"""
巨量云图API封装 (T-023)

封装GetContentMaterialAnalysisInfo接口调用，获取视频分析数据。
"""

import logging
from datetime import datetime, timedelta
from typing import Dict, List, Optional, Any

import httpx

from app.config import settings
from app.services.session_pool import session_pool, get_session_with_retry

logger = logging.getLogger(__name__)

# 巨量云图API基础URL
YUNTU_BASE_URL = "https://yuntu.oceanengine.com"

# 触发点ID列表（固定值）
TRIGGER_POINT_IDS = ["610000", "610300", "610301"]


class YuntuAPIError(Exception):
    """巨量云图API错误"""

    def __init__(self, message: str, status_code: int = 0, response_data: Any = None):
        self.message = message
        self.status_code = status_code
        self.response_data = response_data
        super().__init__(self.message)


class SessionInvalidError(YuntuAPIError):
    """SessionID失效错误"""

    pass


async def call_yuntu_api(
    item_id: str,
    publish_time: datetime,
    industry_id: str,
    session_id: Optional[str] = None,
) -> Dict[str, Any]:
    """
    调用巨量云图GetContentMaterialAnalysisInfo接口。

    Args:
        item_id: 视频ID
        publish_time: 发布时间
        industry_id: 行业ID
        session_id: 可选的sessionid，不提供则从池中获取

    Returns:
        Dict: API响应数据

    Raises:
        SessionInvalidError: SessionID失效时抛出
        YuntuAPIError: API调用失败时抛出
    """
    # 获取sessionid
    if session_id is None:
        session_id = await get_session_with_retry()
        if session_id is None:
            raise YuntuAPIError("Failed to get valid session")

    # 构造请求参数
    # end_date = start_date + 30天
    start_date = publish_time.strftime("%Y-%m-%d")
    end_date = (publish_time + timedelta(days=30)).strftime("%Y-%m-%d")

    request_data = {
        "is_my_video": "0",
        "object_id": item_id,
        "object_type": 2,
        "start_date": start_date,
        "end_date": end_date,
        "assist_type": 3,
        "assist_video_type": 3,
        "industry_id_list": [industry_id] if industry_id else [],
        "trigger_point_id_list": TRIGGER_POINT_IDS,
    }

    # 构造请求头
    headers = {
        "Content-Type": "application/json",
        "Cookie": f"sessionid={session_id}",
    }

    try:
        async with httpx.AsyncClient(timeout=settings.YUNTU_API_TIMEOUT) as client:
            response = await client.post(
                f"{YUNTU_BASE_URL}/yuntu_common/api/content/trigger_analysis/GetContentMaterialAnalysisInfo",
                json=request_data,
                headers=headers,
            )

            # 检查SessionID是否失效
            if response.status_code in (401, 403):
                logger.warning(f"Session invalid: {session_id[:8]}...")
                raise SessionInvalidError(
                    f"Session invalid: {response.status_code}",
                    status_code=response.status_code,
                )

            if response.status_code != 200:
                raise YuntuAPIError(
                    f"API returned {response.status_code}",
                    status_code=response.status_code,
                    response_data=response.text,
                )

            data = response.json()

            # 检查业务错误码
            if data.get("code") != 0:
                error_msg = data.get("message", "Unknown error")
                raise YuntuAPIError(
                    f"API business error: {error_msg}",
                    status_code=response.status_code,
                    response_data=data,
                )

            return data

    except httpx.TimeoutException:
        logger.error(f"Yuntu API timeout for item_id: {item_id}")
        raise YuntuAPIError("API request timeout")
    except httpx.RequestError as e:
        logger.error(f"Yuntu API request error: {e}")
        raise YuntuAPIError(f"API request error: {e}")


async def get_video_analysis(
    item_id: str,
    publish_time: datetime,
    industry_id: str,
    max_retries: int = 3,
) -> Dict[str, Any]:
    """
    获取视频分析数据，支持SessionID失效自动重试 (T-022)。

    Args:
        item_id: 视频ID
        publish_time: 发布时间
        industry_id: 行业ID
        max_retries: 最大重试次数

    Returns:
        Dict: 视频分析数据

    Raises:
        YuntuAPIError: 所有重试失败后抛出
    """
    last_error = None

    for attempt in range(max_retries):
        # 从池中获取sessionid
        session_id = await get_session_with_retry()
        if session_id is None:
            last_error = YuntuAPIError("Failed to get valid session")
            continue

        try:
            result = await call_yuntu_api(
                item_id=item_id,
                publish_time=publish_time,
                industry_id=industry_id,
                session_id=session_id,
            )
            return result

        except SessionInvalidError:
            # SessionID失效，从池中移除并重试
            session_pool.remove(session_id)
            logger.info(
                f"Session invalid, retrying... attempt {attempt + 1}/{max_retries}"
            )
            last_error = SessionInvalidError("All sessions invalid")
            continue

        except YuntuAPIError as e:
            last_error = e
            logger.error(f"Yuntu API error on attempt {attempt + 1}: {e.message}")
            # 非SessionID问题，不再重试
            break

    raise last_error or YuntuAPIError("Unknown error after retries")


def parse_analysis_response(data: Dict[str, Any]) -> Dict[str, Any]:
    """
    解析巨量云图API响应，提取关键指标。

    Args:
        data: API原始响应数据

    Returns:
        Dict: 结构化的分析数据
    """
    result_data = data.get("data", {})

    return {
        # 触达指标
        "total_show_cnt": result_data.get("total_show_cnt", 0),  # 总曝光数
        "natural_show_cnt": result_data.get("natural_show_cnt", 0),  # 自然曝光数
        "ad_show_cnt": result_data.get("ad_show_cnt", 0),  # 加热曝光数
        "total_play_cnt": result_data.get("total_play_cnt", 0),  # 总播放数
        "natural_play_cnt": result_data.get("natural_play_cnt", 0),  # 自然播放数
        "ad_play_cnt": result_data.get("ad_play_cnt", 0),  # 加热播放数
        "effective_play_cnt": result_data.get("effective_play_cnt", 0),  # 有效播放数
        # A3指标
        "a3_increase_cnt": result_data.get("a3_increase_cnt", 0),  # 新增A3
        "ad_a3_increase_cnt": result_data.get("ad_a3_increase_cnt", 0),  # 加热新增A3
        "natural_a3_increase_cnt": result_data.get("natural_a3_increase_cnt", 0),  # 自然新增A3
        # 搜索指标
        "after_view_search_uv": result_data.get("after_view_search_uv", 0),  # 看后搜人数
        "after_view_search_pv": result_data.get("after_view_search_pv", 0),  # 看后搜次数
        "brand_search_uv": result_data.get("brand_search_uv", 0),  # 品牌搜索人数
        "product_search_uv": result_data.get("product_search_uv", 0),  # 商品搜索人数
        "return_search_cnt": result_data.get("return_search_cnt", 0),  # 回搜次数
        # 费用指标
        "cost": result_data.get("cost", 0),  # 总花费
        "natural_cost": result_data.get("natural_cost", 0),  # 自然花费
        "ad_cost": result_data.get("ad_cost", 0),  # 加热花费
    }