Agent/examples/content_finder/tools/content_eval.py

"""
内容评估工具 - 评估内容质量和相关性
"""

from typing import Dict, Any, List
from agent.tools import tool, ToolResult, ToolContext


@tool(description="评估内容质量和相关性")
async def evaluate_content(
    content_items: List[Dict[str, Any]],
    keywords: List[str],
    ctx: ToolContext = None,
) -> ToolResult:
    """
    评估内容质量和相关性

    Args:
        content_items: 内容列表
        keywords: 搜索关键词
        ctx: 工具上下文
    """
    evaluated_items = []

    for item in content_items:
        score = _calculate_quality_score(item, keywords)
        item["quality_score"] = score
        item["relevance_score"] = _calculate_relevance_score(item, keywords)
        evaluated_items.append(item)

    # 按质量分数排序
    evaluated_items.sort(key=lambda x: x["quality_score"], reverse=True)

    return ToolResult(
        title="内容评估完成",
        output=f"已评估 {len(evaluated_items)} 条内容",
        data={"evaluated_items": evaluated_items},
    )


def _calculate_quality_score(item: Dict[str, Any], keywords: List[str]) -> float:
    """
    计算内容质量分数

    考虑因素：
    1. 播放量、点赞数、评论数、分享数
    2. 互动率（点赞/播放、评论/播放）
    3. 发布时间（新鲜度）
    """
    stats = item.get("stats", {})
    views = stats.get("views", 0)
    likes = stats.get("likes", 0)
    comments = stats.get("comments", 0)
    shares = stats.get("shares", 0)

    # 基础分数：基于绝对数值
    base_score = (
        (views / 10000) * 0.3 +
        (likes / 1000) * 0.3 +
        (comments / 100) * 0.2 +
        (shares / 100) * 0.2
    )

    # 互动率加成
    engagement_rate = (likes + comments + shares) / max(views, 1)
    engagement_bonus = engagement_rate * 20

    # 总分
    total_score = min(base_score + engagement_bonus, 100)

    return round(total_score, 2)


def _calculate_relevance_score(item: Dict[str, Any], keywords: List[str]) -> float:
    """
    计算内容相关性分数

    考虑因素：
    1. 标题中关键词匹配度
    2. 描述中关键词匹配度
    3. 标签匹配度
    """
    title = item.get("title", "").lower()
    description = item.get("description", "").lower()
    tags = [t.lower() for t in item.get("tags", [])]

    score = 0.0
    keyword_count = len(keywords)

    for keyword in keywords:
        keyword_lower = keyword.lower()

        # 标题匹配（权重最高）
        if keyword_lower in title:
            score += 40 / keyword_count

        # 描述匹配
        if keyword_lower in description:
            score += 30 / keyword_count

        # 标签匹配
        if keyword_lower in tags:
            score += 30 / keyword_count

    return min(score, 100)