how/script/data_processing/match_inspiration_features.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
灵感点特征匹配脚本

从解构任务列表中提取灵感点的特征，与人设灵感特征进行匹配，
使用 relation_analyzer 模块分析特征之间的语义关系。
"""

import json
import asyncio
from pathlib import Path
from typing import Dict, List
import sys
from tqdm import tqdm

# 添加项目根目录到路径
project_root = Path(__file__).parent.parent.parent
sys.path.insert(0, str(project_root))

from lib.hybrid_similarity import compare_phrases_cartesian
from script.data_processing.path_config import PathConfig

# 全局进度条
progress_bar = None


async def process_single_point(
    point: Dict,
    point_type: str,
    persona_features: List[Dict],
    category_mapping: Dict = None,
    model_name: str = None
) -> Dict:
    """
    处理单个点 - 使用笛卡尔积批量计算（优化版）

    Args:
        point: 点数据（灵感点/关键点/目的点）
        point_type: 点类型（"灵感点"/"关键点"/"目的点"）
        persona_features: 人设特征列表
        category_mapping: 特征分类映射字典
        model_name: 使用的模型名称

    Returns:
        包含 how 步骤列表的点数据
    """
    global progress_bar

    point_name = point.get("名称", "")
    feature_list = point.get("特征列表", [])

    # 如果没有特征，直接返回
    if not feature_list or not persona_features:
        result = point.copy()
        result["how步骤列表"] = []
        return result

    # 提取特征名称和人设名称列表
    feature_names = [f.get("特征名称", "") for f in feature_list]
    persona_names = [pf["特征名称"] for pf in persona_features]

    # 定义进度回调函数
    def on_llm_progress(count: int):
        """LLM完成一个任务时的回调"""
        if progress_bar:
            progress_bar.update(count)

    # 核心优化：使用混合模型笛卡尔积一次计算M×N
    # max_concurrent 控制的是底层 LLM 的全局并发数
    similarity_results = await compare_phrases_cartesian(
        feature_names,      # M个特征
        persona_names,      # N个人设
        max_concurrent=100,  # LLM最大并发数（全局共享）
        progress_callback=on_llm_progress  # 传递进度回调
    )
    # similarity_results[i][j] = {"相似度": float, "说明": str}


    # 构建匹配结果（使用模块返回的完整结果）
    feature_match_results = []

    for i, feature_item in enumerate(feature_list):
        feature_name = feature_item.get("特征名称", "")
        feature_weight = feature_item.get("权重", 1.0)

        # 该特征与所有人设的匹配结果
        match_results = []
        for j, persona_feature in enumerate(persona_features):
            persona_name = persona_feature["特征名称"]
            persona_level = persona_feature["人设特征层级"]

            # 直接使用模块返回的完整结果
            similarity_result = similarity_results[i][j]

            # 判断特征类型和分类
            feature_type = "分类"  # 默认为分类
            categories = []

            if category_mapping:
                # 先在标签特征中查找
                is_tag_feature = False
                for ft in ["灵感点", "关键点", "目的点"]:
                    if ft in category_mapping:
                        type_mapping = category_mapping[ft]
                        if persona_name in type_mapping:
                            feature_type = "标签"
                            categories = type_mapping[persona_name].get("所属分类", [])
                            is_tag_feature = True
                            break

                # 如果不是标签特征，检查是否是分类特征
                if not is_tag_feature:
                    all_categories = set()
                    for ft in ["灵感点", "关键点", "目的点"]:
                        if ft in category_mapping:
                            for _fname, fdata in category_mapping[ft].items():
                                cats = fdata.get("所属分类", [])
                                all_categories.update(cats)

                    if persona_name in all_categories:
                        feature_type = "分类"
                        categories = []

            # 去重分类
            unique_categories = list(dict.fromkeys(categories))

            match_result = {
                "人设特征名称": persona_name,
                "人设特征层级": persona_level,
                "特征类型": feature_type,
                "特征分类": unique_categories,
                "匹配结果": similarity_result  # 直接使用模块返回的结果
            }
            match_results.append(match_result)

        feature_match_results.append({
            "特征名称": feature_name,
            "权重": feature_weight,
            "匹配结果": match_results
        })

    # 构建 how 步骤（保持不变）
    step_name_mapping = {
        "灵感点": "灵感特征分别匹配人设特征",
        "关键点": "关键特征分别匹配人设特征",
        "目的点": "目的特征分别匹配人设特征"
    }

    how_step = {
        "步骤名称": step_name_mapping.get(point_type, f"{point_type}特征分别匹配人设特征"),
        "特征列表": list(feature_match_results)
    }

    result = point.copy()
    result["how步骤列表"] = [how_step]

    return result


async def process_single_task(
    task: Dict,
    task_index: int,
    total_tasks: int,
    all_persona_features: List[Dict],
    category_mapping: Dict = None,
    model_name: str = None
) -> Dict:
    """
    处理单个任务

    Args:
        task: 任务数据
        task_index: 任务索引（从1开始）
        total_tasks: 总任务数
        all_persona_features: 所有人设特征列表（包含三种层级）
        category_mapping: 特征分类映射字典
        model_name: 使用的模型名称

    Returns:
        包含 how 解构结果的任务
    """
    global progress_bar

    post_id = task.get("帖子id", "")

    # 获取 what 解构结果
    what_result = task.get("what解构结果", {})

    # 计算当前帖子的总匹配任务数
    current_task_match_count = 0
    for point_type in ["灵感点", "关键点", "目的点"]:
        point_list = what_result.get(f"{point_type}列表", [])
        for point in point_list:
            feature_count = len(point.get("特征列表", []))
            current_task_match_count += feature_count * len(all_persona_features)

    # 创建当前帖子的进度条
    progress_bar = tqdm(
        total=current_task_match_count,
        desc=f"[{task_index}/{total_tasks}] {post_id}",
        unit="匹配",
        ncols=100
    )

    # 构建 how 解构结果
    how_result = {}

    # 串行处理灵感点、关键点和目的点
    for point_type in ["灵感点", "关键点", "目的点"]:
        point_list_key = f"{point_type}列表"
        point_list = what_result.get(point_list_key, [])

        if point_list:
            updated_point_list = []
            # 串行处理每个点
            for point in point_list:
                result = await process_single_point(
                    point=point,
                    point_type=point_type,
                    persona_features=all_persona_features,
                    category_mapping=category_mapping,
                    model_name=model_name
                )
                updated_point_list.append(result)

            # 添加到 how 解构结果
            how_result[point_list_key] = updated_point_list

    # 关闭当前帖子的进度条
    if progress_bar:
        progress_bar.close()

    # 更新任务
    updated_task = task.copy()
    updated_task["how解构结果"] = how_result

    return updated_task


async def process_task_list(
    task_list: List[Dict],
    persona_features_dict: Dict,
    category_mapping: Dict = None,
    model_name: str = None,
    output_dir: Path = None
) -> List[Dict]:
    """
    处理整个解构任务列表（串行执行，每个帖子处理完立即保存）

    Args:
        task_list: 解构任务列表
        persona_features_dict: 人设特征字典（包含灵感点、目的点、关键点）
        category_mapping: 特征分类映射字典
        model_name: 使用的模型名称
        output_dir: 输出目录（如果提供，每个帖子处理完立即保存）

    Returns:
        包含 how 解构结果的任务列表
    """
    # 合并三种人设特征（灵感点、关键点、目的点）
    all_features = []

    for feature_type in ["灵感点", "关键点", "目的点"]:
        # 获取该类型的标签特征
        type_features = persona_features_dict.get(feature_type, [])

        # 为每个特征添加层级信息
        for feature in type_features:
            feature_with_level = feature.copy()
            feature_with_level["人设特征层级"] = feature_type
            all_features.append(feature_with_level)

        print(f"人设{feature_type}标签特征数量: {len(type_features)}")

        # 从分类映射中提取该类型的分类特征
        if category_mapping and feature_type in category_mapping:
            type_categories = set()
            for _, feature_data in category_mapping[feature_type].items():
                categories = feature_data.get("所属分类", [])
                type_categories.update(categories)

            # 转换为特征格式并添加层级信息
            for cat in sorted(type_categories):
                all_features.append({
                    "特征名称": cat,
                    "人设特征层级": feature_type
                })

            print(f"人设{feature_type}分类特征数量: {len(type_categories)}")

    print(f"总特征数量（三种类型的标签+分类）: {len(all_features)}")

    # 计算总匹配任务数（灵感点、关键点和目的点）
    total_match_count = 0
    for task in task_list:
        what_result = task.get("what解构结果", {})
        for point_type in ["灵感点", "关键点", "目的点"]:
            point_list = what_result.get(f"{point_type}列表", [])
            for point in point_list:
                feature_count = len(point.get("特征列表", []))
                total_match_count += feature_count * len(all_features)

    print(f"处理灵感点、关键点和目的点特征")
    print(f"总匹配任务数: {total_match_count:,}")
    print()

    # 串行处理所有任务（一个接一个，每个处理完立即保存）
    updated_task_list = []
    for i, task in enumerate(task_list, 1):
        updated_task = await process_single_task(
            task=task,
            task_index=i,
            total_tasks=len(task_list),
            all_persona_features=all_features,
            category_mapping=category_mapping,
            model_name=model_name
        )
        updated_task_list.append(updated_task)

        # 立即保存当前帖子的结果
        if output_dir:
            post_id = updated_task.get("帖子id", "unknown")
            output_file = output_dir / f"{post_id}_how.json"

            with open(output_file, "w", encoding="utf-8") as f:
                json.dump(updated_task, f, ensure_ascii=False, indent=4)

            print(f"  ✓ 已保存: {output_file.name}")

    return updated_task_list


async def main():
    """主函数"""
    # 使用路径配置
    config = PathConfig()

    # 确保输出目录存在
    config.ensure_dirs()

    # 获取路径
    task_list_file = config.task_list_file
    persona_features_file = config.feature_source_mapping_file
    category_mapping_file = config.feature_category_mapping_file
    output_dir = config.how_results_dir

    print(f"账号: {config.account_name}")
    print(f"任务列表文件: {task_list_file}")
    print(f"人设特征文件: {persona_features_file}")
    print(f"分类映射文件: {category_mapping_file}")
    print(f"输出目录: {output_dir}")
    print()

    print(f"读取解构任务列表: {task_list_file}")
    with open(task_list_file, "r", encoding="utf-8") as f:
        task_list_data = json.load(f)

    print(f"读取人设特征: {persona_features_file}")
    with open(persona_features_file, "r", encoding="utf-8") as f:
        persona_features_data = json.load(f)

    print(f"读取特征分类映射: {category_mapping_file}")
    with open(category_mapping_file, "r", encoding="utf-8") as f:
        category_mapping = json.load(f)

    # 获取任务列表
    task_list = task_list_data.get("解构任务列表", [])
    print(f"总任务数: {len(task_list)}")

    # 处理任务列表（每个帖子处理完立即保存）
    updated_task_list = await process_task_list(
        task_list=task_list,
        persona_features_dict=persona_features_data,
        category_mapping=category_mapping,
        model_name=None,  # 使用默认模型
        output_dir=output_dir  # 传递输出目录，启用即时保存
    )

    print("\n完成!")

    # 打印统计信息
    total_inspiration_points = 0
    total_key_points = 0
    total_purpose_points = 0
    total_inspiration_features = 0
    total_key_features = 0
    total_purpose_features = 0

    for task in updated_task_list:
        how_result = task.get("how解构结果", {})

        # 统计灵感点
        inspiration_list = how_result.get("灵感点列表", [])
        total_inspiration_points += len(inspiration_list)
        for point in inspiration_list:
            total_inspiration_features += len(point.get("特征列表", []))

        # 统计关键点
        key_list = how_result.get("关键点列表", [])
        total_key_points += len(key_list)
        for point in key_list:
            total_key_features += len(point.get("特征列表", []))

        # 统计目的点
        purpose_list = how_result.get("目的点列表", [])
        total_purpose_points += len(purpose_list)
        for point in purpose_list:
            total_purpose_features += len(point.get("特征列表", []))

    print(f"\n统计:")
    print(f"  处理的帖子数: {len(updated_task_list)}")
    print(f"  处理的灵感点数: {total_inspiration_points}")
    print(f"  处理的灵感点特征数: {total_inspiration_features}")
    print(f"  处理的关键点数: {total_key_points}")
    print(f"  处理的关键点特征数: {total_key_features}")
    print(f"  处理的目的点数: {total_purpose_points}")
    print(f"  处理的目的点特征数: {total_purpose_features}")


if __name__ == "__main__":
    asyncio.run(main())