video_decode/test/agents/test_script_orthogonal_with_real_data.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
使用真实数据测试脚本正交分析Agent
"""

import sys
import os
import json
from pathlib import Path
from dotenv import load_dotenv

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

# 加载环境变量
load_dotenv(project_root / ".env")

from src.components.agents.script_orthogonal_analysis_agent import ScriptOrthogonalAnalysisAgent


def load_script_result(file_path: str) -> dict:
    """加载脚本解析结果文件

    Args:
        file_path: JSON文件路径

    Returns:
        解析后的JSON数据
    """
    with open(file_path, 'r', encoding='utf-8') as f:
        return json.load(f)


def test_with_real_data():
    """使用真实数据测试脚本正交分析Agent"""

    print("=" * 100)
    print("使用真实数据测试脚本正交分析Agent")
    print("=" * 100)

    # 加载真实数据
    script_result_path = project_root / "examples/阿里多多酱/output/script_result_20251118_152626.json"
    print(f"\n📁 加载脚本解析结果: {script_result_path}")

    script_result = load_script_result(str(script_result_path))

    # 提取所需数据
    topic_description = script_result.get("选题描述", {})
    content_weight = script_result.get("图文权重", {})
    script_understanding = script_result.get("脚本理解", {})

    # 构建state
    state = {
        "text": {
            "title": "当代年轻人对食物的双标日常",
            "body": "#讨好型水果[话题]#"
        },
        "images": script_understanding.get("图片列表", []),
        "topic_selection_understanding": topic_description,
        "content_weight": content_weight,
        "script_sections": {
            "内容品类": script_understanding.get("内容品类", ""),
            "段落列表": script_understanding.get("段落列表", [])
        },
        "script_elements": {
            "元素列表": script_understanding.get("元素列表", [])
        }
    }

    print(f"\n✓ 数据加载成功")
    print(f"  - 选题主题: {topic_description.get('主题', 'N/A')}")
    print(f"  - 内容品类: {script_understanding.get('内容品类', 'N/A')}")
    print(f"  - 段落数量: {len(script_understanding.get('段落列表', []))}")
    print(f"  - 元素数量: {len(script_understanding.get('元素列表', []))}")
    print(f"  - 图片数量: {len(script_understanding.get('图片列表', []))}")

    # 初始化Agent
    print("\n🤖 初始化ScriptOrthogonalAnalysisAgent...")
    agent = ScriptOrthogonalAnalysisAgent()
    print("  ✓ Agent初始化成功")

    # 执行正交分析
    print("\n⚙️  开始执行正交分析...")
    print("-" * 100)
    result = agent.process(state)
    print("-" * 100)

    # 输出结果
    print("\n" + "=" * 100)
    print("📊 正交分析结果")
    print("=" * 100)

    orthogonal_matrix = result.get("正交矩阵", [])
    element_type_list = result.get("元素类型列表", [])

    print(f"\n✓ 正交矩阵生成成功")
    print(f"  - 段落行数: {len(orthogonal_matrix)}")
    print(f"  - 元素类型列数: {len(element_type_list)}")

    print(f"\n📋 元素类型列表 (共{len(element_type_list)}个):")
    for idx, element_type in enumerate(element_type_list, 1):
        print(f"  {idx}. {element_type}")

    print(f"\n📋 正交矩阵预览 (前3个段落):")
    for idx, row in enumerate(orthogonal_matrix[:3], 1):
        print(f"\n  段落 {idx}: {row['段落']}")
        print(f"  内容范围: {row['内容范围']}")
        print(f"  元素类型分析:")
        for element_type, analysis in row.get('元素类型分析', {}).items():
            if analysis:  # 只显示非空的分析
                print(f"    - {element_type}: {analysis[:80]}{'...' if len(analysis) > 80 else ''}")

    # 保存结果
    output_dir = project_root / "examples/阿里多多酱/output"
    output_path = output_dir / "orthogonal_analysis_result.json"

    print(f"\n💾 保存结果到: {output_path}")
    with open(output_path, 'w', encoding='utf-8') as f:
        json.dump(result, f, ensure_ascii=False, indent=2)

    print(f"  ✓ 结果保存成功")

    # 生成表格形式的输出（Markdown格式）
    markdown_output_path = output_dir / "orthogonal_analysis_table.md"
    print(f"\n📝 生成Markdown表格: {markdown_output_path}")

    with open(markdown_output_path, 'w', encoding='utf-8') as f:
        f.write("# 脚本正交分析矩阵\n\n")
        f.write(f"**选题主题**: {topic_description.get('主题', 'N/A')}\n\n")
        f.write(f"**内容品类**: {script_understanding.get('内容品类', 'N/A')}\n\n")

        # 生成表格
        f.write("## 正交矩阵表格\n\n")

        # 表头
        header = "| 段落 |"
        separator = "|------|"
        for element_type in element_type_list:
            header += f" {element_type} |"
            separator += "------|"
        f.write(header + "\n")
        f.write(separator + "\n")

        # 表格内容
        for row in orthogonal_matrix:
            line = f"| {row['段落']} |"
            for element_type in element_type_list:
                analysis = row.get('元素类型分析', {}).get(element_type, "")
                # 处理换行和特殊字符
                analysis = analysis.replace("\n", " ").replace("|", "\\|")
                line += f" {analysis} |"
            f.write(line + "\n")

        # 添加详细说明
        f.write("\n## 段落内容范围详情\n\n")
        for idx, row in enumerate(orthogonal_matrix, 1):
            f.write(f"### {idx}. {row['段落']}\n\n")
            f.write(f"**内容范围**:\n")
            for content in row['内容范围']:
                f.write(f"- {content}\n")
            f.write("\n")

    print(f"  ✓ Markdown表格生成成功")

    # 测试总结
    print("\n" + "=" * 100)
    print("✅ 测试完成总结")
    print("=" * 100)
    print(f"  ✓ 成功加载真实数据文件")
    print(f"  ✓ 成功提取 {len(orthogonal_matrix)} 个段落")
    print(f"  ✓ 成功提取 {len(element_type_list)} 个元素类型")
    print(f"  ✓ 成功生成正交分析矩阵")
    print(f"  ✓ 结果已保存为JSON和Markdown格式")
    print(f"\n  📂 输出文件:")
    print(f"     - JSON: {output_path}")
    print(f"     - Markdown: {markdown_output_path}")

    print("\n" + "🎉" * 50)
    print("测试成功完成!")
    print("🎉" * 50 + "\n")

    return result


if __name__ == "__main__":
    try:
        test_with_real_data()
    except Exception as e:
        print(f"\n❌ 测试失败: {e}\n")
        import traceback
        traceback.print_exc()
        exit(1)