proposa-ai/scripts/test_step5.py

"""Step 5 人工验证：对 Step4 实际落盘的章节内容做审核与修复。

输入（均使用已生成中间产物，不重跑 Step1~Step4）：
  - Step2 分析结果 step2_info.pkl（评分项、废标项）
  - Step3 大纲与评分/废标映射 step3_outline_report.json
  - Step4 实际章节 DOCX step4_chapters/

用法（按顺序运行）:
  1. uv run python scripts/test_step1.py；
  2. uv run python scripts/test_step2.py；
  3. uv run python scripts/test_step3.py；
  4. uv run python scripts/test_step4.py；
  5. uv run python scripts/test_step5.py。
"""

import glob
import json
import logging
import os
import pickle
import re
import sys
import xml.etree.ElementTree as ET
from collections import Counter

from _bootstrap import PROJECT_ROOT
from models import BidOutline, Chapter, ChapterType


OUTPUT_DIR = os.environ.get(
    "PROPOSA_STEP5_OUTPUT_DIR",
    os.environ.get("PROPOSA_WORK_DIR", "output/171-上海群众艺术馆"),
)
os.environ.setdefault("BID_LLM_CACHE_DIR", os.path.join(OUTPUT_DIR, ".llm_cache"))
os.environ.setdefault("BID_LLM_CACHE", "0")
STEP2_INFO_FILE = os.environ.get(
    "PROPOSA_STEP2_INFO_FILE",
    os.path.join(OUTPUT_DIR, "step2_info.pkl"),
)
STEP3_REPORT_FILE = os.environ.get(
    "PROPOSA_STEP3_REPORT_FILE",
    os.path.join(OUTPUT_DIR, "step3_outline_report.json"),
)
STEP4_CHAPTERS_DIR = os.environ.get(
    "PROPOSA_STEP4_CHAPTERS_DIR",
    os.path.join(OUTPUT_DIR, "step4_chapters"),
)
STEP5_OUTLINE_FILE = os.environ.get(
    "PROPOSA_STEP5_REVIEWED_OUTLINE_FILE",
    os.path.join(OUTPUT_DIR, "step5_outline.pkl"),
)
STEP5_CHAPTERS_DIR = os.environ.get(
    "PROPOSA_STEP5_CHAPTERS_DIR",
    os.path.join(OUTPUT_DIR, "step5_chapters"),
)

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(message)s",
    datefmt="%H:%M:%S",
    stream=sys.stderr,
)


def _load_step2_info(path):
    if not os.path.isfile(path):
        raise FileNotFoundError(
            f"Step2 分析结果不存在: {path}\n"
            "请先按顺序运行 uv run python scripts/test_step1.py 和 "
            "uv run python scripts/test_step2.py。"
        )
    with open(path, "rb") as file:
        payload = pickle.load(file)
    analysis = payload.get("analysis")
    project_data = payload.get("project_data")
    if analysis is None:
        raise RuntimeError(f"Step2 文件缺少 analysis: {path}")
    return analysis, project_data


def _load_step3_report(path):
    if not os.path.isfile(path):
        raise FileNotFoundError(
            f"Step3 大纲/映射不存在: {path}\n"
            "请先按顺序运行 uv run python scripts/test_step1.py 至 "
            "uv run python scripts/test_step3.py。"
        )
    with open(path, "r", encoding="utf-8") as file:
        return json.load(file)


def _find_manifest(chapters_dir):
    for candidate in glob.glob(
        os.path.join(chapters_dir, "chapters", "*", "manifest.json")
    ):
        return candidate
    return ""


def _load_chapter_texts(chapters_dir):
    """读取 Step4 实际落盘的每个章节 DOCX 文本。"""
    from doc_reader import read_file

    manifest_path = _find_manifest(chapters_dir)
    records = []
    if manifest_path:
        with open(manifest_path, "r", encoding="utf-8") as file:
            manifest = json.load(file)
        records = manifest.get("chapters", [])
    else:
        for path in glob.glob(
            os.path.join(chapters_dir, "chapters", "*", "*.docx")
        ):
            base = os.path.basename(path)
            chapter_id = base.split("_", 1)[0]
            records.append({"id": chapter_id, "artifact_path": path})

    texts = {}
    for record in records:
        chapter_id = str(record.get("id", ""))
        path = record.get("artifact_path", "")
        if not chapter_id or not path or not os.path.isfile(path):
            continue
        content = read_file(path)
        if content:
            texts[chapter_id] = content
    return texts


def _load_step4_records(chapters_dir):
    """读取 Step4 manifest 中的章节文件记录（id/标题/源路径/状态）。"""
    manifest_path = _find_manifest(chapters_dir)
    if not manifest_path:
        records = []
        for path in glob.glob(
            os.path.join(chapters_dir, "chapters", "*", "*.docx")
        ):
            base = os.path.basename(path)
            records.append({
                "id": base.split("_", 1)[0],
                "title": base.rsplit("_", 1)[0],
                "artifact_path": path,
                "status": "complete",
            })
        return records
    with open(manifest_path, "r", encoding="utf-8") as file:
        manifest = json.load(file)
    return [
        {
            "id": str(record.get("id", "")),
            "title": str(record.get("title", "")),
            "artifact_path": str(record.get("artifact_path", "")),
            "status": str(record.get("status", "complete")),
        }
        for record in manifest.get("chapters", [])
        if record.get("artifact_path")
    ]


def _chapter_type_for(final_id: str) -> ChapterType:
    return (
        ChapterType.BUSINESS
        if final_id in {"index", "1", "2"}
        else ChapterType.TECHNICAL
    )


def _build_outline_from_report(report, chapter_texts):
    mappings = list(report.get("heading_mappings", []) or [])
    entries = list(report.get("evaluation_index_entries", []) or [])

    nodes = {}
    for mapping in mappings:
        final_id = str(mapping.get("final_id", ""))
        if not final_id:
            continue
        level = 1 if final_id == "index" else 1 + final_id.count(".")
        nodes[final_id] = Chapter(
            id=final_id,
            title=str(mapping.get("final_title", "") or ""),
            chapter_type=_chapter_type_for(final_id),
            level=level,
            template_original_title=str(mapping.get("template_title", "") or ""),
            template_original_id=str(mapping.get("template_id", "") or ""),
            template_chapter_id=(
                str(mapping.get("template_id", "") or "")
                if level == 1 else ""
            ),
            content_generation_mode=str(
                mapping.get("content_generation_mode", "") or ""
            ),
            direct_scoring_bindings=list(mapping.get("scoring_bindings", []) or []),
            direct_rejection_bindings=list(mapping.get("rejection_bindings", []) or []),
            structure_locked=bool(
                mapping.get("scoring_bindings") or mapping.get("rejection_bindings")
            ),
        )

    for entry in entries:
        heading_id = str(entry.get("final_heading_id", "") or "")
        node = nodes.get(heading_id)
        if node is None:
            continue
        if entry.get("entry_type") == "scoring":
            criterion_id = str(entry.get("criterion_id", "") or "")
            if criterion_id and criterion_id not in node.related_criteria:
                node.related_criteria.append(criterion_id)
            if criterion_id and criterion_id not in node.direct_scoring_criteria:
                node.direct_scoring_criteria.append(criterion_id)
        elif entry.get("entry_type") == "rejection":
            source_id = str(entry.get("source_id", "") or "")
            if source_id and source_id not in node.related_rejections:
                node.related_rejections.append(source_id)

    roots = []
    for mapping in mappings:
        final_id = str(mapping.get("final_id", ""))
        parent_id = str(mapping.get("parent_final_id", "") or "")
        node = nodes.get(final_id)
        if node is None:
            continue
        parent = nodes.get(parent_id) if parent_id else None
        if parent is not None:
            node.chapter_type = parent.chapter_type
            if node not in parent.children:
                parent.children.append(node)
        else:
            roots.append(node)

    for chapter in roots:
        chapter.generated_content = chapter_texts.get(chapter.id, "")

    outline = BidOutline(
        project_name=str(report.get("project", "") or "投标项目"),
        chapters=roots,
    )
    outline.heading_mappings = list(mappings)
    outline.evaluation_index_entries = list(entries)
    return outline


def _safe_project_name(name):
    return (
        re.sub(r'[<>:"/\\|?*]+', "_", name or "投标项目").strip(" ._")
        or "投标项目"
    )


def _normalize_docx_text(text):
    return re.sub(r"[\s\u3000]+", "", text or "")


_EMPTY_LABEL_RE = re.compile(
    r"^(项目名称|项目编号|招标编号|招标项目编号|包号|包件号|包件名称|包名|"
    r"服务内容|服务要求|服务期限)[：:]\s*$"
)
_PROJECT_LABEL_RE = re.compile(
    r"^(项目名称|项目编号|招标编号|招标项目编号|包号|包件号|包件名称|包名|"
    r"服务内容|服务要求|服务期限)[：:]\s*(.*)$"
)
_TITLE_PREFIX_RE = re.compile(
    r"^(第[一二三四五六七八九十百\d]+章|"
    r"[一二三四五六七八九十百]+、|"
    r"[（(][一二三四五六七八九十百\d]+[）)]|"
    r"\d+[.、．]|"
    r"[（(]\d+[）)]|"
    r"\d+\)|[a-zA-Z][.、．)])"
)
_PROTECTED_SIGNATURE_RE = re.compile(
    r"^\s*(投标人授权代表签字|投标人（公章）|投标人\(公章\)|日期|法定代表人|"
    r"授权代表|签署人)[：:（(]"
)
_PACKAGE_LABELS = {"包号", "包件", "包件号", "包件名称", "包名"}
_PROJECT_LABEL_ALIASES = {
    "招标编号": "项目编号",
    "招标项目编号": "项目编号",
    "包件号": "包号",
    "包件名称": "包号",
    "包名": "包号",
}


def _heading_level(paragraph):
    style = getattr(getattr(paragraph, "style", None), "name", "") or ""
    match = re.match(r"(?:Heading|标题)\s*([1-4])", style, re.I)
    return int(match.group(1)) if match else 0


def _clear_paragraph_text(paragraph):
    for run in list(paragraph.runs):
        run.text = ""


def _ensure_body_indent(paragraph):
    """给正文段落写入首行缩进 2 字符（Word firstLineChars=200）。"""
    W_NS = "http://schemas.openxmlformats.org/wordprocessingml/2006/main"
    p_pr = paragraph.find(f"{{{W_NS}}}pPr")
    if p_pr is None:
        p_pr = paragraph.makeelement(f"{{{W_NS}}}pPr", {})
        paragraph.insert(0, p_pr)
    ind = p_pr.find(f"{{{W_NS}}}ind")
    if ind is None:
        ind = p_pr.makeelement(f"{{{W_NS}}}ind", {})
        p_pr.append(ind)
    ind.set(f"{{{W_NS}}}firstLineChars", "200")
    ind.set(f"{{{W_NS}}}firstLine", "480")


def _apply_body_format(paragraph, W_NS):
    """给正文段落应用宋体小四、黑色、1.5 倍行距与首行缩进。"""
    p_pr = paragraph.find(f"{{{W_NS}}}pPr")
    if p_pr is None:
        p_pr = ET.SubElement(paragraph, f"{{{W_NS}}}pPr")

    spacing = p_pr.find(f"{{{W_NS}}}spacing")
    if spacing is None:
        spacing = ET.SubElement(p_pr, f"{{{W_NS}}}spacing")
    spacing.set(f"{{{W_NS}}}line", "360")
    spacing.set(f"{{{W_NS}}}lineRule", "auto")
    spacing.set(f"{{{W_NS}}}before", "0")
    spacing.set(f"{{{W_NS}}}after", "0")

    ind = p_pr.find(f"{{{W_NS}}}ind")
    if ind is None:
        ind = ET.SubElement(p_pr, f"{{{W_NS}}}ind")
    ind.set(f"{{{W_NS}}}firstLineChars", "200")
    ind.set(f"{{{W_NS}}}firstLine", "480")

    for run in paragraph.findall(f"{{{W_NS}}}r"):
        r_pr = run.find(f"{{{W_NS}}}rPr")
        if r_pr is None:
            r_pr = ET.SubElement(run, f"{{{W_NS}}}rPr")
        r_fonts = r_pr.find(f"{{{W_NS}}}rFonts")
        if r_fonts is None:
            r_fonts = ET.SubElement(r_pr, f"{{{W_NS}}}rFonts")
        r_fonts.set(f"{{{W_NS}}}ascii", "Times New Roman")
        r_fonts.set(f"{{{W_NS}}}hAnsi", "Times New Roman")
        r_fonts.set(f"{{{W_NS}}}eastAsia", "宋体")

        for tag, value in (("sz", "24"), ("szCs", "24")):
            node = r_pr.find(f"{{{W_NS}}}{tag}")
            if node is None:
                node = ET.SubElement(r_pr, f"{{{W_NS}}}{tag}")
            node.set(f"{{{W_NS}}}val", value)

        color = r_pr.find(f"{{{W_NS}}}color")
        if color is None:
            color = ET.SubElement(r_pr, f"{{{W_NS}}}color")
        color.set(f"{{{W_NS}}}val", "000000")


def _clean_chapter_docx(doc, has_packages):
    """直接在 step4 章节 DOCX 上删除包号行和重复正文段落。

    has_packages 由 Step2 的 LLM 提取结果决定；只有确认项目无实际包号信息时，
    才删除章节中的包号/包件文字。
    """
    _PACKAGE_LABELS = {"包号", "包件", "包件号", "包件名称", "包名"}
    seen = set()
    for paragraph in doc.paragraphs:
        text = (paragraph.text or "").strip()
        if not text:
            continue
        level = _heading_level(paragraph)
        if level == 0:
            if not has_packages and package_re.search(text):
                _clear_paragraph_text(paragraph)
                continue
            key = _normalize_docx_text(text)
            if len(key) >= 12 and key in seen:
                _clear_paragraph_text(paragraph)
                continue
            if len(key) >= 12:
                seen.add(key)

    if has_packages:
        return
    for table in doc.tables:
        for row in table.rows:
            for cell in row.cells:
                for paragraph in cell.paragraphs:
                    if package_re.search(paragraph.text or ""):
                        _clear_paragraph_text(paragraph)


def _find_heading_paragraph(doc, node):
    candidates = [
        _normalize_docx_text(value)
        for value in (node.title, getattr(node, "template_original_title", ""))
        if value
    ]
    for paragraph in doc.paragraphs:
        if _heading_level(paragraph) and _normalize_docx_text(paragraph.text) in candidates:
            return paragraph
    return None


def _insert_node_supplement(doc, node):
    """把 Step5 生成的补充正文插入到对应绑定标题之后，不改动原生内容块。"""
    content = (getattr(node, "supplement_content", "") or "").strip()
    if not content:
        content = (getattr(node, "generated_content", "") or "").strip()
    if not content:
        return 0

    anchor = _find_heading_paragraph(doc, node)
    if anchor is None:
        return 0

    existing = {
        _normalize_docx_text(paragraph.text)
        for paragraph in doc.paragraphs
    }
    lines = []
    for raw in content.splitlines():
        line = raw.strip()
        key = _normalize_docx_text(line)
        if key and key not in existing:
            lines.append(line)
            existing.add(key)
    if not lines:
        return 0

    node_level = int(getattr(node, "level", 2) or 2)
    boundary = None
    anchor_index = None
    for index, paragraph in enumerate(doc.paragraphs):
        if paragraph is anchor:
            anchor_index = index
            continue
        if anchor_index is not None:
            level = _heading_level(paragraph)
            if level and level <= node_level:
                boundary = paragraph
                break

    if boundary is not None:
        for line in reversed(lines):
            boundary.insert_paragraph_before(line)
    else:
        for line in lines:
            doc.add_paragraph(line)
    return len(lines)


def _has_package_info(analysis):
    """依据 Step2 LLM 提取结果判断项目是否真的存在包号/包件信息。"""
    from step5_reviewing.policies import analysis_has_package_info
    return analysis_has_package_info(analysis)


def _safe_copy_and_clean_chapter(source, target, has_packages, remove_keys=None, fills=None):
    """用标准库重写 document.xml，避免 python-docx/lxml 在大型章节上 0xC0000005。"""
    import xml.etree.ElementTree as ET
    import zipfile

    W_NS = "http://schemas.openxmlformats.org/wordprocessingml/2006/main"
    package_re = re.compile(r"(包号|包件)")
    document_name = "word/document.xml"
    remove_keys = set(remove_keys or [])
    fills = list(fills or [])

    with zipfile.ZipFile(source, "r") as source_zip:
        names = set(source_zip.namelist())
        document_out = None
        if document_name in names:
            root = ET.fromstring(source_zip.read(document_name))
            body = None
            for element in root.iter():
                if element.tag == f"{{{W_NS}}}body":
                    body = element
                    break
            if body is not None:
                seen = set()
                seen_filled_labels = set()
                for paragraph in list(body):
                    if paragraph.tag != f"{{{W_NS}}}p":
                        continue
                    text_nodes = [
                        node for node in paragraph.iter()
                        if node.tag == f"{{{W_NS}}}t"
                    ]
                    text = "".join(node.text or "" for node in text_nodes)
                    if not text.strip():
                        continue
                    if _PROTECTED_SIGNATURE_RE.match(text.strip()):
                        continue
                    key = _normalize_docx_text(text)
                    p_pr = paragraph.find(f"{{{W_NS}}}pPr")
                    style_val = ""
                    if p_pr is not None:
                        p_style = p_pr.find(f"{{{W_NS}}}pStyle")
                        if p_style is not None:
                            style_val = p_style.get(f"{{{W_NS}}}val", "")
                    is_title = bool(
                        _TITLE_PREFIX_RE.match(text.strip())
                        or re.match(r"(?:Heading|标题)\s*[1-4]", style_val, re.I)
                    )
                    if is_title:
                        seen = set()
                        seen_filled_labels = set()
                        continue
                    if key in remove_keys:
                        for node in text_nodes:
                            node.text = ""
                        continue
                    label_match = _PROJECT_LABEL_RE.match(text.strip())
                    if label_match:
                        raw_label = label_match.group(1).strip()
                        label_value = label_match.group(2).strip()
                        label = _PROJECT_LABEL_ALIASES.get(raw_label, raw_label)
                        if label_value:
                            seen_filled_labels.add(label)
                            continue
                        if label in seen_filled_labels:
                            for node in text_nodes:
                                node.text = ""
                            continue
                        if raw_label in _PACKAGE_LABELS and not has_packages:
                            for node in text_nodes:
                                node.text = ""
                            continue
                        continue
                    if len(key) >= 12 and key in seen:
                        for node in text_nodes:
                            node.text = ""
                        continue
                    if len(key) >= 12:
                        seen.add(key)
                    for fill in fills:
                        label = str(fill.get("label", "")).strip()
                        value = str(fill.get("value", "")).strip()
                        if not label or not value:
                            continue
                        head = re.split(r"[:：]", text.strip(), maxsplit=1)[0].strip()
                        if _normalize_docx_text(head) == _normalize_docx_text(label):
                            for node in text_nodes:
                                node.text = ""
                            if text_nodes:
                                text_nodes[0].text = f"{label}：{value}"
                            break
                    _apply_body_format(paragraph, W_NS)
            document_out = ET.tostring(
                root, encoding="utf-8", xml_declaration=True
            )

    with zipfile.ZipFile(source, "r") as source_zip:
        with zipfile.ZipFile(
            target, "w", compression=zipfile.ZIP_STORED
        ) as target_zip:
            for info in source_zip.infolist():
                data = (
                    document_out
                    if info.filename == document_name and document_out is not None
                    else source_zip.read(info.filename)
                )
                target_zip.writestr(info, data)


def _collect_redundant_actions(report, chapter_id):
    remove_keys = set()
    fills = []
    if report is None:
        return remove_keys, fills
    for issue in getattr(report, "issues", []) or []:
        if issue.issue_type != "redundant_info" or issue.chapter_id != chapter_id:
            continue
        try:
            payload = json.loads(issue.suggestion or "{}")
        except Exception:
            continue
        for item in payload.get("removals", []) or []:
            if isinstance(item, dict) and item.get("text"):
                raw = str(item["text"]).strip()
                if _EMPTY_LABEL_RE.match(raw):
                    remove_keys.add(_normalize_docx_text(raw))
        for item in payload.get("fills", []) or []:
            if isinstance(item, dict):
                fills.append(item)
    return remove_keys, fills


def _write_step5_chapters_from_step4(records, outline, analysis, base_dir, report=None):
    """以 step4 章节 DOCX 为底稿，在其上应用 Step5 修复并写出 step5_chapters。"""
    chapter_dir = os.path.join(
        base_dir, "chapters", _safe_project_name(outline.project_name)
    )
    os.makedirs(chapter_dir, exist_ok=True)
    has_packages = _has_package_info(analysis)
    manifest = []

    for record in records:
        source = record.get("artifact_path", "")
        chapter_id = str(record.get("id", ""))
        if not source or not os.path.isfile(source):
            continue
        target = os.path.join(chapter_dir, os.path.basename(source))
        remove_keys, fills = _collect_redundant_actions(report, chapter_id)
        _safe_copy_and_clean_chapter(
            source, target, has_packages, remove_keys, fills
        )
        manifest.append({
            "id": chapter_id,
            "title": record.get("title", ""),
            "artifact_path": os.path.abspath(target),
            "status": record.get("status", "complete"),
        })

    manifest_path = os.path.join(chapter_dir, "manifest.json")
    with open(manifest_path, "w", encoding="utf-8") as file:
        json.dump(
            {
                "project_name": outline.project_name,
                "expected_count": len(records),
                "completed_count": len(manifest),
                "chapters": manifest,
            },
            file,
            ensure_ascii=False,
            indent=2,
        )
    return [record["artifact_path"] for record in manifest]


def main():
    analysis, project_data = _load_step2_info(STEP2_INFO_FILE)
    report = _load_step3_report(STEP3_REPORT_FILE)
    step4_records = _load_step4_records(STEP4_CHAPTERS_DIR)
    chapter_texts = _load_chapter_texts(STEP4_CHAPTERS_DIR)
    outline = _build_outline_from_report(report, chapter_texts)

    print("=" * 60)
    print("Step 5a: 加载实际章节内容、Step3 大纲与评分/废标映射")
    print("=" * 60)
    print(f"章节数: {len(outline.chapters)}")
    print(f"评分项: {len(analysis.scoring_criteria)}；废标项: {len(analysis.rejection_items)}")
    print(f"Step3 评分/废标索引映射: {len(outline.evaluation_index_entries)} 条")
    print(f"Step3 标题关系映射: {len(outline.heading_mappings)} 条")
    print(f"已读取实际章节文件: {len(chapter_texts)} 个")
    for chapter in outline.chapters:
        print(
            f"  {chapter.id} {chapter.title}: 实际正文 "
            f"{len(chapter.generated_content or '')} 字符"
        )

    print()
    print("=" * 60)
    print("Step 5b: 内容审核")
    print("=" * 60)

    from step5_reviewing import review_content, auto_fix_issues

    review_report, outline = review_content(outline, analysis)

    print()
    print("=== 审核报告 ===")
    print(f"审核结果: {'[PASS] 通过' if review_report.passed else '[FAIL] 未通过'}")
    print(f"总字数:   {review_report.total_word_count:,}")
    print(f"问题总数: {len(review_report.issues)}")
    print(
        "问题类型分布:",
        dict(Counter(i.issue_type for i in review_report.issues)),
    )

    errors = [i for i in review_report.issues if i.severity == "error"]
    warnings = [i for i in review_report.issues if i.severity == "warning"]
    print(f"  错误: {len(errors)}")
    for issue in errors:
        print(f"    [ERR] [{issue.chapter_id}] {issue.description[:120]}")
        if issue.suggestion:
            print(f"        建议: {issue.suggestion[:120]}")
    print(f"  警告: {len(warnings)}")
    for issue in warnings[:8]:
        print(f"    [WARN] [{issue.chapter_id}] {issue.description[:120]}")

    if not review_report.passed:
        print()
        print("=" * 60)
        print("Step 5c: 自动修复")
        print("=" * 60)
        outline = auto_fix_issues(
            outline,
            analysis,
            review_report,
            project_data=project_data,
        )
        report2, outline = review_content(outline, analysis)
        print(f"\n修复后审核: {'[PASS] 通过' if report2.passed else '[FAIL] 仍有问题'}")
        print(f"修复后字数: {report2.total_word_count:,}")
        print(f"剩余问题:   {len(report2.issues)}")

    print()
    print("=" * 60)
    print("Step 5d: 在 step4 各章 DOCX 基础上写出 step5_chapters")
    print("=" * 60)
    artifact_paths = _write_step5_chapters_from_step4(
        step4_records,
        outline,
        analysis,
        STEP5_CHAPTERS_DIR,
        report=review_report,
    )
    print(f"已写出 {len(artifact_paths)} 个章节 DOCX 到: {STEP5_CHAPTERS_DIR}")

    os.makedirs(os.path.dirname(STEP5_OUTLINE_FILE), exist_ok=True)
    with open(STEP5_OUTLINE_FILE, "wb") as file:
        pickle.dump(outline, file)
    print(f"\nStep5 修复后大纲已保存: {STEP5_OUTLINE_FILE}")
    print("Step 5 测试完成 [OK]")


if __name__ == "__main__":
    main()