proposa-ai/scripts/test_step3.py

"""
Step 3 测试：投书目录生成

读取 Step2 保存的 step2_info.pkl，再调用 generate_outline_safe 生成目录，
并把最终目录保存为 step3_outline.pkl 供 test_step4.py 复用。
用法: uv run python scripts/test_step3.py
"""

import logging
import os
import pickle
import sys

from _bootstrap import PROJECT_ROOT

# ---------- 测试数据（环境变量可覆盖） ----------
TEMPLATE_PATH = os.environ.get(
    "PROPOSA_TEMPLATE_PATH",
    "src/templates/申勤投标模板.docx",
)
REFERENCE_BID = os.environ.get(
    "PROPOSA_REFERENCE_BID",
    "test_data/171-上海群众艺术馆/参考投标文件/物业管理费项目投标文件.docx",
)
OUTPUT_DIR = os.environ.get(
    "PROPOSA_STEP3_OUTPUT_DIR",
    os.environ.get("PROPOSA_WORK_DIR", "output/171-上海群众艺术馆"),
)
STEP2_INFO_FILE = os.environ.get(
    "PROPOSA_STEP2_INFO_FILE",
    os.path.join(OUTPUT_DIR, "step2_info.pkl"),
)
STEP3_OUTLINE_FILE = os.environ.get(
    "PROPOSA_STEP3_OUTLINE_FILE",
    os.path.join(OUTPUT_DIR, "step3_outline.pkl"),
)
os.environ["BID_OUTLINE_CACHE_DIR"] = os.path.join(OUTPUT_DIR, ".outline_cache")
os.environ["BID_LLM_CACHE_DIR"] = os.path.join(OUTPUT_DIR, ".llm_cache")

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

# ---------- Step 2: 复用已落盘的分析结果 ----------
print("=" * 60)
print("Step 2: 复用已落盘的分析结果（Step 3 的前置）")
print("=" * 60)

if not os.path.isfile(STEP2_INFO_FILE):
    raise FileNotFoundError(
        f"Step2 分析结果不存在: {STEP2_INFO_FILE}\n"
        "请先运行 uv run python scripts/test_step2.py。"
    )
with open(STEP2_INFO_FILE, "rb") as file:
    step2_payload = pickle.load(file)
analysis = step2_payload["analysis"]
pd = step2_payload["project_data"]
print(
    f"复用 Step2 分析: {len(analysis.scoring_criteria)} 评分项, "
    f"{len(analysis.rejection_items)} 废标项"
)

# ---------- Step 3: 生成目录 ----------
print()
print("=" * 60)
print("Step 3: 生成投书目录")
print("=" * 60)

from step3_outlining import generate_outline_safe
from models import ChapterType

outline = generate_outline_safe(
    analysis,
    pd,
    template_path=TEMPLATE_PATH,
    reference_bid_path=REFERENCE_BID,
)

# generate_outline_safe 已在返回前写出报告并执行硬门禁；失败时不会进入这里。
os.makedirs(OUTPUT_DIR, exist_ok=True)
with open(STEP3_OUTLINE_FILE, "wb") as file:
    pickle.dump(outline, file)
report_md = os.path.join(OUTPUT_DIR, "step3_outline_report.md")
report_json = os.path.join(OUTPUT_DIR, "step3_outline_report.json")
print(f"大纲对照报告: {report_md}")
print(f"机器可读映射: {report_json}")
print(f"Step3 最终目录: {STEP3_OUTLINE_FILE}")

# ---------- 打印结果 ----------
print()
print("=== 目录结构 ===")
print(f"项目:       {outline.project_name}")
print(f"总目标字数: {outline.total_word_count_target:,}")
print()

biz_chs = [c for c in outline.chapters if c.chapter_type == ChapterType.BUSINESS]
tech_chs = [c for c in outline.chapters if c.chapter_type == ChapterType.TECHNICAL]
appendix_chs = [c for c in outline.chapters if c.chapter_type == ChapterType.APPENDIX]

if biz_chs:
    print("【商务部分】")
    for ch in biz_chs:
        children_info = f" ({len(ch.children)} 节)" if ch.children else ""
        print(f"  第{ch.id}章 {ch.title} — 目标 {ch.word_count_target:,} 字{children_info}")
        for child in ch.children:
            print(f"    {child.id} {child.title} — {child.word_count_target:,} 字")

if tech_chs:
    print()
    print(f"【技术部分】")
    for ch in tech_chs:
        children_info = f" ({len(ch.children)} 节)" if ch.children else ""
        print(f"  第{ch.id}章 {ch.title} — 目标 {ch.word_count_target:,} 字{children_info}")
        for child in ch.children:
            print(f"    {child.id} {child.title} — {child.word_count_target:,} 字")

if appendix_chs:
    print()
    print("【附件】")
    for ch in appendix_chs:
        print(f"  {ch.id} {ch.title}")

total = sum(ch.word_count_target for ch in outline.chapters)
print(f"\n章节总字数目标: {total:,} (配置最低: 170,000)")

print()
print("Step 3 测试完成 [OK]")