# -*- coding: utf-8 -*-
"""四份交付文档 · 跨文档一致性自检。仅读取，不改文件。"""
import re
from pathlib import Path

D = Path(r"D:\桌面\金融\客服agent")
OUT = Path(r"D:\桌面\金融\_consistency.txt")

FILES = {
    "需求": D / "D2.2-客服Agent需求文档.html",
    "计划": D / "D2.3-客服Agent开发计划.html",
    "Todo": D / "D2.1-客服Agent执行Todolist.md",
    "知识库": D / "D2.4-客服Agent知识库设计方案.html",
}

texts = {}
lines = []

lines.append("== 一、文件基本情况 ==")
for name, p in FILES.items():
    if p.exists():
        t = p.read_text(encoding="utf-8")
        texts[name] = t
        lines.append(f"  {name:4s} {p.name}  {t.count(chr(10)) + 1} 行 / {len(t)} 字符")
    else:
        lines.append(f"  [MISSING] {p}")

lines.append("")
lines.append("== 二、关键事实在各文档中的出现次数（0 = 遗漏）==")
FACTS = [
    ("功能需求 48 条", r"48 条"),
    ("FR-CS-048", r"FR-CS-048"),
    ("非功能需求 21 条", r"21 条"),
    ("NFR-CS-021", r"NFR-CS-021"),
    ("51 项", r"51 项"),
    ("7 批次 / 7 个批次", r"7 个?批次"),
    ("12 步", r"12 步"),
    ("6 文件 / 六文件", r"(六文件|6 个?文件)"),
    ("组 2 四文件", r"组 2|组2"),
    ("三集合", r"三集合"),
    ("阈值 0.75", r"0\.75"),
    ("阈值 0.70", r"0\.70"),
    ("阈值 0.65", r"0\.65"),
    ("over-fetch", r"over-fetch"),
    ("不变量", r"不变量"),
    ("actor.py", r"actor\.py"),
    ("knowledge_tier.py", r"knowledge_tier\.py"),
    ("三条红线", r"三条红线"),
    ("subject_type", r"subject_type"),
    ("零 DDL", r"零 ?DDL"),
]
hdr = "  " + "事实".ljust(26) + "".join(n.center(8) for n in FILES)
lines.append(hdr)
for label, pat in FACTS:
    row = "  " + label.ljust(26)
    for name in FILES:
        t = texts.get(name, "")
        row += str(len(re.findall(pat, t))).center(8)
    lines.append(row)

lines.append("")
lines.append("== 三、投顾残留检查（应全部出现在「已清除/作废」语境）==")
for name, t in texts.items():
    if name == "知识库":
        continue
    hits = [m.start() for m in re.finditer("投顾", t)]
    bad = []
    for h in hits:
        ctx = t[max(0, h - 60): h + 60].replace("\n", " ")
        if not re.search(r"已(整体)?清除|作废|不再|已删除|清除后|已移除|已下架|原|投顾模块已", ctx):
            bad.append(ctx)
    lines.append(f"  {name}: 命中 {len(hits)} 处，其中疑似「仍在依赖」{len(bad)} 处")
    for b in bad[:5]:
        lines.append(f"      ⚠ …{b}…")

lines.append("")
lines.append("== 四、HTML 锚点自检（TOC href vs 正文 id）==")
for name in ("需求", "计划", "知识库"):
    t = texts.get(name, "")
    hrefs = set(re.findall(r'class="toc-item" href="#([^"]+)"', t))
    ids = set(re.findall(r'<h[1-4] id="([^"]+)"', t))
    missing = sorted(hrefs - ids)
    lines.append(f"  {name}: TOC {len(hrefs)} 条 / 正文 id {len(ids)} 个 / 失效 {len(missing)}")
    for m in missing:
        lines.append(f"      ⚠ 失效锚点: {m}")

lines.append("")
lines.append("== 五、四文档交叉引用检查 ==")
EXPECT = {
    "需求": ["客服Agent开发计划", "客服Agent执行Todolist", "客服Agent知识库设计方案"],
    "计划": ["客服Agent需求文档", "客服Agent执行Todolist", "客服Agent知识库设计方案"],
    "知识库": ["客服Agent需求文档"],
}
for name, refs in EXPECT.items():
    t = texts.get(name, "")
    for r in refs:
        lines.append(f"  {name} → {r}: {'✅' if r in t else '❌ 缺'}")

OUT.write_text("\n".join(lines), encoding="utf-8")