#!/usr/bin/env python3 import hashlib import json import re import sys from datetime import datetime, timezone from pathlib import Path PROJECT = Path(__file__).resolve().parents[1] STORY = PROJECT / "story" OUTLINE = PROJECT / "script" / "outline.json" IDENTITY = PROJECT / "script" / "identity-registry.json" REPORT = STORY / "promotion-report.json" WORD_RE = re.compile(r"[\wÀ-ỹĐđ]+", re.UNICODE) HEADING_RE = re.compile(r"^\s*(?:chương|phần|cảnh|chapter|part|scene)\s+(?:\d+|[ivxlcdm]+)\s*[:.\-]?", re.I) PRODUCTION_RE = re.compile(r"\[(?:sfx|music|nhạc|âm thanh|pause|transition)\]", re.I) def sha(text): return hashlib.sha256(text.encode("utf-8")).hexdigest() def words(text): return WORD_RE.findall(text) def exact_join(texts): return "\n\n".join(text.strip() for text in texts).strip() + "\n" def identity_hits(text, registry): scrubbed = text fragments = set() for entity in registry["characters"]: allowed = sorted(set(entity.get("allowed_narrative_names", []) + entity.get("allowed_title_names", []) + entity.get("role_address", []) + [entity["full_name"]]), key=len, reverse=True) for value in allowed: scrubbed = re.sub(rf"(?= 50: return hits return hits def reveal_hits(chapters): checks = [ (8, "forged_waivers", r"bảy (?:giấy|chữ ký).{0,180}?(?:giả|giả mạo)|giấy từ bỏ.{0,180}?(?:giả|giả mạo)"), (5, "seven_compartments_function", r"bảy chìa.{0,180}?mở bảy ngăn"), (8, "transfer_condition", r"chỉ được bán khi.{0,220}?(?:đồng thuận|quỹ hoàn trả)|điều kiện chuyển quyền.{0,220}?(?:đồng thuận|hoàn trả)"), (9, "inspection_false_signature", r"Tử Dương.{0,240}?(?:ký|xác nhận).{0,180}?(?:chưa đến|không đến).{0,120}?(?:hiện trường|căn nhà)"), ] hits = [] for before_count, key, pattern in checks: early = "\n".join(chapters[:before_count]) for match in re.finditer(pattern, early, re.I | re.S): context = early[max(0, match.start()-100):match.end()+100] if key == "forged_waivers" and re.search(r"(?:chưa|không)\s+(?:đủ|thể).{0,80}?(?:chứng minh|xác định)|(?:nghi|nếu).{0,100}$", context[:100], re.I | re.S): continue hits.append({"reveal": key, "before_chapter": before_count + 1, "excerpt": context}) break return hits def main(): issues = [] outline = json.loads(OUTLINE.read_text(encoding="utf-8")) registry = json.loads(IDENTITY.read_text(encoding="utf-8")) expected = len(outline["chapters"]) paths = [STORY / "chapters" / f"{i:02d}.txt" for i in range(1, expected + 1)] chapter_texts, rows = [], [] for i, path in enumerate(paths, 1): if not path.exists(): issues.append(f"missing_chapter:{i:02d}") chapter_texts.append("") continue text = path.read_text(encoding="utf-8") chapter_texts.append(text) rows.append({"chapter": i, "path": str(path.relative_to(PROJECT)), "words": len(words(text)), "sha256": sha(text)}) if not text.strip(): issues.append(f"empty_chapter:{i:02d}") for line_no, line in enumerate(text.splitlines(), 1): if HEADING_RE.match(line): issues.append(f"heading:{i:02d}:{line_no}") if PRODUCTION_RE.search(line): issues.append(f"production_marker:{i:02d}:{line_no}") narration_path = STORY / "spoken-narration.txt" narration = narration_path.read_text(encoding="utf-8") if narration_path.exists() else "" if not narration: issues.append("missing_or_empty_spoken_narration") expected_text = exact_join(chapter_texts) if all(chapter_texts) else "" if narration and expected_text and narration != expected_text: issues.append("spoken_narration_not_exact_sequential_join") count = len(words(narration)) low, high = outline["allowed_word_range"] if not low <= count <= high: issues.append(f"word_count_out_of_range:{count}") for line_no, line in enumerate(narration.splitlines(), 1): if HEADING_RE.match(line): issues.append(f"heading_in_spoken:{line_no}") if PRODUCTION_RE.search(line): issues.append(f"production_marker_in_spoken:{line_no}") id_hits = identity_hits(narration, registry) if narration else [] if id_hits: issues.append(f"standalone_identity_fragments:{len(id_hits)}") protected = reveal_hits(chapter_texts) if all(chapter_texts) else [] if protected: issues.append(f"protected_reveal_too_early:{len(protected)}") dead_now = [] if narration: for m in re.finditer(r"Tô Nhã Cầm\s+(?:đi|bước|mở|đứng|ngồi|gọi|nói|đưa|gửi)\b", narration, re.I): context = narration[max(0,m.start()-120):m.end()+140] if not re.search(r"(?:nhớ|hồi ức|năm ấy|ngày trước|lá thư|bản ghi|giọng.*?vang|trong mơ)", context, re.I | re.S): dead_now.append({"line": narration.count("\n",0,m.start())+1,"context":context.strip()}) if dead_now: issues.append(f"deceased_character_outside_historical_container:{len(dead_now)}") report = {"version":1,"status":"passed" if not issues else "failed","verified":not issues,"canonical_title":outline["canonical_title"],"expected_chapters":expected,"chapters_present":sum(p.exists() for p in paths),"chapter_rows":rows,"spoken_narration":str(narration_path.relative_to(PROJECT)),"total_words":count,"target_range":[low,high],"estimated_minutes_at_baseline":round(count/233.33,3) if count else 0,"spoken_sha256":sha(narration) if narration else None,"identity_hits":id_hits,"protected_reveal_hits":protected,"deceased_container_hits":dead_now,"issues":issues,"checked_at":datetime.now(timezone.utc).isoformat()} REPORT.write_text(json.dumps(report,ensure_ascii=False,indent=2)+"\n",encoding="utf-8") print(json.dumps({"verified":report["verified"],"chapters":report["chapters_present"],"total_words":count,"issues":issues[:20],"report":str(REPORT)},ensure_ascii=False)) return 0 if report["verified"] else 1 if __name__ == "__main__": sys.exit(main())