novel-maker 2.2.0

package/skill/scripts/auditor/pre_audit.py

@@ -0,0 +1,203 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+预审计管线 — 一键运行所有可自动化的审计维度
+
+将审计师需要手动执行的 5 个检查脚本整合为一个命令，
+输出结构化 JSON，审计师只需读取结果做 AI 判断，无需读取全文。
+
+用法:
+    python scripts/auditor/pre_audit.py novels/volume-01/chapters/ch15.md
+    python scripts/auditor/pre_audit.py novels/volume-01/chapters/ch15.md --json
+"""
+
+import argparse
+import json
+import os
+import sys
+import subprocess
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..', 'common'))
+from nm_utils import (
+    read_chapter, extract_characters, detect_hook,
+    detect_structure, estimate_pacing,
+    clean_markdown, generate_summary
+)
+
+
+def _run_script(script_name, args_list):
+    """运行子脚本并返回结果"""
+    script_path = os.path.join(os.path.dirname(__file__), script_name)
+    if not os.path.exists(script_path):
+        return {'error': f'{script_name} 不存在'}
+    try:
+        result = subprocess.run(
+            [sys.executable, script_path] + args_list,
+            capture_output=True, text=True, encoding='utf-8', timeout=30
+        )
+        if result.returncode == 0:
+            try:
+                return json.loads(result.stdout)
+            except json.JSONDecodeError:
+                return {'output': result.stdout.strip()}
+        else:
+            return {'error': result.stderr.strip() or f'{script_name} 返回码 {result.returncode}'}
+    except subprocess.TimeoutExpired:
+        return {'error': f'{script_name} 执行超时'}
+    except Exception as e:
+        return {'error': str(e)}
+
+
+def _check_wordcount(chapter_path, min_words=2000, max_words=4000):
+    """字数检查"""
+    raw, title, clean, wc = read_chapter(chapter_path)
+    status = 'OK'
+    if wc < min_words:
+        status = f'不足（{wc}<{min_words}）'
+    elif wc > max_words:
+        status = f'超标（{wc}>{max_words}）'
+    return {'word_count': wc, 'min': min_words, 'max': max_words, 'status': status}
+
+
+def _check_hook(chapter_path):
+    """章末钩子检查"""
+    with open(chapter_path, encoding='utf-8') as f:
+        text = f.read()
+    hook = detect_hook(text)
+    has_hook = hook['type'] not in ('未检测', '未知')
+    return {
+        'has_hook': has_hook,
+        'hook_type': hook['type'],
+        'cues': hook.get('cues', {}),
+        'tail_preview': hook.get('tail_preview', '')
+    }
+
+
+def _check_pacing(chapter_path):
+    """节奏评估"""
+    with open(chapter_path, encoding='utf-8') as f:
+        text = f.read()
+    pacing_label, pacing_score = estimate_pacing(text)
+    structure = detect_structure(text)
+    return {
+        'pacing': pacing_label,
+        'pacing_score': pacing_score,
+        'structure': structure
+    }
+
+
+def _check_ai_style(chapter_path):
+    """AI味检测（调用 style_check.py）"""
+    return _run_script('style_check.py', [chapter_path, '--json'])
+
+
+def _check_consistency(chapter_path, truth_dir=None):
+    """一致性扫描（调用 consistency_scan.py）"""
+    args = [chapter_path]
+    if truth_dir:
+        args.extend(['--truth-dir', truth_dir])
+    args.append('--json')
+    return _run_script('consistency_scan.py', args)
+
+
+def _check_characters(chapter_path):
+    """角色提取"""
+    with open(chapter_path, encoding='utf-8') as f:
+        text = f.read()
+    clean = clean_markdown(text)
+    chars = extract_characters(clean)
+    return {'characters': chars, 'count': len(chars)}
+
+
+def _check_summary(chapter_path):
+    """章节摘要"""
+    with open(chapter_path, encoding='utf-8') as f:
+        text = f.read()
+    summary = generate_summary(text, first_n=150, last_n=150)
+    return summary
+
+
+def run_pre_audit(chapter_path, truth_dir=None, min_words=2000, max_words=4000):
+    """运行完整预审计管线"""
+    if not os.path.exists(chapter_path):
+        return {'error': f'章节文件不存在: {chapter_path}'}
+
+    results = {
+        'chapter': os.path.basename(chapter_path),
+        'checks': {}
+    }
+
+    # 1. 字数检查
+    results['checks']['word_count'] = _check_wordcount(chapter_path, min_words, max_words)
+
+    # 2. 角色提取
+    results['checks']['characters'] = _check_characters(chapter_path)
+
+    # 3. 章末钩子
+    results['checks']['hook'] = _check_hook(chapter_path)
+
+    # 4. 节奏评估
+    results['checks']['pacing'] = _check_pacing(chapter_path)
+
+    # 5. 章节摘要
+    results['checks']['summary'] = _check_summary(chapter_path)
+
+    # 6. AI味检测
+    results['checks']['ai_style'] = _check_ai_style(chapter_path)
+
+    # 7. 一致性扫描（可选）
+    if truth_dir:
+        results['checks']['consistency'] = _check_consistency(chapter_path, truth_dir)
+
+    # 汇总
+    issues = []
+    wc = results['checks']['word_count']
+    if wc['status'] != 'OK':
+        issues.append(f"字数: {wc['status']}")
+    if not results['checks']['hook']['has_hook']:
+        issues.append("章末无钩子")
+    if results['checks']['pacing']['pacing_score'] >= 4:
+        issues.append(f"节奏偏高: {results['checks']['pacing']['pacing']}")
+
+    results['summary'] = {
+        'total_checks': len(results['checks']),
+        'issues': issues,
+        'issue_count': len(issues)
+    }
+
+    return results
+
+
+def main():
+    parser = argparse.ArgumentParser(description='预审计管线')
+    parser.add_argument('chapter', help='章节文件路径')
+    parser.add_argument('--truth-dir', help='真相文件目录')
+    parser.add_argument('--min-words', type=int, default=2000, help='最少字数（默认2000）')
+    parser.add_argument('--max-words', type=int, default=4000, help='最多字数（默认4000）')
+    parser.add_argument('--json', action='store_true', help='JSON输出')
+    args = parser.parse_args()
+
+    results = run_pre_audit(args.chapter, args.truth_dir, args.min_words, args.max_words)
+
+    if args.json:
+        print(json.dumps(results, ensure_ascii=False, indent=2))
+    else:
+        if 'error' in results:
+            print(f"错误: {results['error']}")
+            return
+        print(f"=== 预审计报告: {results['chapter']} ===\n")
+        for name, check in results['checks'].items():
+            if isinstance(check, dict) and 'error' not in check:
+                print(f"✓ {name}: OK")
+            elif isinstance(check, dict) and 'error' in check:
+                print(f"✗ {name}: {check['error']}")
+        if results['summary']['issues']:
+            print(f"\n发现 {results['summary']['issue_count']} 个问题:")
+            for issue in results['summary']['issues']:
+                print(f"  - {issue}")
+        else:
+            print("\n全部通过!")
+
+
+if __name__ == '__main__':
+    main()

package/skill/scripts/auditor/style_check.py

@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+AI写作风格检测
+检测 AI 高频词密度、连续"了/的"、句式重复、对话比例。
+"""
+
+import os
+import sys
+import json
+import re
+from collections import Counter
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..', 'common'))
+from nm_utils import list_chapters, extract_title, count_chinese
+
+
+AI_FREQ_WORDS = [
+    '苦涩', '复杂', '微妙', '难以言喻', '心中涌起', '内心深处',
+    '五味杂陈', '百感交集', '缓缓', '默默', '轻轻',
+    '深吸一口气', '苦笑一声', '皱了皱眉', '点了点头',
+    '犹如', '宛若', '好似', '映入眼帘', '传入耳中',
+    '不知何时', '不知不觉', '不禁', '不由得',
+    '一丝', '一抹', '仿佛', '似乎',
+]
+
+AI_PATTERNS = {
+    "连续'了'字": re.compile(r'了.{0,3}了.{0,3}了'),
+    "连续'的'字": re.compile(r'的.{0,3}的.{0,3}的'),
+    "不禁": re.compile(r'不禁[感到觉得想起]'),
+    "不由得": re.compile(r'不由得[想起觉得]'),
+    "仿佛一般": re.compile(r'仿佛.{1,10}一般'),
+    "似乎样子": re.compile(r'似乎.{1,10}的样子'),
+}
+
+
+def analyze_chapter_style(text, filepath=''):
+    """Analyze a single chapter for AI-like patterns."""
+    word_count = count_chinese(text)
+    if word_count == 0:
+        return None
+
+    title = extract_title(text) or os.path.basename(filepath)
+    issues = []
+
+    for pattern_name, pattern in AI_PATTERNS.items():
+        matches = pattern.findall(text)
+        if matches:
+            issues.append({
+                "type": "ai_pattern",
+                "pattern": pattern_name,
+                "count": len(matches),
+                "examples": [m[:30] for m in matches[:2]]
+            })
+
+    word_freq = Counter(re.findall(r'[\u4e00-\u9fff]{2}', text))
+    top_words = word_freq.most_common(15)
+    repetitive = [(w, c) for w, c in top_words if c > 5 and w not in ('我们', '你们', '他们', '自己', '什么')]
+    for w, c in repetitive:
+        issues.append({
+            "type": "repetitive_word",
+            "word": w,
+            "count": c
+        })
+
+    ai_word_count = 0
+    ai_found_words = []
+    for word in AI_FREQ_WORDS:
+        count = len(re.findall(re.escape(word), text))
+        if count > 0:
+            ai_word_count += count
+            ai_found_words.append({"word": word, "count": count})
+
+    ai_density = ai_word_count / max(word_count / 1000, 1)
+    if ai_density > 3:
+        issues.append({
+            "type": "ai_density",
+            "ai_words_per_1000": round(ai_density, 1),
+            "found_words": ai_found_words
+        })
+
+    paragraphs = [p.strip() for p in text.split('\n') if p.strip() and not p.strip().startswith('#')]
+    dialogue_lines = sum(1 for p in paragraphs if re.match(r'^["\u201c]', p))
+    dialogue_ratio = dialogue_lines / max(len(paragraphs), 1) * 100
+
+    sentences = re.split(r'[。！？]', text)
+    avg_sentence_len = sum(len(s) for s in sentences) / max(len(sentences), 1)
+
+    return {
+        "file": os.path.basename(filepath),
+        "title": title,
+        "word_count": word_count,
+        "paragraphs": len(paragraphs),
+        "dialogue_ratio_pct": round(dialogue_ratio, 1),
+        "avg_sentence_length": round(avg_sentence_len, 1),
+        "ai_density_per_1000": round(ai_density, 1),
+        "ai_words_found": ai_found_words,
+        "repetitive_words": [{"word": w, "count": c} for w, c in repetitive],
+        "issues": issues,
+        "top_words": [{"word": w, "count": c} for w, c in top_words[:8]],
+    }
+
+
+def main():
+    import argparse
+    parser = argparse.ArgumentParser(description="AI style check")
+    parser.add_argument("path", help="Chapter file or chapters directory")
+    parser.add_argument("--all", action="store_true", help="Analyze all chapters in directory")
+    parser.add_argument("--json", action="store_true", help="Output as JSON")
+    parser.add_argument("--recent", type=int, help="Only analyze last N chapters")
+    args = parser.parse_args()
+
+    path = args.path
+    if not os.path.exists(path):
+        print(f"Error: {path} does not exist", file=sys.stderr)
+        sys.exit(1)
+
+    if os.path.isdir(path) or args.all:
+        if os.path.isfile(path):
+            chapters_dir = os.path.dirname(path)
+        else:
+            chapters_dir = path
+        chapter_files = list_chapters(chapters_dir, recent_n=args.recent)
+    else:
+        chapter_files = [path]
+
+    results = []
+    for cf in chapter_files:
+        try:
+            with open(cf, 'r', encoding='utf-8') as f:
+                text = f.read()
+            analysis = analyze_chapter_style(text, cf)
+            if analysis:
+                results.append(analysis)
+        except Exception as e:
+            if args.json:
+                results.append({"file": os.path.basename(cf), "error": str(e)})
+
+    if args.json:
+        print(json.dumps(results, ensure_ascii=False, indent=2))
+    else:
+        for r in results:
+            print(f"\n=== {r['title']} ({r['word_count']}字) ===")
+            print(f"  对话占比: {r['dialogue_ratio_pct']}%  |  平均句长: {r['avg_sentence_length']}字  |  AI词密度: {r['ai_density_per_1000']}/千字")
+            if r['issues']:
+                for issue in r['issues']:
+                    if issue['type'] == 'ai_pattern':
+                        print(f"  ⚠ {issue['pattern']}: 出现{issue['count']}次")
+                    elif issue['type'] == 'repetitive_word':
+                        print(f"  ⚠ 重复词汇: 「{issue['word']}」出现{issue['count']}次")
+                    elif issue['type'] == 'ai_density':
+                        print(f"  ⚠ AI词密度过高: {issue['ai_words_per_1000']}/千字")
+            else:
+                print("  ✅ 无明显AI味")
+
+
+if __name__ == "__main__":
+    main()