feat: Release v0.10 - Modular Architecture & External Config

- Refactor codebase into src/ (preprocessing, translation, assembly) - Add pipeline/ scripts for individual stages - Externalize configuration to config/config.yaml - Fix Cover Image preservation - Update documentation and manuals
2026-01-31 22:49:44 +08:00
parent 9ef82393be
commit 7a93c52b42
306 changed files with 30313 additions and 1071 deletions
@@ -0,0 +1,10 @@
+{
+  "translation": {
+    "system": "你是一位精通中英文的专业翻译家。你的任务是将英文书籍内容翻译成中文。\n\n【核心要求】\n1. 准确传达原文含义，语言流畅自然，符合中文阅读习惯。\n2. 保持原文的风格和语气。\n\n【格式要求 - 极其重要】\n1. 每行格式：#N: 译文（N是行号，必须原样保留）\n2. 输入多少行，输出必须是相同数量的行\n3. **占位符规则**：原文中的 φXφ 和 φ/Xφ 标记必须原样保留在译文中\n   - 成对标记：\"φ1φBoldφ/1φ\" → \"φ1φ粗体φ/1φ\"\n   - 单体标记：\"φ2φ\" 表示公式或符号，保持位置不变\n   - 尾注锚点：文末的 \"φ3φ\" 是超链接，必须保留\n\n【禁止事项】\n- 禁止修改 #N: 行号\n- 禁止删除或修改任何 φXφ 标记\n- 禁止添加解释或注释\n- 禁止合并或拆分行",
+    "user_template": "请翻译以下段落（务必原样保留所有 φnφ 格式标记）：\n\n{{content}}"
+  },
+  "glossary_extraction": {
+    "system": "你是一位资深的文学编辑和领域专家。你的任务是分析书籍样本，提取关键术语并制定统一的译名表。",
+    "user_template": "请阅读以下书籍片段（包含前言和正文采样）。\n\n任务：\n1. 识别文中出现的人名（如 'Masa', 'Steve Jobs'）、地名、机构名。\n2. 识别特定的行业术语或关键概念。\n3. 为上述词汇提供标准的中文译名。如果像 'Masa' 这样的昵称有对应的全名（如孙正义），请务必使用全名。\n\n请以 JSON 格式输出，格式如下：\n{\n  \"Masa\": \"孙正义\",\n  \"Apple\": \"苹果公司\",\n  ...\n}\n\n书籍片段：\n\n{{content}}"
+  }
+}
@@ -0,0 +1,31 @@
+from bs4 import BeautifulSoup
+import zipfile
+
+def debug_clean():
+    epub_path = "input/Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao).epub"
+    file_in_zip = "OEBPS/c9.xhtml"
+    
+    with zipfile.ZipFile(epub_path, 'r') as zf:
+        content = zf.read(file_in_zip).decode('utf-8')
+        
+    print(f"Original Head length: {len(content)}")
+    if "<head>" in content:
+        print("Original has <head>")
+        
+    soup = BeautifulSoup(content, 'html.parser')
+    cleaned = str(soup)
+    
+    print(f"Cleaned Head length: {len(cleaned)}")
+    if "<head>" in cleaned:
+        print("Cleaned has <head>")
+        start = cleaned.find("<head>")
+        end = cleaned.find("</head>")
+        print(f"Cleaned Head content: {cleaned[start:end+7]}")
+    else:
+        print("Cleaned MISSING <head>")
+        # check for self-closing head
+        if "<head/>" in cleaned:
+             print("Cleaned has <head/> (empty)")
+
+if __name__ == "__main__":
+    debug_clean()
@@ -0,0 +1,141 @@
+import argparse
+import sys
+import os
+import asyncio
+from pathlib import Path
+from dotenv import load_dotenv
+
+from src.epub_cleaner import EpubCleaner
+from src.book_profiler import BookProfiler
+from src.fine_grained_extractor import FineGrainedExtractor
+from src.manifest_manager import ManifestManager
+from src.translator import Translator
+from src.llm_client import LLMClient
+from src.backfill_engine import BackfillEngine
+from src.bilingual_builder import BilingualBuilder
+from src.data_model import BookStructure
+from src.utils import setup_logger, ensure_directory
+from src.exceptions import EpubTranslatorError
+
+logger = setup_logger("main")
+
+# Unified work directory structure
+# .work/
+#   ├── {book_name}/
+#   │   ├── book_structure.json
+#   │   ├── manifest.json
+#   │   ├── assets/
+#   │   └── chunks/
+
+def get_work_dirs(input_path: Path) -> dict:
+    """Get work directory paths for a specific book."""
+    book_name = input_path.stem
+    work_root = Path(".work") / book_name
+    
+    return {
+        "root": work_root,
+        "structure": work_root / "book_structure.json",
+        "manifest": work_root / "manifest.json",
+        "assets": work_root / "assets",
+        "chunks": work_root / "chunks",
+    }
+
+
+async def run_pipeline(args):
+    input_path = Path(args.input_epub)
+    output_dir = Path(args.output_dir)
+    
+    # Get work directories for this book
+    work = get_work_dirs(input_path)
+    ensure_directory(work["root"])
+    ensure_directory(output_dir)
+    
+    # Load Environment
+    load_dotenv()
+    api_key = os.getenv("OPENAI_API_KEY")
+    base_url = os.getenv("OPENAI_BASE_URL", "https://api.openai.com/v1")
+    
+    if not api_key:
+        logger.warning("OPENAI_API_KEY not found in .env. LLM features may fail.")
+
+    try:
+        # 1. Preprocessing - Reuse book_structure.json if exists
+        if work["structure"].exists() and not args.force_clean:
+            logger.info(f"Reusing existing book_structure: {work['structure']}")
+            structure = BookStructure.load(work["structure"])
+        else:
+            logger.info("Cleaning EPUB and generating book_structure...")
+            cleaner = EpubCleaner(input_path, work["root"])
+            book_structure_json = cleaner.clean()
+            structure = BookStructure.load(book_structure_json)
+
+        # 2. Extraction
+        extractor = FineGrainedExtractor()
+        manifest_entries = extractor.extract(structure)
+        
+        # 3. Manifest Management
+        manifest_manager = ManifestManager(work["manifest"])
+        manifest_manager.load()  # Load existing if any
+        manifest_manager.add_entries(manifest_entries)
+        manifest_manager.save()
+        
+        # 4. Translation
+        if not args.skip_translation:
+            if not api_key:
+                logger.error("Cannot translate without API Key. Use --skip-translation to test pipeline.")
+                sys.exit(1)
+                
+            llm_client = LLMClient(api_key=api_key, base_url=base_url, model=args.model)
+            
+            # Profiling
+            profiler = BookProfiler(llm_client)
+            profile = await profiler.analyze(manifest_manager.entries)
+            logger.info(f"Book Profile: {profile}")
+            
+            # Translation
+            translator = Translator(llm_client)
+            await translator.translate(manifest_manager.entries, profile)
+            manifest_manager.save()
+            
+            await llm_client.close()
+        else:
+             logger.info("Skipping translation step.")
+        
+        # 5. Backfill
+        backfiller = BackfillEngine()
+        updated_structure = backfiller.backfill(structure, manifest_manager.entries, mode=args.mode)
+        
+        # 6. Assembly - Pass original EPUB for TOC preservation
+        builder = BilingualBuilder(work["root"], original_epub_path=input_path)
+        output_filename = f"bilingual_{input_path.name}"
+        output_path = output_dir / output_filename
+        
+        created_epub = builder.build(updated_structure, output_path)
+        
+        logger.info(f"Pipeline completed! Output: {created_epub}")
+        
+    except EpubTranslatorError as e:
+        logger.error(f"An error occurred: {e}")
+        sys.exit(1)
+    except Exception as e:
+        logger.critical(f"Unexpected error: {e}")
+        import traceback
+        traceback.print_exc()
+        sys.exit(1)
+
+def main():
+    parser = argparse.ArgumentParser(description="EPUB Bilingual Translator")
+    parser.add_argument("input_epub", help="Path to the input EPUB file")
+    parser.add_argument("--output-dir", default="output", help="Directory for output files")
+    parser.add_argument("--model", default="gpt-3.5-turbo", help="LLM Model to use")
+    parser.add_argument("--mode", default="bilingual", choices=["bilingual", "target_only"], help="Output mode")
+    parser.add_argument("--skip-translation", action="store_true", help="Skip LLM translation (for testing)")
+    parser.add_argument("--force-clean", action="store_true", help="Force re-clean EPUB even if book_structure exists")
+    
+    args = parser.parse_args()
+    
+    asyncio.run(run_pipeline(args))
+
+if __name__ == "__main__":
+    main()
+
@@ -0,0 +1,51 @@
+import sys
+from pathlib import Path
+import traceback
+
+# Add src to path
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from src.data_model import BookStructure, ManifestEntry
+from src.manifest_manager import ManifestManager
+from src.backfill_engine import BackfillEngine
+from src.bilingual_builder import BilingualBuilder
+from src.utils import setup_logger
+
+logger = setup_logger("build_final")
+
+def build_final():
+    BOOK_NAME = "Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao)"
+    WORK_DIR = Path(f".work/{BOOK_NAME}")
+    MANIFEST_PATH = WORK_DIR / "manifest.json"
+    STRUCTURE_PATH = WORK_DIR / "book_structure.json"
+    OUTPUT_EPUB = Path("output/final_verification.epub")
+    INPUT_EPUB = Path(f"input/{BOOK_NAME}.epub")
+
+    if not MANIFEST_PATH.exists() or not STRUCTURE_PATH.exists():
+        logger.error("Missing manifest or structure. Run translation first.")
+        return
+
+    # 1. Load Data
+    logger.info("Loading structure and manifest...")
+    structure = BookStructure.load(STRUCTURE_PATH)
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+
+    # 2. Backfill
+    logger.info("Backfilling translations...")
+    backfiller = BackfillEngine()
+    structure = backfiller.backfill(structure, manager.entries, mode="bilingual")
+
+    # 3. Build
+    logger.info(f"Building final EPUB to {OUTPUT_EPUB}...")
+    builder = BilingualBuilder(WORK_DIR, original_epub_path=INPUT_EPUB)
+    builder.build(structure, OUTPUT_EPUB)
+    
+    logger.info("Build complete.")
+
+if __name__ == "__main__":
+    try:
+        build_final()
+    except Exception as e:
+        traceback.print_exc()
+        print(f"CRITICAL ERROR: {e}")
@@ -0,0 +1,212 @@
+#!/usr/bin/env python3
+"""
+Placeholder Backfill Test Script
+Tests placeholder restoration on translated entries.
+
+Usage:
+    python scripts/test_backfill.py --chapter 38
+    python scripts/test_backfill.py --chapter 38 --limit 10
+"""
+
+import argparse
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from src.manifest_manager import ManifestManager
+from src.format_restorer import FormatRestorer
+
+# Configuration - Use unified .work directory
+BOOK_NAME = "Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao)"
+WORK_DIR = Path(f".work/{BOOK_NAME}")
+MANIFEST_PATH = WORK_DIR / "manifest.json"
+
+
+def build_toc_from_manifest(manager: ManifestManager) -> list:
+    """Build TOC from manifest entries."""
+    files = {}
+    for entry in manager.entries:
+        fp = entry.file_path
+        if fp not in files:
+            files[fp] = {'count': 0, 'first_text': ''}
+        files[fp]['count'] += 1
+        if not files[fp]['first_text'] and entry.original_text:
+            files[fp]['first_text'] = entry.original_text[:40].replace('\n', ' ')
+    
+    toc = []
+    for i, (fp, info) in enumerate(sorted(files.items()), 1):
+        toc.append({
+            'index': i,
+            'href': fp,
+            'title': info['first_text'] or f"File {i}",
+            'paragraphs': info['count']
+        })
+    return toc
+
+
+def get_chapter_entries(manager: ManifestManager, chapter_index: int) -> tuple:
+    """Get entries for a specific chapter by index."""
+    toc = build_toc_from_manifest(manager)
+    
+    if chapter_index < 1 or chapter_index > len(toc):
+        print(f"错误: 章节编号 {chapter_index} 无效 (范围: 1-{len(toc)})")
+        return None, None
+    
+    chapter = toc[chapter_index - 1]
+    href = chapter['href']
+    
+    entries = [e for e in manager.entries if e.file_path == href]
+    return chapter, entries
+
+
+def test_backfill(chapter: dict, entries: list, limit: int = None):
+    """Test placeholder restoration for a chapter."""
+    
+    print(f"\n" + "=" * 70)
+    print(f"占位符回填测试 - 章节 #{chapter['index']}: {chapter['title'][:40]}...")
+    print("=" * 70)
+    
+    # Filter entries with translation and placeholders
+    translated = [e for e in entries if e.translated_text]
+    with_placeholders = [e for e in translated if e.placeholders and len(e.placeholders) > 0]
+    
+    print(f"\n统计:")
+    print(f"  总段落: {len(entries)}")
+    print(f"  已翻译: {len(translated)}")
+    print(f"  有占位符: {len(with_placeholders)}")
+    
+    if not translated:
+        print("\n⚠️ 该章节没有已翻译的内容!")
+        return
+    
+    # Test restoration
+    restorer = FormatRestorer()
+    
+    success_count = 0
+    fail_count = 0
+    results = []
+    
+    test_entries = with_placeholders[:limit] if limit else with_placeholders
+    
+    print(f"\n测试 {len(test_entries)} 个带占位符的段落:")
+    print("-" * 70)
+    
+    for i, entry in enumerate(test_entries, 1):
+        original = entry.original_text
+        translated = entry.translated_text
+        placeholders = entry.placeholders
+        
+        # Get non-internal placeholders
+        visible_ph = {k: v for k, v in placeholders.items() if not k.startswith('_')}
+        
+        # Perform restoration
+        restored, success = restorer.restore(translated, placeholders)
+        
+        if success:
+            success_count += 1
+            status = "✅"
+        else:
+            fail_count += 1
+            status = "❌"
+        
+        results.append({
+            'index': i,
+            'entry_id': entry.entry_id,
+            'original': original,
+            'translated': translated,
+            'restored': restored,
+            'placeholders': visible_ph,
+            'success': success
+        })
+        
+        # Print summary
+        print(f"\n[{i}] {status} {entry.entry_id[-40:]}")
+        print(f"    占位符: {list(visible_ph.keys())}")
+        print(f"    原文: {original[:50]}...")
+        print(f"    译文: {translated[:50]}...")
+        
+        if not success:
+            print(f"    还原: {restored[:50]}...")
+            # Show what placeholders are missing
+            missing = []
+            for k in visible_ph.keys():
+                if k.isdigit():
+                    if f"φ{k}φ" not in translated and f"φ/{k}φ" not in translated:
+                        missing.append(k)
+            if missing:
+                print(f"    缺失: {missing}")
+    
+    # Summary
+    print("\n" + "=" * 70)
+    print(f"测试结果汇总")
+    print("=" * 70)
+    print(f"  成功: {success_count}/{len(test_entries)}")
+    print(f"  失败: {fail_count}/{len(test_entries)}")
+    
+    if fail_count > 0:
+        print(f"\n失败案例详情:")
+        for r in results:
+            if not r['success']:
+                print(f"\n  [{r['index']}] {r['entry_id'][-50:]}")
+                print(f"      原文: {r['original'][:60]}...")
+                print(f"      译文: {r['translated'][:60]}...")
+                print(f"      还原: {r['restored'][:60]}...")
+                print(f"      占位符: {r['placeholders']}")
+    
+    # Also test entries without visible placeholders (only _prefix/_suffix)
+    prefix_suffix_only = [e for e in translated 
+                         if e.placeholders 
+                         and all(k.startswith('_') for k in e.placeholders.keys())]
+    
+    if prefix_suffix_only:
+        print(f"\n\n额外测试: 只有 _prefix/_suffix 的段落 ({len(prefix_suffix_only)} 个)")
+        print("-" * 70)
+        
+        ps_success = 0
+        ps_fail = 0
+        
+        for entry in prefix_suffix_only[:5]:  # Test first 5
+            restored, success = restorer.restore(entry.translated_text, entry.placeholders)
+            if success:
+                ps_success += 1
+                status = "✅"
+            else:
+                ps_fail += 1
+                status = "❌"
+            
+            print(f"  {status} {entry.entry_id[-40:]}")
+            if '_prefix' in entry.placeholders:
+                print(f"      _prefix: {entry.placeholders['_prefix'][:30]}...")
+            if '_suffix' in entry.placeholders:
+                print(f"      _suffix: {entry.placeholders['_suffix'][:30]}...")
+        
+        print(f"\n  结果: {ps_success}/{min(5, len(prefix_suffix_only))} 成功")
+
+
+def main():
+    parser = argparse.ArgumentParser(description="测试占位符回填")
+    parser.add_argument("--chapter", "-c", type=int, required=True, help="章节编号")
+    parser.add_argument("--limit", "-l", type=int, default=20, help="测试数量限制 (默认20)")
+    args = parser.parse_args()
+    
+    if not MANIFEST_PATH.exists():
+        print(f"错误: Manifest 不存在: {MANIFEST_PATH}")
+        return
+    
+    # Load manifest
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    print(f"已加载 manifest: {len(manager.entries)} 条目")
+    
+    # Get chapter
+    chapter, entries = get_chapter_entries(manager, args.chapter)
+    if not chapter:
+        return
+    
+    # Test backfill
+    test_backfill(chapter, entries, args.limit)
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,25 @@
+import asyncio
+import os
+import httpx
+
+async def test_conn():
+    print(f"HTTP_PROXY: {os.environ.get('http_proxy')}")
+    print(f"HTTPS_PROXY: {os.environ.get('https_proxy')}")
+    print(f"ALL_PROXY: {os.environ.get('all_proxy')}")
+    
+    url = "https://api.gpt.ge/v1/models"
+    headers = {"Authorization": f"Bearer {os.environ.get('V3_API_KEY')}"}
+    
+    print(f"Connecting to {url}...")
+    try:
+        async with httpx.AsyncClient(timeout=10, follow_redirects=True) as client:
+            resp = await client.get(url, headers=headers)
+            print(f"Status: {resp.status_code}")
+            print(f"Headers: {resp.headers}")
+    except Exception as e:
+        print(f"Error: {e}")
+
+if __name__ == "__main__":
+    from dotenv import load_dotenv
+    load_dotenv()
+    asyncio.run(test_conn())
@@ -0,0 +1,40 @@
+import asyncio
+import os
+import httpx
+from openai import AsyncOpenAI
+
+async def test_openai():
+    proxy_url = os.environ.get("http_proxy")
+    print(f"Using proxy: {proxy_url}")
+    
+    http_client = httpx.AsyncClient(
+        proxy=proxy_url,
+        timeout=30.0,
+        follow_redirects=True
+    )
+
+    client = AsyncOpenAI(
+        base_url="https://api.gpt.ge/v1",
+        api_key=os.environ.get("V3_API_KEY"),
+        http_client=http_client
+    )
+    
+    print("Sending request...")
+    try:
+        response = await client.chat.completions.create(
+            model="gpt-3.5-turbo",
+            messages=[{"role": "user", "content": "Hello"}],
+            max_tokens=5
+        )
+        print(f"Response: {response.choices[0].message.content}")
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        print(f"Error: {e}")
+    finally:
+        await http_client.aclose()
+
+if __name__ == "__main__":
+    from dotenv import load_dotenv
+    load_dotenv()
+    asyncio.run(test_openai())
@@ -0,0 +1,211 @@
+#!/usr/bin/env python3
+"""
+Debug script to show chunk content and test short ID strategy.
+"""
+
+import asyncio
+import json
+import os
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from dotenv import load_dotenv
+from src.manifest_manager import ManifestManager
+from src.llm_client import LLMClient
+from src.data_model import ManifestEntry
+
+load_dotenv()
+
+# Configuration
+MANIFEST_PATH = Path("cache/Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao)_manifest.json")
+API_KEY = os.getenv("V3_API_KEY")
+BASE_URL = "https://api.gpt.ge/v1"
+MODEL = "gemini-3-flash-preview"
+EXTRA_HEADERS = {"x-foo": "true"}
+
+
+def show_current_chunk_format():
+    """Show current chunk format (problematic)."""
+    print("\n" + "="*60)
+    print("当前 Chunk 格式 (有问题)")
+    print("="*60)
+    
+    # Load manifest
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    
+    # Get sample entries with placeholders
+    samples = [e for e in manager.entries if e.placeholders][:3]
+    
+    print("\n发送给 LLM 的格式 (当前):")
+    print("-"*60)
+    for item in samples:
+        context = item.context or "BODY"
+        print(f"{item.entry_id} [{context}] {item.original_text[:50]}...")
+    
+    print("\n问题分析:")
+    print("  1. entry_id 太长 (包含 UUID): 容易被 LLM 截断或修改")
+    print("  2. [BODY] context 没必要发送")
+    print("  3. 依赖 LLM 精确复制长 ID，不可靠")
+
+
+def show_proposed_chunk_format():
+    """Show proposed short ID chunk format."""
+    print("\n" + "="*60)
+    print("建议的 Chunk 格式 (短 ID)")
+    print("="*60)
+    
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    
+    samples = [e for e in manager.entries if e.placeholders][:5]
+    
+    print("\n发送给 LLM 的格式 (建议):")
+    print("-"*60)
+    
+    # Build with short IDs
+    id_map = {}  # short_id -> entry_id
+    for i, item in enumerate(samples, 1):
+        short_id = f"#{i}"
+        id_map[short_id] = item.entry_id
+        text = item.original_text[:60]
+        print(f"{short_id}: {text}...")
+    
+    print("\n期望 LLM 返回的格式:")
+    print("-"*60)
+    print("#1: φ1φpenguinrandomhouse.com（保持不翻译）")
+    print("#2: 该产品在欧盟的产品安全授权代表为 φ1φPenguin Random House Irelandφ/1φ...")
+    print("#3: φ1φ献辞")
+    print("#4: φ1φ题记")  
+    print("#5: φ1φ作者说明")
+    
+    print("\n优势:")
+    print("  1. 短 ID (#1, #2...) 不会被 LLM 弄乱")
+    print("  2. 去掉了无用的 context 标签")
+    print("  3. 解析更可靠：用正则 ^#(\\d+): 匹配")
+    print("  4. ID 映射表保留在代码中，用于还原")
+    
+    print("\n映射表 (代码内保留):")
+    for short_id, full_id in id_map.items():
+        print(f"  {short_id} -> {full_id[:50]}...")
+
+
+async def test_short_id_translation():
+    """Test translation with short ID format."""
+    print("\n" + "="*60)
+    print("测试短 ID 翻译")
+    print("="*60)
+    
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    
+    # Get 5 entries with varied content
+    samples = [e for e in manager.entries if len(e.original_text) > 20][:5]
+    
+    # Build prompt with short IDs
+    id_map = {}
+    lines = []
+    for i, item in enumerate(samples, 1):
+        short_id = f"#{i}"
+        id_map[short_id] = item.entry_id
+        text = item.original_text.replace('\n', ' ').strip()
+        lines.append(f"{short_id}: {text}")
+    
+    user_prompt = "\n".join(lines)
+    
+    print("\n发送给 LLM 的 Prompt:")
+    print("-"*60)
+    print(user_prompt)
+    
+    # Create client
+    from openai import AsyncOpenAI
+    client = AsyncOpenAI(
+        base_url=BASE_URL,
+        api_key=API_KEY,
+        default_headers=EXTRA_HEADERS
+    )
+    
+    system_prompt = """You are a professional English to Chinese translator.
+
+Translate each line to Chinese. Keep the format:
+- Each line starts with #N: (keep this ID exactly)
+- Preserve any φXφ placeholders exactly as-is
+- Only output translations, no explanations
+
+Example input:
+#1: Hello world
+#2: φ1φClick hereφ/1φ to continue
+
+Example output:
+#1: 你好世界
+#2: φ1φ点击这里φ/1φ 继续"""
+    
+    print("\n发送请求...")
+    
+    try:
+        resp = await client.chat.completions.create(
+            model=MODEL,
+            messages=[
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_prompt}
+            ],
+            temperature=0.3,
+        )
+        raw_response = resp.choices[0].message.content.strip()
+        
+        print("\nLLM 返回:")
+        print("-"*60)
+        print(raw_response)
+        
+        # Parse with short ID
+        print("\n解析结果:")
+        print("-"*60)
+        import re
+        results = {}
+        for line in raw_response.split("\n"):
+            line = line.strip()
+            match = re.match(r'^#(\d+):\s*(.+)$', line)
+            if match:
+                short_id = f"#{match.group(1)}"
+                translation = match.group(2)
+                if short_id in id_map:
+                    full_id = id_map[short_id]
+                    results[full_id] = translation
+                    print(f"  {short_id} -> {translation[:40]}...")
+        
+        print(f"\n成功解析: {len(results)}/{len(samples)}")
+        
+    finally:
+        await client.close()
+
+
+async def main():
+    print("="*60)
+    print("Chunk ID 策略分析与测试")
+    print("="*60)
+    
+    if not MANIFEST_PATH.exists():
+        print(f"ERROR: Manifest not found at {MANIFEST_PATH}")
+        return
+    
+    # 1. Show current format (problems)
+    show_current_chunk_format()
+    
+    # 2. Show proposed format
+    show_proposed_chunk_format()
+    
+    # 3. Test short ID translation
+    if API_KEY:
+        await test_short_id_translation()
+    else:
+        print("\n跳过测试 (V3_API_KEY 未设置)")
+    
+    print("\n" + "="*60)
+    print("分析完成")
+    print("="*60)
+
+
+if __name__ == "__main__":
+    asyncio.run(main())
@@ -0,0 +1,355 @@
+#!/usr/bin/env python3
+"""
+Translation Pipeline Debug Script
+Shows visible results at each step of the translation process.
+"""
+
+import asyncio
+import json
+import os
+import sys
+from pathlib import Path
+
+# Add project root to path
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from dotenv import load_dotenv
+
+from src.manifest_manager import ManifestManager
+from src.llm_client import LLMClient
+from src.book_profiler import BookProfiler
+from src.translator import Translator
+from src.format_restorer import FormatRestorer
+from src.data_model import ManifestEntry, BookProfile
+
+load_dotenv()
+
+# Configuration
+MANIFEST_PATH = Path("cache/Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao)_manifest.json")
+API_KEY = os.getenv("V3_API_KEY")
+BASE_URL = "https://api.gpt.ge/v1"
+MODEL = "gemini-3-flash-preview"
+EXTRA_HEADERS = {"x-foo": "true"}
+
+# Chunk size configuration
+MAX_CHUNK_SIZE = 15  # Maximum entries per chunk
+
+
+def group_entries_by_file(entries: list) -> dict:
+    """Group manifest entries by their source file (chapter)."""
+    grouped = {}
+    for entry in entries:
+        file_path = entry.file_path
+        if file_path not in grouped:
+            grouped[file_path] = []
+        grouped[file_path].append(entry)
+    return grouped
+
+
+def create_chapter_aware_chunks(entries: list, max_size: int = MAX_CHUNK_SIZE) -> list:
+    """
+    Create chunks that respect chapter boundaries.
+    Returns list of (file_path, chunk_entries) tuples.
+    """
+    grouped = group_entries_by_file(entries)
+    chunks = []
+    
+    for file_path, file_entries in grouped.items():
+        # Split this file's entries into chunks of max_size
+        for i in range(0, len(file_entries), max_size):
+            chunk = file_entries[i:i + max_size]
+            chunks.append((file_path, chunk))
+    
+    return chunks
+
+
+def get_file_type(file_path: str) -> str:
+    """Determine the type of content based on file name."""
+    fname = file_path.lower()
+    if any(k in fname for k in ['toc', 'contents', 'nav']):
+        return 'toc'
+    elif any(k in fname for k in ['title', 'cover']):
+        return 'cover'
+    elif any(k in fname for k in ['copyright', 'colophon']):
+        return 'legal'
+    elif any(k in fname for k in ['author', 'about']):
+        return 'author_bio'
+    elif any(k in fname for k in ['index', 'bibliography', 'endnote', 'footnote']):
+        return 'reference'
+    else:
+        return 'body'
+
+
+async def step1_load_manifest():
+    """Step 1: Load manifest and show statistics."""
+    print("\n" + "="*60)
+    print("STEP 1: Loading Manifest")
+    print("="*60)
+    
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    
+    entries = manager.entries
+    untranslated = [e for e in entries if not e.translated_text]
+    
+    print(f"  Total entries: {len(entries)}")
+    print(f"  Untranslated:  {len(untranslated)}")
+    
+    # Show grouping by file
+    grouped = group_entries_by_file(entries)
+    print(f"  Unique files:  {len(grouped)}")
+    
+    # Show sample entry
+    if entries:
+        sample = entries[0]
+        print(f"\n  Sample entry:")
+        print(f"    ID:       {sample.entry_id}")
+        print(f"    File:     {sample.file_path}")
+        print(f"    Original: {sample.original_text[:80]}...")
+        print(f"    Placeholders: {sample.placeholders}")
+    
+    return manager
+
+
+async def step2_profile_book(manager: ManifestManager, llm_client: LLMClient):
+    """Step 2: Generate book profile."""
+    print("\n" + "="*60)
+    print("STEP 2: Generating Book Profile")
+    print("="*60)
+    
+    profiler = BookProfiler(llm_client)
+    profile = await profiler.analyze(manager.entries)
+    
+    print(f"  Title:       {profile.title}")
+    print(f"  Author:      {profile.author}")
+    print(f"  Genre:       {profile.genre}")
+    print(f"  Keywords:    {profile.keywords}")
+    print(f"  Style Guide: {profile.style_guide[:200]}..." if profile.style_guide else "  Style Guide: (none)")
+    
+    return profile
+
+
+async def step3_create_chunks(manager: ManifestManager):
+    """Step 3: Create chapter-aware chunks."""
+    print("\n" + "="*60)
+    print("STEP 3: Creating Chapter-Aware Chunks")
+    print("="*60)
+    
+    untranslated = [e for e in manager.entries if not e.translated_text]
+    chunks = create_chapter_aware_chunks(untranslated)
+    
+    print(f"  Total chunks: {len(chunks)}")
+    
+    # Show chunk distribution
+    print(f"\n  Chunk distribution by file type:")
+    type_counts = {}
+    for file_path, chunk_entries in chunks:
+        ftype = get_file_type(file_path)
+        type_counts[ftype] = type_counts.get(ftype, 0) + 1
+    
+    for ftype, count in sorted(type_counts.items()):
+        print(f"    {ftype}: {count} chunks")
+    
+    # Show first 3 chunks
+    print(f"\n  First 3 chunks:")
+    for i, (file_path, chunk_entries) in enumerate(chunks[:3]):
+        ftype = get_file_type(file_path)
+        print(f"    [{i}] {file_path} ({ftype}): {len(chunk_entries)} entries")
+        if chunk_entries:
+            print(f"        First: {chunk_entries[0].original_text[:50]}...")
+    
+    return chunks
+
+
+async def step4_translate_sample(chunks: list, llm_client: LLMClient, profile: BookProfile):
+    """Step 4: Translate a sample chunk and show results."""
+    print("\n" + "="*60)
+    print("STEP 4: Translating Sample Chunk")
+    print("="*60)
+    
+    if not chunks:
+        print("  No chunks to translate!")
+        return
+    
+    # Pick a proper body chapter (skip first few files which are usually cover/copyright/toc)
+    sample_chunk = None
+    skip_prefixes = ['cM', 'c9', 'c18']  # Cover, title, contents pages
+    for file_path, chunk_entries in chunks:
+        # Skip non-body files and known cover/toc files
+        ftype = get_file_type(file_path)
+        fname = Path(file_path).stem
+        if ftype == 'body' and fname not in skip_prefixes and len(chunk_entries) > 3:
+            sample_chunk = (file_path, chunk_entries[:5])  # Limit to 5 entries for demo
+            break
+    
+    if not sample_chunk:
+        # Fallback to any body chunk
+        for file_path, chunk_entries in chunks:
+            if get_file_type(file_path) == 'body':
+                sample_chunk = (file_path, chunk_entries[:5])
+                break
+    
+    if not sample_chunk:
+        sample_chunk = chunks[0]
+        sample_chunk = (sample_chunk[0], sample_chunk[1][:5])
+    
+    file_path, entries = sample_chunk
+    ftype = get_file_type(file_path)
+    
+    print(f"  Selected chunk: {file_path} ({ftype})")
+    print(f"  Entries: {len(entries)}")
+    
+    # Show entries before translation
+    print(f"\n  === Before Translation ===")
+    for i, entry in enumerate(entries):
+        print(f"  [{i}] {entry.entry_id}")
+        print(f"      Original: {entry.original_text[:60]}...")
+        if entry.placeholders:
+            print(f"      Placeholders: {list(entry.placeholders.keys())}")
+    
+    # Translate
+    print(f"\n  Translating...")
+    results = await llm_client.translate_chunk(
+        entries,
+        instruction=profile.style_guide,
+        mode="bilingual"
+    )
+    
+    # Apply results and show
+    print(f"\n  === After Translation ===")
+    restorer = FormatRestorer()
+    for i, entry in enumerate(entries):
+        if entry.entry_id in results:
+            translated = results[entry.entry_id]
+            entry.translated_text = translated
+            
+            print(f"  [{i}] {entry.entry_id}")
+            print(f"      Original:    {entry.original_text[:50]}...")
+            print(f"      Translated:  {translated[:50]}...")
+            
+            # Restore format
+            if entry.placeholders:
+                restored, success = restorer.restore(translated, entry.placeholders)
+                print(f"      Restored OK: {success}")
+                if not success:
+                    print(f"      Restored:    {restored[:50]}...")
+        else:
+            print(f"  [{i}] MISSING: {entry.entry_id}")
+    
+    return entries
+
+
+async def step5_test_placeholders(manager: ManifestManager, llm_client: LLMClient, profile: BookProfile):
+    """Step 5: Test placeholder handling with entries that have placeholders."""
+    print("\n" + "="*60)
+    print("STEP 5: Testing Placeholder Handling")
+    print("="*60)
+    
+    # Find entries with placeholders
+    entries_with_ph = [e for e in manager.entries if e.placeholders and len(e.placeholders) > 1]
+    
+    print(f"  Entries with placeholders: {len(entries_with_ph)}")
+    
+    if not entries_with_ph:
+        print("  No entries with placeholders found!")
+        return
+    
+    # Pick 5 diverse samples
+    samples = entries_with_ph[:5]
+    
+    print(f"\n  === Selected Samples ({len(samples)}) ===")
+    for i, entry in enumerate(samples):
+        ph_keys = [k for k in entry.placeholders.keys() if not k.startswith('_')]
+        print(f"  [{i}] {entry.entry_id}")
+        print(f"      Original: {entry.original_text[:60]}...")
+        print(f"      Placeholders: {ph_keys}")
+    
+    # Translate
+    print(f"\n  Translating {len(samples)} entries with placeholders...")
+    results = await llm_client.translate_chunk(
+        samples,
+        instruction=profile.style_guide,
+        mode="bilingual"
+    )
+    
+    # Show results with restoration
+    print(f"\n  === Translation Results ===")
+    restorer = FormatRestorer()
+    success_count = 0
+    
+    for i, entry in enumerate(samples):
+        print(f"\n  [{i}] {entry.entry_id}")
+        print(f"      Original:     {entry.original_text[:50]}...")
+        
+        if entry.entry_id in results:
+            translated = results[entry.entry_id]
+            print(f"      Translated:   {translated[:50]}...")
+            
+            # Check if placeholders are preserved
+            ph_keys = [k for k in entry.placeholders.keys() if not k.startswith('_')]
+            preserved = all(f"φ{k}φ" in translated or f"φ/{k}φ" in translated for k in ph_keys if k.isdigit())
+            print(f"      PH Preserved: {preserved}")
+            
+            # Restore format
+            restored, success = restorer.restore(translated, entry.placeholders)
+            print(f"      Restore OK:   {success}")
+            
+            if success:
+                success_count += 1
+            else:
+                print(f"      Restored:     {restored[:50]}...")
+        else:
+            print(f"      MISSING from results!")
+    
+    print(f"\n  Summary: {success_count}/{len(samples)} restored successfully")
+
+
+async def main():
+    """Run all steps."""
+    print("="*60)
+    print("TRANSLATION PIPELINE DEBUG")
+    print("="*60)
+    
+    if not API_KEY:
+        print("ERROR: V3_API_KEY not found in .env")
+        return
+    
+    if not MANIFEST_PATH.exists():
+        print(f"ERROR: Manifest not found at {MANIFEST_PATH}")
+        print("Run the main pipeline first to generate the manifest.")
+        return
+    
+    # Initialize LLM client
+    llm_client = LLMClient(
+        api_key=API_KEY,
+        base_url=BASE_URL,
+        model=MODEL,
+        extra_headers=EXTRA_HEADERS
+    )
+    
+    try:
+        # Step 1: Load manifest
+        manager = await step1_load_manifest()
+        
+        # Step 2: Profile book
+        profile = await step2_profile_book(manager, llm_client)
+        
+        # Step 3: Create chunks
+        chunks = await step3_create_chunks(manager)
+        
+        # Step 4: Translate sample (simple text)
+        await step4_translate_sample(chunks, llm_client, profile)
+        
+        # Step 5: Test placeholders
+        await step5_test_placeholders(manager, llm_client, profile)
+        
+        print("\n" + "="*60)
+        print("DEBUG COMPLETE")
+        print("="*60)
+        
+    finally:
+        await llm_client.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())
@@ -0,0 +1,353 @@
+#!/usr/bin/env python3
+"""
+Chapter Translation Test Script
+Translate a complete chapter to test the full pipeline.
+
+Usage:
+    python scripts/translate_chapter.py --show-toc           # 显示章节目录
+    python scripts/translate_chapter.py --chapter 5          # 翻译第5章
+    python scripts/translate_chapter.py --chapter 5 --test   # 测试模式，只翻译前2个chunk
+"""
+
+import argparse
+import asyncio
+import json
+import os
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from dotenv import load_dotenv
+from src.manifest_manager import ManifestManager
+from src.llm_client import LLMClient
+from src.book_profiler import BookProfiler
+from src.format_restorer import FormatRestorer
+from src.data_model import ManifestEntry, BookStructure
+
+load_dotenv()
+
+# Configuration - Use unified .work directory
+BOOK_NAME = "Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao)"
+WORK_DIR = Path(f".work/{BOOK_NAME}")
+MANIFEST_PATH = WORK_DIR / "manifest.json"
+STRUCTURE_PATH = WORK_DIR / "book_structure.json"
+CHUNK_DIR = WORK_DIR / "chunks"
+
+API_KEY = os.getenv("V3_API_KEY")
+BASE_URL = "https://api.gpt.ge/v1"
+MODEL = "gpt-4o-mini" # "gemini-3-flash-preview"
+EXTRA_HEADERS = {"x-foo": "true"}
+
+# Chunk config - around 5000 chars per chunk
+CHUNK_SIZE_CHARS = 5000
+
+
+def load_toc_from_structure() -> list:
+    """Load TOC from book structure for readable chapter names."""
+    if not STRUCTURE_PATH.exists():
+        return []
+    
+    try:
+        structure = BookStructure.load(STRUCTURE_PATH)
+        # Build TOC from spine order with chapter titles
+        toc = []
+        for i, item_id in enumerate(structure.spine, 1):
+            if item_id in structure.resources:
+                resource = structure.resources[item_id]
+                href = resource.href
+                
+                # Try to extract title from content
+                title = extract_title_from_html(resource.content) if resource.content else None
+                
+                toc.append({
+                    'index': i,
+                    'item_id': item_id,
+                    'href': href,
+                    'title': title or f"Chapter {i}"
+                })
+        return toc
+    except Exception as e:
+        print(f"警告: 无法加载书籍结构: {e}")
+        return []
+
+
+def extract_title_from_html(html: str) -> str:
+    """Extract title from HTML content."""
+    from bs4 import BeautifulSoup
+    soup = BeautifulSoup(html, 'html.parser')
+    
+    # Try h1, h2, h3 in order
+    for tag in ['h1', 'h2', 'h3']:
+        elem = soup.find(tag)
+        if elem:
+            return elem.get_text().strip()[:50]
+    
+    # Try first paragraph
+    p = soup.find('p')
+    if p:
+        text = p.get_text().strip()[:50]
+        if text:
+            return text + "..."
+    
+    return None
+
+
+def build_toc_from_manifest(manager: ManifestManager) -> list:
+    """Build TOC from manifest entries."""
+    files = {}
+    for entry in manager.entries:
+        fp = entry.file_path
+        if fp not in files:
+            files[fp] = {
+                'count': 0,
+                'first_text': '',
+                'total_chars': 0
+            }
+        files[fp]['count'] += 1
+        files[fp]['total_chars'] += len(entry.original_text)
+        if not files[fp]['first_text'] and entry.original_text:
+            files[fp]['first_text'] = entry.original_text[:40].replace('\n', ' ')
+    
+    toc = []
+    for i, (fp, info) in enumerate(sorted(files.items()), 1):
+        toc.append({
+            'index': i,
+            'href': fp,
+            'title': info['first_text'] or f"File {i}",
+            'paragraphs': info['count'],
+            'chars': info['total_chars']
+        })
+    return toc
+
+
+def show_toc(manager: ManifestManager):
+    """Display TOC with chapter numbers."""
+    toc = build_toc_from_manifest(manager)
+    
+    print("\n" + "=" * 70)
+    print("章节目录 (Table of Contents)")
+    print("=" * 70)
+    print(f"{'#':>3} | {'段落':>5} | {'字符':>6} | 章节标题")
+    print("-" * 70)
+    
+    for item in toc:
+        title = item['title'][:45] if len(item['title']) > 45 else item['title']
+        print(f"{item['index']:3d} | {item['paragraphs']:5d} | {item['chars']:6d} | {title}")
+    
+    print("-" * 70)
+    print(f"共 {len(toc)} 个章节")
+    print("\n用法: python scripts/translate_chapter.py --chapter <编号>")
+    print("示例: python scripts/translate_chapter.py --chapter 5")
+
+
+def get_chapter_entries(manager: ManifestManager, chapter_index: int) -> tuple:
+    """Get entries for a specific chapter by index."""
+    toc = build_toc_from_manifest(manager)
+    
+    if chapter_index < 1 or chapter_index > len(toc):
+        print(f"错误: 章节编号 {chapter_index} 无效 (范围: 1-{len(toc)})")
+        return None, None
+    
+    chapter = toc[chapter_index - 1]
+    href = chapter['href']
+    
+    entries = [e for e in manager.entries if e.file_path == href]
+    return chapter, entries
+
+
+def create_char_based_chunks(entries: list, chunk_size: int = CHUNK_SIZE_CHARS) -> list:
+    """
+    Create chunks based on character count (~5000 chars each).
+    Returns list of entry lists.
+    """
+    chunks = []
+    current_chunk = []
+    current_size = 0
+    
+    for entry in entries:
+        text_len = len(entry.original_text)
+        
+        # If adding this entry exceeds limit and we have content, start new chunk
+        if current_size + text_len > chunk_size and current_chunk:
+            chunks.append(current_chunk)
+            current_chunk = []
+            current_size = 0
+        
+        current_chunk.append(entry)
+        current_size += text_len
+    
+    if current_chunk:
+        chunks.append(current_chunk)
+    
+    return chunks
+
+
+async def translate_chapter(chapter: dict, entries: list, manager: ManifestManager, 
+                           llm_client: LLMClient, profile, test_mode: bool = False):
+    """Translate a complete chapter."""
+    
+    print(f"\n开始翻译章节 #{chapter['index']}: {chapter['title'][:40]}...")
+    print(f"  文件: {chapter['href']}")
+    print(f"  总段落: {len(entries)}")
+    
+    # Filter untranslated
+    untranslated = [e for e in entries if not e.translated_text]
+    print(f"  待翻译: {len(untranslated)}")
+    
+    if not untranslated:
+        print("  ✅ 该章节已全部翻译!")
+        return
+    
+    # Create character-based chunks
+    chunks = create_char_based_chunks(untranslated)
+    print(f"  分块: {len(chunks)} 个 Chunk (约{CHUNK_SIZE_CHARS}字符/块)")
+    
+    if test_mode:
+        print("  [测试模式] 只翻译前2个 Chunk")
+        chunks = chunks[:2]
+    
+    # Show chunk stats
+    for i, chunk in enumerate(chunks, 1):
+        total_chars = sum(len(e.original_text) for e in chunk)
+        print(f"    Chunk {i}: {len(chunk)} 段落, {total_chars} 字符")
+    
+    # Translate
+    restorer = FormatRestorer()
+    total_success = 0
+    total_failed = 0
+    
+    for i, chunk in enumerate(chunks, 1):
+        chunk_chars = sum(len(e.original_text) for e in chunk)
+        print(f"\n  翻译 Chunk {i}/{len(chunks)} ({len(chunk)} 段, {chunk_chars} 字符)...")
+        
+        try:
+            results = await llm_client.translate_chunk(
+                chunk,
+                instruction=profile.style_guide if hasattr(profile, 'style_guide') else None,
+                mode="bilingual"
+            )
+            
+            # Apply results
+            chunk_success = 0
+            chunk_failed = 0
+            
+            for entry in chunk:
+                if entry.entry_id in results:
+                    translated = results[entry.entry_id]
+                    entry.translated_text = translated
+                    
+                    # Verify placeholder restoration
+                    if entry.placeholders:
+                        _, success = restorer.restore(translated, entry.placeholders)
+                        if success:
+                            chunk_success += 1
+                        else:
+                            chunk_failed += 1
+                            print(f"    ⚠️ 占位符还原警告: {entry.entry_id[-30:]}")
+                    else:
+                        chunk_success += 1
+                else:
+                    chunk_failed += 1
+                    print(f"    ❌ 缺失: {entry.entry_id[-30:]}")
+            
+            total_success += chunk_success
+            total_failed += chunk_failed
+            print(f"    ✓ 成功: {chunk_success}, 失败: {chunk_failed}")
+            
+            # Save after each chunk
+            manager.save()
+                    
+        except Exception as e:
+            print(f"    ❌ Chunk {i} 翻译失败: {e}")
+            total_failed += len(chunk)
+    
+    print(f"\n翻译完成:")
+    print(f"  ✅ 成功: {total_success}")
+    print(f"  ❌ 失败: {total_failed}")
+    
+    # Show sample results
+    print(f"\n翻译样例 (前3段):")
+    print("-" * 60)
+    translated_entries = [e for e in entries if e.translated_text][:3]
+    for entry in translated_entries:
+        orig = entry.original_text[:40].replace('\n', ' ')
+        trans = entry.translated_text[:40].replace('\n', ' ') if entry.translated_text else "(无)"
+        print(f"  原: {orig}...")
+        print(f"  译: {trans}...")
+        print()
+
+
+async def main():
+    parser = argparse.ArgumentParser(description="翻译指定章节")
+    parser.add_argument("--show-toc", action="store_true", help="显示章节目录")
+    parser.add_argument("--chapter", "-c", type=int, help="章节编号 (从1开始)")
+    parser.add_argument("--test", "-t", action="store_true", help="测试模式 (只翻译前2个chunk)")
+    args = parser.parse_args()
+    
+    if not MANIFEST_PATH.exists():
+        print(f"错误: Manifest 不存在: {MANIFEST_PATH}")
+        print("请先运行主管道生成 manifest。")
+        return
+    
+    # Load manifest
+    manager = ManifestManager(MANIFEST_PATH)
+    manager.load()
+    print(f"已加载 manifest: {len(manager.entries)} 条目")
+    
+    # Show TOC
+    if args.show_toc or not args.chapter:
+        show_toc(manager)
+        return
+    
+    if not API_KEY:
+        print("错误: V3_API_KEY 未设置")
+        return
+    
+    # Get chapter entries
+    chapter, entries = get_chapter_entries(manager, args.chapter)
+    if not chapter:
+        return
+    
+    # Initialize LLM client
+    from src.utils import ensure_directory
+    ensure_directory(CHUNK_DIR)
+    
+    llm_client = LLMClient(
+        api_key=API_KEY,
+        base_url=BASE_URL,
+        model=MODEL,
+        extra_headers=EXTRA_HEADERS,
+        chunk_dir=CHUNK_DIR
+    )
+    
+    try:
+        # Generate profile
+        print("\n生成书籍 Profile... (Skipping for debug)")
+        # profiler = BookProfiler(llm_client)
+        # profile = await profiler.analyze(manager.entries)
+        # print(f"  风格: {profile.style_guide[:80] if profile.style_guide else '(无)'}...")
+        
+        class DummyProfile:
+            style_guide = "Keep technical terms. Translate accurately."
+        profile = DummyProfile()
+        
+        # Translate chapter
+        await translate_chapter(chapter, entries, manager, llm_client, profile, args.test)
+        
+        print(f"\n✅ Manifest 已保存: {MANIFEST_PATH}")
+        print(f"✅ Chunk 文件保存在: {CHUNK_DIR}")
+        
+    finally:
+        await llm_client.close()
+
+
+if __name__ == "__main__":
+    print("DEBUG: Script started execution")
+    try:
+        asyncio.run(main())
+        print("DEBUG: Script finished execution")
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        print(f"CRITICAL ERROR: {e}")
@@ -0,0 +1,47 @@
+import sys
+from pathlib import Path
+import traceback
+
+# Add src to path
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from src.epub_cleaner import EpubCleaner
+from src.bilingual_builder import BilingualBuilder
+from src.utils import setup_logger
+from src.data_model import BookStructure
+
+logger = setup_logger("verify_toc")
+
+def verify_fix():
+    INPUT_EPUB = Path("input/Empire of AI Dreams and Nightmares in Sam Altmans OpenAI (Karen Hao).epub")
+    OUTPUT_EPUB = Path("output/verify_toc.epub")
+    WORK_DIR = Path("tmp/verify_toc")
+
+    if not INPUT_EPUB.exists():
+        logger.error(f"Input not found: {INPUT_EPUB}")
+        return
+
+    # Clean
+    logger.info("Step 1: Cleaning EPUB...")
+    cleaner = EpubCleaner(INPUT_EPUB, WORK_DIR)
+    json_path = cleaner.clean()
+    
+    # Load structure
+    with open(json_path, 'r') as f:
+        structure = BookStructure.model_validate_json(f.read())
+
+    # Build
+    logger.info("Step 2: Building EPUB with TOC preservation...")
+    builder = BilingualBuilder(WORK_DIR, original_epub_path=INPUT_EPUB)
+    builder.build(structure, OUTPUT_EPUB)
+    
+    logger.info(f"Step 3: EPUB generated at {OUTPUT_EPUB}")
+
+if __name__ == "__main__":
+    print("Starting verification script...")
+    try:
+        verify_fix()
+        print("Verification script finished.")
+    except Exception as e:
+        traceback.print_exc()
+        print(f"CRITICAL ERROR: {e}")
@@ -0,0 +1,96 @@
+from pathlib import Path
+from typing import List, Dict, Optional
+from bs4 import BeautifulSoup
+
+from src.data_model import BookStructure, ManifestEntry
+from src.format_restorer import FormatRestorer
+from src.utils import setup_logger
+
+logger = setup_logger("backfill_engine")
+
+class BackfillEngine:
+    """
+    Applies translations back to the BookStructure.
+    """
+    
+    def __init__(self):
+        self.restorer = FormatRestorer()
+        
+    def backfill(self, structure: BookStructure, manifest_entries: List[ManifestEntry], mode: str = "bilingual") -> BookStructure:
+        """
+        Modifies the BookStructure in-place with translations.
+        
+        Args:
+            structure: The BookStructure (from book_structure.json).
+            manifest_entries: List of translations.
+            mode: 'bilingual' or 'target_only'.
+        """
+        logger.info(f"Backfilling with mode: {mode}")
+        
+        # Index manifest by file and element ID for faster lookup
+        # Map: file_path -> element_id -> ManifestEntry
+        manifest_map: Dict[str, Dict[str, ManifestEntry]] = {}
+        for entry in manifest_entries:
+            if not entry.translated_text:
+                continue # Skip untranslated entries
+                
+            if entry.file_path not in manifest_map:
+                manifest_map[entry.file_path] = {}
+            manifest_map[entry.file_path][entry.element_id] = entry
+            
+        # Iterate resources in structure
+        for item_id, resource in structure.resources.items():
+            if resource.media_type != "application/xhtml+xml" or resource.href not in manifest_map:
+                continue
+                
+            file_entries = manifest_map[resource.href]
+            if not file_entries:
+                continue
+            
+            logger.debug(f"Processing {resource.href} with {len(file_entries)} translations")
+            
+            soup = BeautifulSoup(resource.content, 'html.parser')
+            modified = False
+            
+            for element_id, entry in file_entries.items():
+                element = soup.find(id=element_id)
+                if not element:
+                    logger.warning(f"Element {element_id} not found in {resource.href}")
+                    continue
+                
+                # Restore formatting
+                restored_html, _ = self.restorer.restore(entry.translated_text, entry.placeholders)
+                
+                # Create translated tag
+                new_tag = soup.new_tag(element.name)
+                # Parse restored HTML to get content nodes
+                inner_soup = BeautifulSoup(restored_html, 'html.parser')
+                if inner_soup.body:
+                    for child in list(inner_soup.body.children):
+                        new_tag.append(child)
+                else:
+                    for child in list(inner_soup.children):
+                        new_tag.append(child)
+                
+                # Copy classes and add 'translation'
+                classes = element.get('class', [])
+                if isinstance(classes, str):
+                    classes = classes.split()
+                new_tag['class'] = classes + ['translation']
+                
+                # Copy style
+                style = element.get('style')
+                if style:
+                    new_tag['style'] = style
+                
+                if mode == "bilingual":
+                    element.insert_after(new_tag)
+                else:
+                    element.replace_with(new_tag)
+                
+                modified = True
+                
+            if modified:
+                resource.content = str(soup)
+                
+        return structure
@@ -0,0 +1,252 @@
+import shutil
+import uuid
+from pathlib import Path
+from ebooklib import epub
+from src.data_model import BookStructure
+from src.utils import setup_logger, ensure_directory
+
+logger = setup_logger("bilingual_builder")
+
+class BilingualBuilder:
+    """
+    Assembles the final EPUB from BookStructure.
+    
+    Preserves the original TOC structure by reading it from the original EPUB.
+    """
+    
+    def __init__(self, work_dir: Path, original_epub_path: Path = None):
+        self.work_dir = work_dir
+        self.assets_dir = work_dir / "assets"
+        self.original_epub_path = original_epub_path
+        self._original_book = None
+        
+    def _load_original_book(self):
+        """Lazy load original book for TOC extraction."""
+        if self._original_book is None and self.original_epub_path and self.original_epub_path.exists():
+            try:
+                self._original_book = epub.read_epub(str(self.original_epub_path))
+                logger.debug(f"Loaded original EPUB for TOC: {self.original_epub_path}")
+            except Exception as e:
+                logger.warning(f"Failed to load original EPUB: {e}")
+        return self._original_book
+    
+    def _sanitize_toc(self, toc):
+        """
+        Ensure all TOC nodes have IDs (for ebooklib compatibility).
+        From v0.08 bilingual_builder.py
+        """
+        result = []
+        for item in toc:
+            if isinstance(item, (epub.Link, epub.Section)):
+                if not getattr(item, 'uid', None):
+                    item.uid = f"navPoint-{uuid.uuid4().hex[:8]}"
+                result.append(item)
+            elif isinstance(item, tuple) and len(item) == 2:
+                # Handle (Section, [children]) structure
+                section, children = item
+                if isinstance(section, (epub.Link, epub.Section)):
+                    if not getattr(section, 'uid', None):
+                        section.uid = f"navPoint-{uuid.uuid4().hex[:8]}"
+                sanitized_children = self._sanitize_toc(children)
+                result.append((section, sanitized_children))
+            else:
+                result.append(item)
+        return result
+        
+    def _validate_and_fix_toc(self, toc, book_items):
+        """
+        Recursively validate and fix TOC links.
+        Removes nodes with broken links that cannot be fixed.
+        """
+        fixed_toc = []
+        for item in toc:
+            if isinstance(item, (epub.Link, epub.Section)):
+                # Check href
+                href = getattr(item, 'href', '')
+                if href:
+                    # Remove anchor for check
+                    clean_href = href.split('#')[0]
+                    # Check if item exists in book (by file_name)
+                    found = False
+                    for existing_item in book_items.values():
+                        if existing_item.file_name == clean_href:
+                            found = True
+                            break
+                        if clean_href.endswith(existing_item.file_name) or existing_item.file_name.endswith(clean_href):
+                             item.href = existing_item.file_name
+                             found = True
+                             break
+                    
+                    if not found:
+                        if 'c0.xhtml' in clean_href:
+                            for existing_item in book_items.values():
+                                if 'titlepage' in existing_item.file_name or 'cover' in existing_item.file_name.lower():
+                                    if existing_item.media_type == "application/xhtml+xml":
+                                        logger.info(f"Fixed TOC link: {href} -> {existing_item.file_name}")
+                                        item.href = existing_item.file_name
+                                        found = True
+                                        break
+                        
+                        if not found:
+                             logger.warning(f"Removing broken TOC link: {href}")
+                             continue 
+            
+            if isinstance(item, tuple) and len(item) == 2:
+                section, children = item
+                fixed_children = self._validate_and_fix_toc(children, book_items)
+                fixed_toc.append((section, fixed_children))
+            else:
+                 fixed_toc.append(item)
+                 
+        return fixed_toc
+
+    def build(self, structure: BookStructure, output_path: Path) -> Path:
+        """
+        Builds the EPUB file.
+        Returns the path to the generated EPUB.
+        """
+        logger.info(f"Building final EPUB: {output_path}")
+        output_path.parent.mkdir(parents=True, exist_ok=True)
+        
+        book = epub.EpubBook()
+        
+        # 1. Metadata
+        book.set_identifier(structure.metadata.identifier or f"uuid-{uuid.uuid4().hex[:12]}")
+        book.set_title(structure.metadata.title)
+        book.set_language(structure.metadata.language)
+        book.add_author(structure.metadata.author)
+        
+        # 2. Copy TOC from original EPUB if available
+        original_book = self._load_original_book()
+        if original_book and hasattr(original_book, 'toc') and original_book.toc:
+            book.toc = self._sanitize_toc(original_book.toc)
+            logger.info("Copied TOC structure from original EPUB")
+            
+
+        
+        # 3. Add Resources - First pass: Collect CSS items
+        items_map = {} # id -> epub_item
+        css_items = [] # List of CSS EpubItem for linking
+        html_items = [] # List of (item_id, EpubHtml) tuples
+        
+        for item_id, resource in structure.resources.items():
+            # Skip NCX - we'll handle it separately
+            if resource.media_type == "application/x-dtbncx+xml":
+                continue
+                
+            if resource.media_type == "application/xhtml+xml":
+                # HTML Item - create but don't add yet (need to add CSS links)
+                item = epub.EpubHtml(
+                    uid=item_id,
+                    file_name=resource.href,
+                    media_type=resource.media_type,
+                    content=resource.content.encode('utf-8')
+                )
+                html_items.append((item_id, item))
+            elif resource.file_path:
+                # Binary/Asset Item
+                asset_full_path = self.work_dir / resource.file_path
+                
+                if not asset_full_path.exists():
+                     logger.warning(f"Asset missing: {asset_full_path}")
+                     continue
+                     
+                with open(asset_full_path, 'rb') as f:
+                    content = f.read()
+                    
+                if "image" in resource.media_type:
+                     item = epub.EpubImage(
+                        uid=item_id,
+                        file_name=resource.href,
+                        media_type=resource.media_type,
+                        content=content
+                     )
+                else:
+                    item = epub.EpubItem(
+                        uid=item_id,
+                        file_name=resource.href,
+                        media_type=resource.media_type,
+                        content=content
+                    )
+                    # Track CSS items
+                    if resource.media_type == "text/css":
+                        css_items.append(item)
+                        
+                book.add_item(item)
+                items_map[item_id] = item
+            else:
+                logger.warning(f"Skipping resource {item_id}: No content or file path.")
+                continue
+        
+        # 4. Add HTML items with CSS links
+        for item_id, item in html_items:
+            html_dir = Path(item.file_name).parent
+            for css_item in css_items:
+                css_path = Path(css_item.file_name)
+                # Calculate relative path from HTML directory to CSS file
+                try:
+                    relative_css_path = Path(css_path).relative_to(html_dir)
+                except ValueError:
+                    # Not a subpath, calculate full relative
+                    # Go up from html_dir, then down to css_path
+                    up_count = len(html_dir.parts)
+                    relative_css_path = Path("/".join([".."] * up_count)) / css_path
+                
+                item.add_link(href=str(relative_css_path), rel='stylesheet', type='text/css')
+            book.add_item(item)
+            items_map[item_id] = item
+
+        # 5. Copy missing items from original EPUB (cover, etc.)
+        # This ensures TOC links don't break
+        if original_book:
+            added_hrefs = {item.file_name for item in items_map.values() if hasattr(item, 'file_name')}
+            
+            for orig_item in original_book.get_items():
+                orig_name = orig_item.get_name()
+                if orig_name not in added_hrefs:
+                    # Skip NCX and NAV - we generate these
+                    if 'ncx' in orig_name.lower() or orig_name.endswith('nav.xhtml'):
+                        continue
+                    
+                    # Copy the item directly
+                    try:
+                        book.add_item(orig_item)
+                        items_map[orig_item.id] = orig_item
+                        logger.debug(f"Copied missing item from original: {orig_name}")
+                    except Exception as e:
+                        logger.warning(f"Failed to copy item {orig_name}: {e}")
+
+        # 6. Spine
+        book.spine = []
+        for item_id in structure.spine:
+             if item_id in items_map:
+                 book.spine.append(items_map[item_id])
+             else:
+                 logger.warning(f"Spine item {item_id} not found in resources.")
+        
+        # Add missing spine items from original
+        if original_book:
+            for spine_id, _ in original_book.spine:
+                if spine_id not in [i.id for i in book.spine]:
+                    orig_item = original_book.get_item_with_id(spine_id)
+                    if orig_item and orig_item.id in items_map:
+                        book.spine.append(items_map[orig_item.id])
+                 
+        # Validate and fix TOC
+        if original_book and hasattr(book, 'toc') and book.toc:
+             try:
+                 book.toc = self._validate_and_fix_toc(book.toc, items_map)
+                 logger.info("Validated and fixed TOC links")
+             except Exception as e:
+                 logger.error(f"Failed to validate TOC: {e}")
+
+        # 7. Navigation - NCX and Nav
+        book.add_item(epub.EpubNcx())
+        book.add_item(epub.EpubNav())
+        
+        # 7. Write
+        epub.write_epub(str(output_path), book)
+        logger.info(f"EPUB created successfully at {output_path}")
+        
+        return output_path
+
@@ -0,0 +1,76 @@
+import json
+import random
+from typing import Dict, List
+from loguru import logger
+from src.llm_client import LLMClient
+from src.manifest_manager import ManifestManager
+from src.data_model import BookProfile
+
+class BookProfiler:
+    def __init__(self, llm_client: LLMClient):
+        self.llm_client = llm_client
+
+    def extract_sample_text(self, entries: List[object], char_limit: int = 3000) -> str:
+        """Extract sample text from manifest entries."""
+        if not entries: return ""
+        
+        # Simple sampling strategy: First few + random middle
+        intro_text = []
+        for entry in entries[:50]:
+            if len(entry.original_text) > 50:
+                intro_text.append(entry.original_text)
+        
+        body_text = []
+        candidates = [e for e in entries[50:] if len(e.original_text) > 80]
+        if candidates:
+            samples = random.sample(candidates, min(5, len(candidates)))
+            body_text = [e.original_text for e in samples]
+            
+        full_text = "\n\n".join(intro_text[:5] + body_text)
+        return full_text[:char_limit]
+
+    async def analyze(self, entries: List[object]) -> BookProfile:
+        """Generate Book Profile."""
+        sample = self.extract_sample_text(entries)
+        if not sample: 
+            return BookProfile(title="Unknown", author="Unknown")
+        
+        logger.info("Generating Book Profile from sample text...")
+        
+        system_prompt = "You are a senior publishing editor. Analyze the text and output JSON."
+        user_prompt = f"""
+Please analyze the following book excerpt.
+Output JSON format:
+{{
+    "title": "Book Title",
+    "author": "Author Name",
+    "genre": "Genre",
+    "style": "Style description",
+    "keywords": ["keyword1", "keyword2"],
+    "style_guide": "Specific instruction for translator"
+}}
+
+Excerpt:
+{sample}
+"""
+        try:
+            response = await self.llm_client.raw_chat_completion(system_prompt, user_prompt)
+            json_str = response.strip()
+            # Basic cleanup
+            if "```json" in json_str:
+                json_str = json_str.split("```json")[1].split("```")[0].strip()
+            elif "```" in json_str:
+                json_str = json_str.split("```")[1].split("```")[0].strip()
+            
+            data = json.loads(json_str)
+            
+            return BookProfile(
+                title=data.get("title", "Unknown"),
+                author=data.get("author", "Unknown"),
+                genre=data.get("genre", "General"),
+                keywords=data.get("keywords", []),
+                style_guide=data.get("style_guide", "")
+            )
+        except Exception as e:
+            logger.error(f"Profile generation failed: {e}")
+            return BookProfile(title="Unknown", author="Unknown")
@@ -0,0 +1,51 @@
+from pathlib import Path
+from typing import List, Optional, Dict, Any
+from pydantic import BaseModel, Field
+
+class ManifestEntry(BaseModel):
+    """Represents a single translatable unit."""
+    entry_id: str = Field(..., description="Global unique ID (e.g. file.html#paragraph_id)")
+    file_path: str = Field(..., description="Internal path in EPUB")
+    element_id: str = Field(..., description="HTML ID (e.g. uuid-1234)")
+    original_text: str
+    placeholders: Dict[str, str] = Field(default_factory=dict)
+    translated_text: Optional[str] = None
+    context: Optional[str] = None
+
+class BookMetaData(BaseModel):
+    title: str = "Unknown Title"
+    author: str = "Unknown Author"
+    language: str = "en"
+    identifier: str = ""
+
+class ResourceItem(BaseModel):
+    href: str
+    media_type: str
+    content: Optional[str] = None  # For text/html
+    file_path: Optional[str] = None # For binary/assets (relative to assets dir)
+    properties: Optional[str] = None
+
+class BookStructure(BaseModel):
+    metadata: BookMetaData
+    spine: List[str] = Field(default_factory=list, description="Ordered list of item IDs in spine")
+    resources: Dict[str, ResourceItem] = Field(default_factory=dict, description="Map of item_id to ResourceItem")
+    
+    @classmethod
+    def load(cls, path: Path) -> "BookStructure":
+        """Load BookStructure from JSON file."""
+        with open(path, 'r', encoding='utf-8') as f:
+            return cls.model_validate_json(f.read())
+    
+    def save(self, path: Path):
+        """Save BookStructure to JSON file."""
+        with open(path, 'w', encoding='utf-8') as f:
+            f.write(self.model_dump_json(indent=2))
+
+class BookProfile(BaseModel):
+    """Represents the profile of the book."""
+    title: str
+    author: str
+    genre: str = "General"
+    keywords: List[str] = Field(default_factory=list)
+    style_guide: str = ""
+
@@ -0,0 +1,178 @@
+import shutil
+import uuid
+import zipfile
+from pathlib import Path
+from typing import Dict, Set
+
+import ebooklib
+from bs4 import BeautifulSoup, Tag
+from ebooklib import epub
+
+from src.data_model import BookStructure, BookMetaData, ResourceItem
+from src.exceptions import CleaningError
+from src.utils import setup_logger, ensure_directory
+
+logger = setup_logger("epub_cleaner")
+
+class EpubCleaner:
+    def __init__(self, input_path: Path, work_dir: Path):
+        self.input_path = input_path
+        self.work_dir = work_dir
+        self.assets_dir = work_dir / "assets"
+        self.json_path = work_dir / "book_structure.json"
+        
+    def clean(self) -> Path:
+        """
+        Cleans the input EPUB and generates book_structure.json.
+        Returns the path to the JSON file.
+        """
+        logger.info(f"Starting cleanup for {self.input_path}")
+        ensure_directory(self.work_dir)
+        ensure_directory(self.assets_dir)
+
+        try:
+            book = epub.read_epub(self.input_path)
+            
+            # 1. Extract Metadata
+            metadata = self._extract_metadata(book)
+            
+            # 2. Process Resources
+            resources = {}
+            # Use zipfile for binary extraction to avoid ebooklib's memory overhead/decoding issues
+            with zipfile.ZipFile(self.input_path, 'r') as zf:
+                # Map ebooklib items to zip entries isn't straightforward directly via name
+                # So we iterate ebooklib items and assume standard structure or handle content bytes
+                
+                for item in book.get_items():
+                    item_id = item.get_id()
+                    file_name = item.get_name()
+                    media_type = item.get_type() # ebooklib constant
+                    
+                    if media_type == ebooklib.ITEM_DOCUMENT:
+                        # Clean HTML
+                        content_str = item.get_content().decode('utf-8')
+                        cleaned_content = self._clean_html(content_str, file_name)
+                        
+                        resources[item_id] = ResourceItem(
+                            href=file_name,
+                            media_type="application/xhtml+xml",
+                            content=cleaned_content
+                        )
+                    elif media_type in (ebooklib.ITEM_IMAGE, ebooklib.ITEM_STYLE, ebooklib.ITEM_FONT, ebooklib.ITEM_COVER):
+                        # Save asset
+                        # Preserve directory structure to avoid collisions
+                        asset_path = self.assets_dir / file_name
+                        ensure_directory(asset_path.parent)
+                        
+                        # Ebooklib might change filenames, safer to use item.get_content()
+                        with open(asset_path, "wb") as f:
+                            f.write(item.get_content())
+                            
+                        resources[item_id] = ResourceItem(
+                            href=file_name,
+                            media_type=self._get_media_type_str(item),
+                            file_path=str(asset_path.relative_to(self.work_dir))
+                        )
+                    elif media_type == ebooklib.ITEM_NAVIGATION:
+                        # NCX or NAV document - preserve for TOC
+                        content_bytes = item.get_content()
+                        asset_path = self.assets_dir / file_name
+                        ensure_directory(asset_path.parent)
+                        
+                        with open(asset_path, "wb") as f:
+                            f.write(content_bytes)
+                        
+                        # Determine media type
+                        if file_name.endswith('.ncx'):
+                            mt = "application/x-dtbncx+xml"
+                        else:
+                            mt = "application/xhtml+xml"
+                        
+                        resources[item_id] = ResourceItem(
+                            href=file_name,
+                            media_type=mt,
+                            file_path=str(asset_path.relative_to(self.work_dir))
+                        )
+                        logger.debug(f"Preserved navigation: {file_name}")
+                    else:
+                        # Skip other items (scripts, etc.)
+                        pass
+
+            # 3. Extract Spine
+            spine_ids = [item[0] for item in book.spine]
+
+            # 4. Construct Structure
+            structure = BookStructure(
+                metadata=metadata,
+                spine=spine_ids,
+                resources=resources
+            )
+
+            # 5. Serialize
+            with open(self.json_path, "w", encoding="utf-8") as f:
+                f.write(structure.model_dump_json(indent=2))
+            
+            logger.info(f"Cleanup finished. Structure saved to {self.json_path}")
+            return self.json_path
+
+        except Exception as e:
+            logger.error(f"Cleaning failed: {e}")
+            raise CleaningError(f"Failed to clean EPUB: {e}") from e
+
+    def _extract_metadata(self, book: epub.EpubBook) -> BookMetaData:
+        title = book.get_metadata('DC', 'title')[0][0] if book.get_metadata('DC', 'title') else "Unknown"
+        author = book.get_metadata('DC', 'creator')[0][0] if book.get_metadata('DC', 'creator') else "Unknown"
+        lang = book.get_metadata('DC', 'language')[0][0] if book.get_metadata('DC', 'language') else "en"
+        ident = book.get_metadata('DC', 'identifier')[0][0] if book.get_metadata('DC', 'identifier') else ""
+        
+        return BookMetaData(
+            title=str(title),
+            author=str(author),
+            language=str(lang),
+            identifier=str(ident)
+        )
+
+    def _get_media_type_str(self, item) -> str:
+        # Helper to map ebooklib type to mime string if needed
+        # ebooklib doesn't expose easy MIME string for all types directly on item object sometimes
+        if hasattr(item, 'media_type'):
+            return item.media_type
+        return "application/octet-stream"
+
+    def _clean_html(self, content: str, filename: str) -> str:
+        soup = BeautifulSoup(content, 'html.parser')
+        
+        # 1. Provide IDs for structural/translatable elements
+        self._ensure_element_ids(soup)
+        
+        # 2. Flatten divs (Disable to prevent style loss)
+        # self._flatten_divs(soup)
+        
+        return str(soup)
+
+    def _ensure_element_ids(self, soup: BeautifulSoup):
+        """
+        Injects UUIDs into p, h1-h6, li tags if they don't have an ID.
+        This provides the anchor for translation backfilling.
+        """
+        targets = soup.find_all(['p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'li'])
+        for tag in targets:
+            if not tag.has_attr('id'):
+                tag['id'] = f"uuid-{uuid.uuid4()}"
+    
+    def _flatten_divs(self, soup: BeautifulSoup):
+        """
+        Converts generic divs containing only inline text/styles to p tags.
+        Recursive strategies can be complex, sticking to simple heuristic from archive.
+        """
+        inline_tags = {'span', 'em', 'strong', 'i', 'b', 'u', 'a', 'br', 'sub', 'sup', 'small'}
+        
+        for div in list(soup.find_all('div')):
+            # If div has no block children, convert to p
+            has_block = any(
+                isinstance(c, Tag) and c.name not in inline_tags 
+                for c in div.children
+            )
+            
+            if not has_block:
+                div.name = 'p'
@@ -0,0 +1,18 @@
+class EpubTranslatorError(Exception):
+    """Base exception for Epub Translator."""
+    pass
+
+class CleaningError(EpubTranslatorError):
+    pass
+
+class ExtractionError(EpubTranslatorError):
+    pass
+
+class TranslationError(EpubTranslatorError):
+    pass
+
+class RestorationError(EpubTranslatorError):
+    pass
+
+class BuildError(EpubTranslatorError):
+    pass
@@ -0,0 +1,132 @@
+import re
+from typing import List, Dict, Any, Optional
+from bs4 import BeautifulSoup
+from loguru import logger
+
+from src.data_model import BookStructure, ManifestEntry, BookProfile
+from src.format_extractor import FormatExtractor
+from src.utils import setup_logger
+
+logger = setup_logger("fine_grained_extractor")
+
+class FineGrainedExtractor:
+    """
+    Extracts translatable text segments from BookStructure.
+    Uses FormatExtractor for detailed content analysis.
+    """
+    
+    SKIP_TRANSLATION_PATTERNS = [
+        r'index\.x?html',
+        r'bibliography\.x?html',
+        r'endnotes?\.x?html',
+        r'footnotes?\.x?html',
+        r'copyright\.x?html',
+    ]
+    
+    TOC_PATTERNS = [
+        r'nav\.x?html',
+        r'toc\.x?html',
+    ]
+
+    def __init__(self, translate_toc: bool = False):
+        self.translate_toc = translate_toc
+        self.format_extractor = FormatExtractor()
+
+    def extract(self, structure: BookStructure, profile: Optional[BookProfile] = None) -> List[ManifestEntry]:
+        """
+        Extracts translatable segments from the BookStructure.
+        Iteration follows the spine order.
+        """
+        logger.info("Starting extraction from BookStructure...")
+        manifest_entries = []
+        
+        # Iterate over spine to maintain order
+        for item_id in structure.spine:
+            if item_id not in structure.resources:
+                logger.warning(f"Item ID {item_id} in spine but not in resources.")
+                continue
+            
+            resource = structure.resources[item_id]
+            
+            # Only process HTML/XHTML
+            if resource.media_type != "application/xhtml+xml" or not resource.content:
+                continue
+            
+            file_path = resource.href
+            doc_type = self._classify_document(file_path)
+            
+            # Parse HTML
+            soup = BeautifulSoup(resource.content, 'html.parser')
+            target_tags = ['p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'li']
+            
+            for element in soup.find_all(target_tags):
+                # Ensure element has ID (should have been done by Cleaner)
+                element_id = element.get('id')
+                if not element_id:
+                    logger.warning(f"Element in {file_path} missing ID, skipping: {element.name}")
+                    continue
+                
+                # Check translation eligibility
+                raw_text = element.get_text(separator=' ', strip=True)
+                if not raw_text.strip():
+                    continue
+                
+                is_decorative = self._is_decorative(raw_text)
+                should_translate = self._should_translate(doc_type, is_decorative, raw_text)
+                
+                if should_translate:
+                    # Extract detailed format
+                    outer_html = str(element)
+                    clean_text, text_with_ph, ph_map, p_type, _ = self.format_extractor.extract(outer_html)
+                    
+                    if clean_text.strip() and text_with_ph.strip():
+                        entry = ManifestEntry(
+                            entry_id=f"{file_path}#{element_id}",
+                            file_path=file_path,
+                            element_id=element_id,
+                            original_text=text_with_ph,
+                            placeholders=ph_map,
+                            context=p_type
+                        )
+                        manifest_entries.append(entry)
+        
+        logger.info(f"Extracted {len(manifest_entries)} entries in total.")
+        return manifest_entries
+
+    def _classify_document(self, file_name: str) -> str:
+        if not file_name: return 'core'
+        fname = file_name.lower()
+        if any(re.search(p, fname) for p in self.SKIP_TRANSLATION_PATTERNS): return 'skip'
+        if any(re.search(p, fname) for p in self.TOC_PATTERNS): return 'toc'
+        return 'core'
+
+    def _should_translate(self, doc_type: str, is_decorative: bool, text: str = "") -> bool:
+        if is_decorative: return False
+        if self._is_roman_numeral(text): return False
+        if doc_type == 'core': return True
+        if doc_type == 'toc': return self.translate_toc
+        return False
+
+    def _is_roman_numeral(self, text: str) -> bool:
+        text = text.strip().upper()
+        if not text: return False
+        pattern = re.compile(r"^M{0,4}(CM|CD|D?C{0,3})(XC|XL|L?X{0,3})(IX|IV|V?I{0,3})$")
+        return bool(pattern.match(text))
+
+    def _is_decorative(self, text: str) -> bool:
+        s = text.strip()
+        if not s: return False
+        if not any(c.isalnum() for c in s): return True
+        if len(s) > 20: return False
+        
+        patterns = [
+            r'^[*\-_~•◆◇■□▪▫●○◉◎⦿⦾]+$',
+            r'^[*\-_~•◆◇■□▪▫●○◉◎⦿⦾\s]+$',
+        ]
+        for p in patterns:
+            if re.match(p, s): return True
+            
+        unique = set(s.replace(' ', ''))
+        if len(unique) <= 3 and (unique & set('*-_~•◆◇■□▪▫●○◉◎⦿⦾')):
+            return True
+        return False
@@ -0,0 +1,372 @@
+import re
+from bs4 import BeautifulSoup, Tag
+from typing import Tuple, Dict, List
+
+class HeadingDetector:
+    """Detects heading types and paragraph roles."""
+    
+    CHAPTER_PATTERNS = [
+        r'^(chapter|chap\.?|part)\s+([0-9]+|[ivxlc]+|[a-z])',
+        r'^(第\s*[0-9一二三四五六七八九十百]+\s*[章节部篇])',
+        r'^(\d+|[IVXLC]+|[A-Z])\.$'
+    ]
+    
+    EPIGRAPH_CLASSES = {
+        'epigraph', 'quote', 'blockquote', 'motto', 
+        'dedication', 'verse', 'poetry', 'poem'
+    }
+
+    def detect(self, element: Tag, text: str) -> str:
+        if self._is_epigraph(element):
+            return "epigraph"
+        tag_name = element.name.lower()
+        if tag_name in ['h1', 'h2']:
+            return "chapter" if self._matches_chapter_pattern(text) else "section"
+        if tag_name == 'h3':
+            return "section"
+        if tag_name in ['h4', 'h5', 'h6']:
+            return "subsection"
+        if self._is_pseudo_heading(element, text):
+            return "subsection"
+        return "body"
+
+    def _is_epigraph(self, element: Tag) -> bool:
+        if element.name == 'blockquote':
+            return True
+        current = element
+        for _ in range(3):
+            if not current: break
+            classes = current.get('class', [])
+            if isinstance(classes, list):
+                classes = ' '.join(classes)
+            if any(k in classes.lower() for k in self.EPIGRAPH_CLASSES):
+                return True
+            current = current.parent
+        return False
+
+    def _matches_chapter_pattern(self, text: str) -> bool:
+        text = text.strip().lower()
+        for pattern in self.CHAPTER_PATTERNS:
+            if re.match(pattern, text, re.IGNORECASE):
+                return True
+        return False
+
+    def _is_pseudo_heading(self, element: Tag, text: str) -> bool:
+        if element.name != 'p':
+            return False
+        text = text.strip()
+        if not text or len(text) > 80:
+            return False
+        children = list(element.children)
+        if len(children) == 1 and isinstance(children[0], Tag):
+            if children[0].name in ['strong', 'b']:
+                return True
+        return False
+
+
+class FormatExtractor:
+    """
+    HTML Format Extractor (Ported from v0.09 v3)
+    Handles inline styles, formulas, and drop caps.
+    """
+
+    FORMULA_CHARS = re.compile(
+        r'^[\d\s\+\-\×\÷\=\(\)\[\]\{\}\<\>\^\*\/\.\,\;\:\'\"\`\~\@\#\$\%\&\|\\'
+        r'αβγδεζηθικλμνξοπρστυφχψωΑΒΓΔΕΖΗΘΙΚΛΜΝΞΟΠΡΣΤΥΦΧΨΩ'
+        r'a-zA-Z]+$'
+    )
+
+    def __init__(self):
+        self.detector = HeadingDetector()
+
+    def extract(self, element_html: str) -> Tuple[str, str, Dict[str, str], str, List[str]]:
+        """
+        Extracts format information.
+        
+        Returns:
+            clean_text: Pure text
+            text_with_placeholders: Text with inline placeholders
+            placeholder_map: Map of placeholders
+            paragraph_type: Detected type
+            endnote_anchors: List of detected endnote IDs
+        """
+        soup = BeautifulSoup(element_html, 'html.parser')
+        root = list(soup.children)[0] if list(soup.children) else soup
+        
+        clean_text = root.get_text().strip()
+        clean_text = re.sub(r'\s+', ' ', clean_text)
+        p_type = self.detector.detect(root, clean_text) if isinstance(root, Tag) else "body"
+        
+        inner_html = root.decode_contents() if isinstance(root, Tag) else str(root)
+        
+        text_with_ph, local_map = self._smart_extract_v3(inner_html)
+        
+        if text_with_ph:
+            text_with_ph = re.sub(r'\s+', ' ', text_with_ph).strip()
+        
+        # Verify integrity
+        stripped_text = self._strip_placeholders(text_with_ph)
+        stripped_text = re.sub(r'\s+', ' ', stripped_text).strip()
+        
+        if not self._verify_content_integrity(clean_text, stripped_text):
+            # Fallback
+            text_with_ph, local_map = self._fallback_extract(inner_html, clean_text)
+        
+        endnote_anchors = []
+        for pid, html in local_map.items():
+            if pid.startswith("_"):
+                continue
+            if re.match(r'<(span|a)\s+id="[a-zA-Z][a-zA-Z0-9]{2,5}"\s*>\s*</\1>', html):
+                endnote_anchors.append(pid)
+        
+        return clean_text, text_with_ph, local_map, p_type, endnote_anchors
+    
+    def _strip_placeholders(self, text: str) -> str:
+        return re.sub(r'φ/?[0-9]+φ', '', text)
+    
+    def _verify_content_integrity(self, clean_text: str, stripped_text: str) -> bool:
+        def normalize(s):
+            s = re.sub(r'\s+', '', s)
+            s = s.lower()
+            return s
+        
+        norm_clean = normalize(clean_text)
+        norm_stripped = normalize(stripped_text)
+        
+        if norm_clean == norm_stripped:
+            return True
+            
+        return False
+    
+    def _fallback_extract(self, inner_html: str, clean_text: str) -> Tuple[str, Dict[str, str]]:
+        return clean_text, {"_prefix": "", "_suffix": ""}
+
+    def _smart_extract_v3(self, inner_html: str) -> Tuple[str, Dict[str, str]]:
+        parts = re.split(r'(<[^>]+>)', inner_html)
+        parts = [p for p in parts if p]
+        
+        if not parts:
+            return "", {"_prefix": "", "_suffix": ""}
+        
+        part_types = []
+        for part in parts:
+            if part.startswith('<'):
+                part_types.append('tag')
+            elif not part.strip():
+                part_types.append('whitespace')
+            elif self._is_translatable_text(part):
+                part_types.append('translatable')
+            else:
+                part_types.append('formula')
+        
+        first_trans_idx = None
+        last_trans_idx = None
+        for i, t in enumerate(part_types):
+            if t == 'translatable':
+                if first_trans_idx is None:
+                    first_trans_idx = i
+                last_trans_idx = i
+        
+        if first_trans_idx is None:
+            return "", {"_prefix": inner_html, "_suffix": ""}
+        
+        # Prefix Separation
+        safe_prefix_end = 0
+        for i in range(first_trans_idx):
+            if part_types[i] == 'tag':
+                tag = parts[i]
+                is_self_closing = tag.endswith('/>') or re.match(r'<(br|hr|img|meta|link)\b', tag, re.I)
+                is_closing = tag.startswith('</')
+                if is_self_closing or is_closing:
+                    safe_prefix_end = i + 1
+                else:
+                    break
+            elif part_types[i] == 'whitespace':
+                safe_prefix_end = i + 1
+            else:
+                break
+        
+        # Suffix Separation
+        safe_suffix_start = len(parts)
+        for i in range(len(parts) - 1, last_trans_idx, -1):
+            if part_types[i] == 'tag':
+                tag = parts[i]
+                is_closing = tag.startswith('</')
+                is_self_closing = tag.endswith('/>') or re.match(r'<(br|hr|img|meta|link)\b', tag, re.I)
+                if is_closing or is_self_closing:
+                    safe_suffix_start = i
+                else:
+                    break
+            elif part_types[i] == 'whitespace':
+                safe_suffix_start = i
+            else:
+                break
+        
+        prefix_parts = parts[:safe_prefix_end]
+        middle_parts = parts[safe_prefix_end:safe_suffix_start]
+        middle_types = part_types[safe_prefix_end:safe_suffix_start]
+        suffix_parts = parts[safe_suffix_start:]
+        
+        # Drop Cap Check
+        if prefix_parts and middle_parts:
+            prefix_parts, middle_parts, middle_types = self._handle_drop_cap(
+                prefix_parts, middle_parts, middle_types
+            )
+        
+        local_map = {}
+        if prefix_parts:
+            local_map["_prefix"] = "".join(prefix_parts)
+        if suffix_parts:
+            local_map["_suffix"] = "".join(suffix_parts)
+        
+        # Middle processing
+        placeholder_counter = 1
+        result_parts = []
+        tag_stack = []
+        
+        i = 0
+        while i < len(middle_parts):
+            part = middle_parts[i]
+            ptype = middle_types[i]
+            
+            if ptype == 'translatable':
+                result_parts.append(part)
+                i += 1
+                
+            elif ptype == 'tag':
+                is_closing = part.startswith('</')
+                if is_closing:
+                    if tag_stack:
+                        open_id, open_tag = tag_stack.pop()
+                        local_map[f"/{open_id}"] = part
+                        result_parts.append(f"φ/{open_id}φ")
+                    else:
+                        pid = str(placeholder_counter)
+                        placeholder_counter += 1
+                        local_map[pid] = part
+                        result_parts.append(f"φ{pid}φ")
+                    i += 1
+                else:
+                    has_translatable_after = False
+                    for j in range(i + 1, len(middle_parts)):
+                        if middle_types[j] == 'translatable':
+                            has_translatable_after = True
+                            break
+                        elif middle_types[j] == 'tag' and middle_parts[j].startswith('</'):
+                            break
+                    
+                    if has_translatable_after:
+                        pid = str(placeholder_counter)
+                        placeholder_counter += 1
+                        local_map[pid] = part
+                        result_parts.append(f"φ{pid}φ")
+                        tag_stack.append((pid, part))
+                        i += 1
+                    else:
+                        block_parts = []
+                        while i < len(middle_parts) and middle_types[i] != 'translatable':
+                            block_parts.append(middle_parts[i])
+                            i += 1
+                        if block_parts:
+                            block_html = "".join(block_parts)
+                            pid = str(placeholder_counter)
+                            placeholder_counter += 1
+                            local_map[pid] = block_html
+                            result_parts.append(f"φ{pid}φ")
+            else:
+                result_parts.append(part)
+                i += 1
+        
+        text_with_ph = "".join(result_parts)
+        text_with_ph = re.sub(r'\s+', ' ', text_with_ph).strip()
+        
+        # Merge consecutive placeholders
+        def merge_match(m):
+            full_match = m.group(0)
+            pids = re.findall(r'φ(/?\d+)φ', full_match)
+            if len(pids) <= 1:
+                return full_match
+            
+            merged_html = ""
+            for pid in pids:
+                if pid in local_map:
+                    merged_html += local_map[pid]
+                    del local_map[pid]
+            
+            new_pid = pids[0] if pids[0].isdigit() else pids[0][1:]
+            local_map[new_pid] = merged_html
+            return f"φ{new_pid}φ"
+        
+        text_with_ph = re.sub(r'(φ/?\d+φ)(φ/?\d+φ)+', merge_match, text_with_ph)
+        
+        return text_with_ph, local_map
+
+    def _handle_drop_cap(self, prefix_parts: List[str], middle_parts: List[str], middle_types: List[str]):
+        if not prefix_parts or not middle_parts:
+            return prefix_parts, middle_parts, middle_types
+        
+        prefix_text = ""
+        for part in prefix_parts:
+            if not part.startswith('<'):
+                prefix_text = part.strip()
+        
+        if not prefix_text or len(prefix_text) != 1 or not prefix_text.isupper():
+            return prefix_parts, middle_parts, middle_types
+        
+        first_middle_text = ""
+        first_middle_idx = -1
+        for i, (part, ptype) in enumerate(zip(middle_parts, middle_types)):
+            if ptype == 'translatable':
+                first_middle_text = part.strip()
+                first_middle_idx = i
+                break
+        
+        if not first_middle_text:
+            return prefix_parts, middle_parts, middle_types
+        
+        is_drop_cap = False
+        first_char = first_middle_text[0] if first_middle_text else ''
+        if first_char.islower() or first_char.isupper():
+            is_drop_cap = True
+        
+        combined = prefix_text + first_middle_text.split()[0] if first_middle_text else ""
+        if not (len(combined) >= 2 and combined.isalpha()):
+            return prefix_parts, middle_parts, middle_types
+            
+        new_prefix = []
+        skip_until_close = False
+        found_letter = False
+        
+        for part in prefix_parts:
+            if part.startswith('<') and not part.startswith('</'):
+                skip_until_close = True
+            elif part.startswith('</'):
+                if skip_until_close:
+                    skip_until_close = False
+                    continue
+                new_prefix.append(part)
+            elif part.strip() == prefix_text:
+                found_letter = True
+                continue
+            else:
+                if not skip_until_close:
+                    new_prefix.append(part)
+        
+        if found_letter:
+            middle_parts = middle_parts.copy()
+            middle_parts[first_middle_idx] = prefix_text + middle_parts[first_middle_idx]
+            prefix_parts = new_prefix
+            
+        return prefix_parts, middle_parts, middle_types
+
+    def _is_translatable_text(self, text: str) -> bool:
+        text = text.strip()
+        if not text:
+            return False
+        if re.search(r'[a-zA-Z]{3,}', text):
+            return True
+        if ' ' in text and re.search(r'[a-zA-Z]', text):
+            return True
+        if re.search(r'\d', text):
+            return True
+        return False
@@ -0,0 +1,70 @@
+import re
+from typing import Dict, Tuple, List, Optional
+from loguru import logger
+from src.utils import setup_logger
+
+logger = setup_logger("format_restorer")
+
+class FormatRestorer:
+    """
+    Restores HTML formatting from placeholders.
+    """
+    
+    PLACEHOLDER_REGEX = re.compile(r'φ(/?\d+)φ')
+
+    def restore(self, text_with_placeholders: str, placeholder_map: Dict[str, str]) -> Tuple[str, bool]:
+        """
+        Restores HTML from text with placeholders.
+        Returns (restored_html, success).
+        """
+        if not placeholder_map:
+            return text_with_placeholders or "", True
+            
+        if not text_with_placeholders:
+            prefix = placeholder_map.get("_prefix", "")
+            suffix = placeholder_map.get("_suffix", "")
+            return prefix + suffix, True
+
+        prefix = placeholder_map.get("_prefix", "")
+        suffix = placeholder_map.get("_suffix", "")
+        
+        inner_map = {k: v for k, v in placeholder_map.items() if not k.startswith("_")}
+        
+        found_ids = set(self.PLACEHOLDER_REGEX.findall(text_with_placeholders))
+        expected_ids = set(inner_map.keys())
+        
+        success = True
+        missing_ids = expected_ids - found_ids
+        if missing_ids:
+            logger.warning(f"Restoration warning: missing placeholders {missing_ids}")
+            success = False
+            
+        unknown_ids = found_ids - expected_ids
+        if unknown_ids:
+            real_unknowns = set()
+            for pid in unknown_ids:
+                if pid.startswith('/') and pid[1:] in expected_ids:
+                    continue
+                real_unknowns.add(pid)
+                
+            if real_unknowns:
+                logger.warning(f"Restoration warning: unknown placeholders {real_unknowns}")
+                success = False
+
+        def replace_match(match):
+            pid = match.group(1)
+            if pid in inner_map:
+                return inner_map[pid]
+            else:
+                return ""
+
+        try:
+            restored_inner = self.PLACEHOLDER_REGEX.sub(replace_match, text_with_placeholders)
+            restored_html = prefix + restored_inner + suffix
+            return restored_html, success
+        except Exception as e:
+            logger.error(f"Restoration failed: {e}")
+            return prefix + self._strip_placeholders(text_with_placeholders) + suffix, False
+
+    def _strip_placeholders(self, text: str) -> str:
+        return self.PLACEHOLDER_REGEX.sub("", text)
@@ -0,0 +1,220 @@
+import asyncio
+import json
+import re
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import List, Dict, Optional
+from openai import AsyncOpenAI
+from tenacity import retry, stop_after_attempt, wait_exponential
+from loguru import logger
+
+from src.data_model import ManifestEntry
+from src.utils import setup_logger, ensure_directory
+
+logger = setup_logger("llm_client")
+
+# Default chunk save directory (can be overridden)
+DEFAULT_CHUNK_DIR = Path("tmp/chunks")
+
+
+class RateLimiter:
+    """Rate limiter for concurrency and RPM."""
+    def __init__(self, requests_per_minute: int, concurrent_requests: int):
+        self.semaphore = asyncio.Semaphore(concurrent_requests)
+        self.min_interval = 60.0 / requests_per_minute if requests_per_minute > 0 else 0
+        self.last_request_time = 0
+        self._lock = asyncio.Lock()
+
+    async def acquire(self):
+        await self.semaphore.acquire()
+        async with self._lock:
+            current_time = time.time()
+            wait_time = self.min_interval - (current_time - self.last_request_time)
+            if wait_time > 0:
+                await asyncio.sleep(wait_time)
+            self.last_request_time = time.time()
+
+    def release(self):
+        self.semaphore.release()
+
+
+class LLMClient:
+    """Generic OpenAI-compatible API Client with short ID strategy."""
+    
+    def __init__(self, api_key: str, base_url: str, model: str = "gpt-3.5-turbo",
+                 requests_per_minute: int = 60, concurrent_requests: int = 5,
+                 extra_headers: Dict = None, chunk_dir: Path = None):
+        
+        # Configure proxy client to avoid SOCKS issues and ensure connectivity
+        import httpx
+        import os
+        
+        # Prefer HTTP proxy if available to avoid missing socksio support
+        proxy_url = os.environ.get("http_proxy") or os.environ.get("https_proxy")
+        http_client = httpx.AsyncClient(
+            proxy=proxy_url,
+            timeout=60.0,
+            follow_redirects=True
+        ) if proxy_url else None
+
+        self.client = AsyncOpenAI(
+            base_url=base_url,
+            api_key=api_key,
+            default_headers=extra_headers,
+            http_client=http_client
+        )
+        self.model = model
+        
+        self.rate_limiter = RateLimiter(requests_per_minute, concurrent_requests)
+        self.prompts = self._load_prompts()
+        self._chunk_counter = 0
+        
+        # Chunk directory for debug output
+        self.chunk_dir = chunk_dir or DEFAULT_CHUNK_DIR
+        ensure_directory(self.chunk_dir)
+        
+    def _load_prompts(self) -> Dict:
+        try:
+            with open("config/prompts.json", "r", encoding="utf-8") as f:
+                return json.load(f)
+        except Exception as e:
+            logger.error(f"Failed to load config/prompts.json: {e}")
+            return {}
+
+    async def translate_chunk(self, items: List[ManifestEntry], glossary: Dict = None, 
+                              instruction: str = None, mode: str = "bilingual") -> Dict[str, str]:
+        """
+        Translate a chunk of items using short ID strategy.
+        Returns: Dict[entry_id, translated_text]
+        """
+        if not items: return {}
+        
+        # Build prompt with short IDs
+        id_map, prompt = self._build_prompt_with_short_ids(items)
+        
+        try:
+            # Build System Prompt
+            base_sys_prompt = self.prompts.get("translation", {}).get("system", 
+                "You are a professional English to Chinese translator.")
+            
+            if instruction:
+                base_sys_prompt += f"\n\nBook Style Guide:\n{instruction}"
+            
+            if glossary:
+                glossary_text = "\n".join([f"{k} -> {v}" for k, v in glossary.items()])
+                base_sys_prompt += f"\n\nTerminology:\n{glossary_text}"
+            
+            # Short ID format instructions
+            base_sys_prompt += """
+
+Output Format:
+- Each line MUST start with #N: (keep this ID exactly as given)
+- Preserve any φXφ or φ/Xφ placeholders EXACTLY as-is
+- Only output translations, no explanations
+- Match the number of output lines to input lines"""
+
+            # Save chunk before translation
+            chunk_id = self._save_chunk("before", prompt, base_sys_prompt)
+            
+            logger.debug(f"Sending request to LLM (Chunk: {chunk_id}, Items: {len(items)})")
+            
+            raw_response = await self._make_request(base_sys_prompt, prompt)
+            
+            # Save chunk after translation
+            self._save_chunk("after", raw_response, base_sys_prompt, chunk_id)
+            
+            # Parse with short ID mapping
+            results = self._parse_short_id_response(raw_response, id_map)
+            return results
+            
+        except Exception as e:
+            logger.error(f"Translation failed: {e}")
+            return {item.entry_id: f"[Error - {str(e)}]" for item in items}
+
+    def _build_prompt_with_short_ids(self, items: List[ManifestEntry]) -> tuple:
+        """
+        Build prompt with short IDs (#1, #2, ...).
+        Returns: (id_map, prompt_text)
+        """
+        id_map = {}  # short_id -> entry_id
+        lines = []
+        
+        for i, item in enumerate(items, 1):
+            short_id = f"#{i}"
+            id_map[short_id] = item.entry_id
+            
+            # Clean text (remove extra whitespace)
+            text = re.sub(r'\s+', ' ', item.original_text).strip()
+            lines.append(f"{short_id}: {text}")
+        
+        return id_map, "\n".join(lines)
+
+    def _parse_short_id_response(self, response: str, id_map: Dict[str, str]) -> Dict[str, str]:
+        """
+        Parse response with short ID format.
+        Returns: Dict[entry_id, translated_text]
+        """
+        results = {}
+        
+        for line in response.split("\n"):
+            line = line.strip()
+            if not line:
+                continue
+                
+            # Match #N: pattern
+            match = re.match(r'^#(\d+):\s*(.+)$', line)
+            if match:
+                short_id = f"#{match.group(1)}"
+                translation = match.group(2).strip()
+                
+                if short_id in id_map:
+                    full_id = id_map[short_id]
+                    results[full_id] = translation
+                else:
+                    logger.warning(f"Unknown short ID in response: {short_id}")
+        
+        return results
+
+    def _save_chunk(self, stage: str, content: str, system_prompt: str = None, 
+                    chunk_id: str = None) -> str:
+        """Save chunk to tmp directory for debugging."""
+        if chunk_id is None:
+            self._chunk_counter += 1
+            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+            chunk_id = f"{timestamp}_{self._chunk_counter:04d}"
+        
+        filename = self.chunk_dir / f"chunk_{chunk_id}_{stage}.txt"
+        
+        with open(filename, "w", encoding="utf-8") as f:
+            if system_prompt and stage == "before":
+                f.write("=== SYSTEM PROMPT ===\n")
+                f.write(system_prompt)
+                f.write("\n\n=== USER PROMPT ===\n")
+            f.write(content)
+        
+        logger.debug(f"Saved chunk: {filename}")
+        return chunk_id
+
+    async def raw_chat_completion(self, system_prompt: str, user_prompt: str) -> str:
+        """Generic chat completion."""
+        return await self._make_request(system_prompt, user_prompt)
+
+    @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=2, max=10))
+    async def _make_request(self, system_prompt: str, user_prompt: str) -> str:
+        await self.rate_limiter.acquire()
+        try:
+            resp = await self.client.chat.completions.create(
+                model=self.model,
+                messages=[
+                    {"role": "system", "content": system_prompt},
+                    {"role": "user", "content": user_prompt}
+                ],
+                temperature=0.3,
+            )
+            return resp.choices[0].message.content.strip()
+        finally:
+            self.rate_limiter.release()
+
+    async def close(self):
+        await self.client.close()
@@ -0,0 +1,74 @@
+from pathlib import Path
+from typing import List, Dict, Optional
+import json
+from src.data_model import ManifestEntry
+from src.utils import setup_logger
+
+logger = setup_logger("manifest_manager")
+
+class ManifestManager:
+    """
+    Manages the translation manifest (Source of Truth).
+    Handles persistence and state updates.
+    """
+    
+    def __init__(self, manifest_path: Path):
+        self.manifest_path = manifest_path
+        self.entries: List[ManifestEntry] = []
+        self._entries_map: Dict[str, ManifestEntry] = {}
+        
+    def load(self):
+        """Loads manifest from disk if it exists."""
+        if not self.manifest_path.exists():
+            logger.info(f"Manifest not found at {self.manifest_path}, starting empty.")
+            return
+
+        try:
+            with open(self.manifest_path, 'r', encoding='utf-8') as f:
+                data = json.load(f)
+                self.entries = [ManifestEntry.model_validate(item) for item in data]
+                self._rebuild_map()
+            logger.info(f"Loaded {len(self.entries)} entries from manifest.")
+        except Exception as e:
+            logger.error(f"Failed to load manifest: {e}")
+            raise
+
+    def save(self):
+        """Saves current state to disk."""
+        try:
+            # Pydantic v2: model_dump(mode='json') or just list dump
+            data = [entry.model_dump(mode='json') for entry in self.entries]
+            with open(self.manifest_path, 'w', encoding='utf-8') as f:
+                json.dump(data, f, indent=2, ensure_ascii=False)
+            logger.info(f"Saved {len(self.entries)} entries to manifest.")
+        except Exception as e:
+            logger.error(f"Failed to save manifest: {e}")
+            raise
+
+    def add_entries(self, new_entries: List[ManifestEntry]):
+        """
+        Adds new entries to the manifest.
+        If an entry with the same ID exists, it keeps the EXISTING one (to preserve translations).
+        """
+        count = 0
+        for entry in new_entries:
+            if entry.entry_id not in self._entries_map:
+                self.entries.append(entry)
+                self._entries_map[entry.entry_id] = entry
+                count += 1
+        
+        if count > 0:
+            logger.info(f"Added {count} new entries to manifest.")
+            
+    def update_translation(self, entry_id: str, translation: str):
+        """Updates translation for a specific entry."""
+        if entry_id in self._entries_map:
+            self._entries_map[entry_id].translated_text = translation
+        else:
+            logger.warning(f"Attempted to update translation for unknown ID: {entry_id}")
+
+    def get_entry(self, entry_id: str) -> Optional[ManifestEntry]:
+        return self._entries_map.get(entry_id)
+
+    def _rebuild_map(self):
+        self._entries_map = {e.entry_id: e for e in self.entries}
@@ -0,0 +1,285 @@
+"""
+Translator Module - Handles translation of ManifestEntry items.
+
+Key features:
+- Character-based chunking (~5000 chars per chunk)
+- Chapter-aware grouping (chunks don't cross file boundaries)
+- Concurrent translation with asyncio.gather
+- Progress tracking and error handling
+"""
+
+import asyncio
+from typing import List, Dict
+from collections import defaultdict
+from loguru import logger
+
+from src.data_model import ManifestEntry, BookProfile
+from src.llm_client import LLMClient
+from src.format_restorer import FormatRestorer
+from src.utils import setup_logger
+
+logger = setup_logger("translator")
+
+# Default chunk size in characters
+DEFAULT_CHUNK_SIZE = 5000
+# Maximum concurrent translations
+MAX_CONCURRENT = 5
+
+
+class Translator:
+    """
+    Translates ManifestEntry items using LLM with chapter-aware chunking.
+    Supports both sequential and concurrent translation modes.
+    """
+    
+    def __init__(self, llm_client: LLMClient, chunk_size: int = DEFAULT_CHUNK_SIZE, 
+                 max_concurrent: int = MAX_CONCURRENT):
+        self.llm_client = llm_client
+        self.chunk_size = chunk_size
+        self.max_concurrent = max_concurrent
+        self.restorer = FormatRestorer()
+        
+    async def translate(self, entries: List[ManifestEntry], profile: BookProfile, 
+                       concurrent: bool = True) -> List[ManifestEntry]:
+        """
+        Translates all untranslated entries.
+        
+        Args:
+            entries: All manifest entries
+            profile: Book profile with style guide
+            concurrent: Use concurrent translation (default True)
+            
+        Returns:
+            The same entries list with translated_text populated
+        """
+        untranslated = [e for e in entries if not e.translated_text]
+        if not untranslated:
+            logger.info("No new entries to translate.")
+            return entries
+            
+        logger.info(f"Found {len(untranslated)} entries to translate")
+        
+        # Group by chapter (file_path)
+        chapters = self._group_by_chapter(untranslated)
+        logger.info(f"Grouped into {len(chapters)} chapters")
+        
+        # Create all chunks
+        all_chunks = []
+        for file_path, chapter_entries in chapters.items():
+            chapter_chunks = self._create_char_based_chunks(chapter_entries)
+            for chunk in chapter_chunks:
+                all_chunks.append((file_path, chunk))
+        
+        total_chunks = len(all_chunks)
+        logger.info(f"Created {total_chunks} chunks (avg ~{self.chunk_size} chars each)")
+        
+        if concurrent:
+            await self._translate_concurrent(all_chunks, profile)
+        else:
+            await self._translate_sequential(all_chunks, profile)
+        
+        translated_count = sum(1 for e in entries if e.translated_text)
+        logger.info(f"Translation complete: {translated_count}/{len(entries)} entries translated")
+        return entries
+    
+    async def _translate_concurrent(self, all_chunks: List, profile: BookProfile):
+        """Translate chunks concurrently with semaphore control."""
+        semaphore = asyncio.Semaphore(self.max_concurrent)
+        completed = [0]  # Use list for mutable counter in closure
+        total = len(all_chunks)
+        success = [0]
+        failed = [0]
+        
+        async def translate_chunk_task(file_path: str, chunk: List[ManifestEntry], idx: int):
+            async with semaphore:
+                try:
+                    results = await self.llm_client.translate_chunk(
+                        chunk,
+                        instruction=profile.style_guide if profile else None,
+                        mode="bilingual"
+                    )
+                    
+                    for entry in chunk:
+                        if entry.entry_id in results:
+                            entry.translated_text = results[entry.entry_id]
+                            success[0] += 1
+                        else:
+                            failed[0] += 1
+                            
+                except Exception as e:
+                    logger.error(f"Chunk {idx} failed: {e}")
+                    failed[0] += len(chunk)
+                finally:
+                    completed[0] += 1
+                    if completed[0] % 5 == 0 or completed[0] == total:
+                        logger.info(f"Progress: {completed[0]}/{total} chunks ({success[0]} translated)")
+        
+        tasks = [
+            translate_chunk_task(file_path, chunk, i) 
+            for i, (file_path, chunk) in enumerate(all_chunks)
+        ]
+        await asyncio.gather(*tasks)
+        
+        logger.info(f"Concurrent translation: {success[0]} success, {failed[0]} failed")
+    
+    async def _translate_sequential(self, all_chunks: List, profile: BookProfile):
+        """Translate chunks sequentially."""
+        total_chunks = len(all_chunks)
+        success_count = 0
+        fail_count = 0
+        
+        for idx, (file_path, chunk) in enumerate(all_chunks):
+            try:
+                results = await self.llm_client.translate_chunk(
+                    chunk,
+                    instruction=profile.style_guide if profile else None,
+                    mode="bilingual"
+                )
+                
+                for entry in chunk:
+                    if entry.entry_id in results:
+                        entry.translated_text = results[entry.entry_id]
+                        success_count += 1
+                    else:
+                        fail_count += 1
+                        logger.warning(f"Missing translation for: {entry.entry_id[-40:]}")
+                        
+            except Exception as e:
+                logger.error(f"Chunk {idx} translation failed: {e}")
+                fail_count += len(chunk)
+            
+            if (idx + 1) % 10 == 0 or idx + 1 == total_chunks:
+                logger.info(f"Progress: {idx + 1}/{total_chunks} chunks ({success_count} entries translated)")
+        
+        logger.info(f"Sequential translation: {success_count} success, {fail_count} failed")
+
+    
+    async def translate_chapter(self, entries: List[ManifestEntry], file_path: str, 
+                                 profile: BookProfile) -> Dict[str, int]:
+        """
+        Translate a single chapter.
+        
+        Args:
+            entries: All entries (will filter by file_path)
+            file_path: Chapter file path to translate
+            profile: Book profile
+            
+        Returns:
+            Dict with 'success' and 'failed' counts
+        """
+        chapter_entries = [e for e in entries if e.file_path == file_path and not e.translated_text]
+        
+        if not chapter_entries:
+            logger.info(f"Chapter {file_path} has no untranslated entries")
+            return {"success": 0, "failed": 0}
+        
+        logger.info(f"Translating chapter: {file_path} ({len(chapter_entries)} entries)")
+        
+        chunks = self._create_char_based_chunks(chapter_entries)
+        logger.info(f"Created {len(chunks)} chunks")
+        
+        success = 0
+        failed = 0
+        
+        for i, chunk in enumerate(chunks, 1):
+            chunk_chars = sum(len(e.original_text) for e in chunk)
+            logger.debug(f"Chunk {i}/{len(chunks)}: {len(chunk)} entries, {chunk_chars} chars")
+            
+            try:
+                results = await self.llm_client.translate_chunk(
+                    chunk,
+                    instruction=profile.style_guide if profile else None,
+                    mode="bilingual"
+                )
+                
+                for entry in chunk:
+                    if entry.entry_id in results:
+                        entry.translated_text = results[entry.entry_id]
+                        
+                        # Verify placeholder preservation
+                        if entry.placeholders:
+                            _, restored_ok = self.restorer.restore(
+                                entry.translated_text, 
+                                entry.placeholders
+                            )
+                            if not restored_ok:
+                                logger.warning(f"Placeholder issue: {entry.entry_id[-40:]}")
+                        
+                        success += 1
+                    else:
+                        failed += 1
+                        
+            except Exception as e:
+                logger.error(f"Chunk {i} failed: {e}")
+                failed += len(chunk)
+        
+        logger.info(f"Chapter done: {success} success, {failed} failed")
+        return {"success": success, "failed": failed}
+    
+    def _group_by_chapter(self, entries: List[ManifestEntry]) -> Dict[str, List[ManifestEntry]]:
+        """Group entries by file_path (chapter)."""
+        chapters = defaultdict(list)
+        for entry in entries:
+            chapters[entry.file_path].append(entry)
+        return dict(chapters)
+    
+    def _create_char_based_chunks(self, entries: List[ManifestEntry]) -> List[List[ManifestEntry]]:
+        """
+        Create chunks based on character count.
+        
+        Each chunk contains approximately self.chunk_size characters.
+        Chunks never cross chapter boundaries (entries from same file only).
+        """
+        chunks = []
+        current_chunk = []
+        current_size = 0
+        
+        for entry in entries:
+            text_len = len(entry.original_text)
+            
+            # If adding this entry exceeds limit and we have content, start new chunk
+            if current_size + text_len > self.chunk_size and current_chunk:
+                chunks.append(current_chunk)
+                current_chunk = []
+                current_size = 0
+            
+            current_chunk.append(entry)
+            current_size += text_len
+        
+        if current_chunk:
+            chunks.append(current_chunk)
+        
+        return chunks
+    
+    def get_chapter_stats(self, entries: List[ManifestEntry]) -> List[Dict]:
+        """
+        Get statistics for each chapter.
+        
+        Returns list of dicts with: file_path, total, translated, pending, chars
+        """
+        chapters = self._group_by_chapter(entries)
+        stats = []
+        
+        for file_path, chapter_entries in sorted(chapters.items()):
+            total = len(chapter_entries)
+            translated = sum(1 for e in chapter_entries if e.translated_text)
+            total_chars = sum(len(e.original_text) for e in chapter_entries)
+            
+            # Get first text as title preview
+            first_text = ""
+            for e in chapter_entries:
+                if e.original_text:
+                    first_text = e.original_text[:40].replace('\n', ' ')
+                    break
+            
+            stats.append({
+                "file_path": file_path,
+                "title": first_text,
+                "total": total,
+                "translated": translated,
+                "pending": total - translated,
+                "chars": total_chars
+            })
+        
+        return stats
+
@@ -0,0 +1,25 @@
+import logging
+from pathlib import Path
+
+def setup_logger(name: str, log_file: Path = None, level=logging.INFO):
+    """Sets up a logger with the given name."""
+    logger = logging.getLogger(name)
+    logger.setLevel(level)
+    
+    formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+    
+    if log_file:
+        file_handler = logging.FileHandler(log_file)
+        file_handler.setFormatter(formatter)
+        logger.addHandler(file_handler)
+        
+    stream_handler = logging.StreamHandler()
+    stream_handler.setFormatter(formatter)
+    logger.addHandler(stream_handler)
+    
+    return logger
+
+def ensure_directory(path: Path):
+    """Ensures a directory exists."""
+    if not path.exists():
+        path.mkdir(parents=True, exist_ok=True)
@@ -0,0 +1 @@
+print("Hello from python")