Improve chart repair logic

@@ -1270,18 +1270,73 @@ class ChapterGenerationNode(BaseNode):
 
         normalized_cells: List[Dict[str, Any]] = []
         for cell in cell_entries:
+            # 检测错误嵌套的 cells 结构：有 cells 但没有 blocks
+            # 需要展平成多个独立的 cells
+            if isinstance(cell, dict) and "cells" in cell and "blocks" not in cell:
+                flattened = self._flatten_all_nested_cells(cell)
+                normalized_cells.extend(flattened)
+            else:
                 sanitized = self._normalize_table_cell(cell)
                 if sanitized:
                     normalized_cells.append(sanitized)
 
         return normalized_cells
 
+    def _flatten_all_nested_cells(self, cell: Dict[str, Any]) -> List[Dict[str, Any]]:
+        """
+        展平错误嵌套的 cells 结构，返回所有展平后的 cells。
+
+        LLM 有时会生成类似这样的错误结构：
+        { "cells": [
+            { "blocks": [...] },
+            { "cells": [
+                { "blocks": [...] },
+                { "cells": [...] }
+              ]
+            }
+          ]
+        }
+
+        应该展平为独立的 cells 列表。
+        """
+        nested_cells = cell.get("cells")
+        if not isinstance(nested_cells, list) or not nested_cells:
+            return [{"blocks": [self._as_paragraph_block("")]}]
+
+        result: List[Dict[str, Any]] = []
+        for nested in nested_cells:
+            if isinstance(nested, dict):
+                if "blocks" in nested and "cells" not in nested:
+                    # 正常的 cell，直接规范化添加
+                    sanitized = self._normalize_table_cell(nested)
+                    if sanitized:
+                        result.append(sanitized)
+                elif "cells" in nested and "blocks" not in nested:
+                    # 继续递归展平嵌套的 cells
+                    result.extend(self._flatten_all_nested_cells(nested))
+                else:
+                    # 其他情况，尝试规范化
+                    sanitized = self._normalize_table_cell(nested)
+                    if sanitized:
+                        result.append(sanitized)
+            elif isinstance(nested, (str, int, float)):
+                result.append({"blocks": [self._as_paragraph_block(str(nested))]})
+
+        return result if result else [{"blocks": [self._as_paragraph_block("")]}]
+
     def _normalize_table_cell(self, cell: Any) -> Dict[str, Any] | None:
         """把各种单元格写法规整为schema认可的形式"""
         if cell is None:
             return {"blocks": [self._as_paragraph_block("")]}
 
         if isinstance(cell, dict):
+            # 检测错误嵌套的 cells 结构：有 cells 但没有 blocks
+            # 这是 LLM 常见的错误，把同级 cell 嵌套进了 cells 数组
+            if "cells" in cell and "blocks" not in cell:
+                # 展平嵌套的 cells 并返回第一个有效 cell
+                # 注意：其余嵌套的 cells 会在 _normalize_table_cells 中被处理
+                return self._flatten_nested_cell(cell)
+
             normalized = dict(cell)
             blocks = self._coerce_cell_blocks(normalized.get("blocks"), normalized)
         elif isinstance(cell, list):
@@ -1297,6 +1352,40 @@ class ChapterGenerationNode(BaseNode):
         normalized["blocks"] = blocks or [self._as_paragraph_block("")]
         return normalized
 
+    def _flatten_nested_cell(self, cell: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        展平错误嵌套的 cell 结构。
+
+        LLM 有时会生成类似这样的错误结构：
+        { "cells": [ { "blocks": [...] }, { "cells": [...] } ] }
+
+        应该返回第一个有效的 cell 内容。
+        """
+        nested_cells = cell.get("cells")
+        if not isinstance(nested_cells, list) or not nested_cells:
+            # 没有有效的嵌套内容，返回空 cell
+            return {"blocks": [self._as_paragraph_block("")]}
+
+        # 递归查找第一个包含 blocks 的有效 cell
+        for nested in nested_cells:
+            if isinstance(nested, dict):
+                if "blocks" in nested:
+                    # 找到有效 cell，递归规范化
+                    return self._normalize_table_cell(nested)
+                elif "cells" in nested:
+                    # 继续递归展平
+                    result = self._flatten_nested_cell(nested)
+                    if result:
+                        return result
+
+        # 没有找到有效内容，尝试从第一个嵌套元素提取文本
+        first_nested = nested_cells[0]
+        if isinstance(first_nested, dict):
+            text = self._extract_block_text(first_nested)
+            return {"blocks": [self._as_paragraph_block(text or "")]}
+
+        return {"blocks": [self._as_paragraph_block("")]}
+
     def _coerce_cell_blocks(
         self, blocks: Any, source: Dict[str, Any] | None
     ) -> List[Dict[str, Any]]:

@@ -605,6 +605,8 @@ class MarkdownRenderer:
         elif isinstance(data_field, dict):
             if isinstance(data_field.get("items"), list):
                 candidates.append(data_field.get("items"))
+            if isinstance(data_field.get("words"), list):
+                candidates.append(data_field.get("words"))
 
         items: List[Dict[str, Any]] = []
         seen: set[str] = set()

+#!/usr/bin/env python3
+"""
+IR 文档验证工具。
+
+命令行工具，用于：
+- 扫描指定 JSON 文件中的所有图表和表格
+- 报告结构问题和数据缺失
+- 支持自动修复常见问题
+- 支持批量处理
+
+使用方法:
+    python -m ReportEngine.scripts.validate_ir chapter-030-section-3-0.json
+    python -m ReportEngine.scripts.validate_ir *.json --fix
+    python -m ReportEngine.scripts.validate_ir ./output/ --recursive --fix --verbose
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Tuple
+from dataclasses import dataclass, field
+
+# 添加项目根目录到路径
+project_root = Path(__file__).parent.parent.parent
+if str(project_root) not in sys.path:
+    sys.path.insert(0, str(project_root))
+
+from loguru import logger
+
+from ReportEngine.utils.chart_validator import (
+    ChartValidator,
+    ChartRepairer,
+    ValidationResult,
+)
+from ReportEngine.utils.table_validator import (
+    TableValidator,
+    TableRepairer,
+    TableValidationResult,
+)
+
+
+@dataclass
+class BlockIssue:
+    """单个 block 的问题"""
+    block_type: str
+    block_id: str
+    path: str
+    errors: List[str] = field(default_factory=list)
+    warnings: List[str] = field(default_factory=list)
+    is_fixable: bool = False
+
+
+@dataclass
+class DocumentReport:
+    """文档验证报告"""
+    file_path: str
+    total_blocks: int = 0
+    chart_count: int = 0
+    table_count: int = 0
+    wordcloud_count: int = 0
+    issues: List[BlockIssue] = field(default_factory=list)
+    fixed_count: int = 0
+
+    @property
+    def has_issues(self) -> bool:
+        return len(self.issues) > 0
+
+    @property
+    def error_count(self) -> int:
+        return sum(len(issue.errors) for issue in self.issues)
+
+    @property
+    def warning_count(self) -> int:
+        return sum(len(issue.warnings) for issue in self.issues)
+
+
+class IRValidator:
+    """IR 文档验证器"""
+
+    def __init__(
+        self,
+        chart_validator: Optional[ChartValidator] = None,
+        table_validator: Optional[TableValidator] = None,
+        chart_repairer: Optional[ChartRepairer] = None,
+        table_repairer: Optional[TableRepairer] = None,
+    ):
+        self.chart_validator = chart_validator or ChartValidator()
+        self.table_validator = table_validator or TableValidator()
+        self.chart_repairer = chart_repairer or ChartRepairer(self.chart_validator)
+        self.table_repairer = table_repairer or TableRepairer(self.table_validator)
+
+    def validate_document(
+        self,
+        document: Dict[str, Any],
+        file_path: str = "<unknown>",
+    ) -> DocumentReport:
+        """
+        验证整个文档。
+
+        Args:
+            document: IR 文档数据
+            file_path: 文件路径（用于报告）
+
+        Returns:
+            DocumentReport: 验证报告
+        """
+        report = DocumentReport(file_path=file_path)
+
+        # 遍历所有章节
+        chapters = document.get("chapters", [])
+        for chapter_idx, chapter in enumerate(chapters):
+            if not isinstance(chapter, dict):
+                continue
+
+            chapter_id = chapter.get("chapterId", f"chapter-{chapter_idx}")
+            blocks = chapter.get("blocks", [])
+
+            self._validate_blocks(
+                blocks,
+                f"chapters[{chapter_idx}].blocks",
+                chapter_id,
+                report,
+            )
+
+        return report
+
+    def _validate_blocks(
+        self,
+        blocks: List[Any],
+        path: str,
+        chapter_id: str,
+        report: DocumentReport,
+    ):
+        """递归验证 blocks 列表"""
+        if not isinstance(blocks, list):
+            return
+
+        for idx, block in enumerate(blocks):
+            if not isinstance(block, dict):
+                continue
+
+            report.total_blocks += 1
+            block_path = f"{path}[{idx}]"
+            block_type = block.get("type", "")
+            block_id = block.get("widgetId") or block.get("id") or f"block-{idx}"
+
+            # 根据类型验证
+            if block_type == "widget":
+                widget_type = (block.get("widgetType") or "").lower()
+                if "chart.js" in widget_type:
+                    report.chart_count += 1
+                    self._validate_chart(block, block_path, block_id, report)
+                elif "wordcloud" in widget_type:
+                    report.wordcloud_count += 1
+                    self._validate_wordcloud(block, block_path, block_id, report)
+
+            elif block_type == "table":
+                report.table_count += 1
+                self._validate_table(block, block_path, block_id, report)
+
+            # 递归处理嵌套 blocks
+            nested_blocks = block.get("blocks")
+            if isinstance(nested_blocks, list):
+                self._validate_blocks(nested_blocks, f"{block_path}.blocks", chapter_id, report)
+
+            # 处理 table rows 中的 blocks
+            if block_type == "table":
+                rows = block.get("rows", [])
+                for row_idx, row in enumerate(rows):
+                    if isinstance(row, dict):
+                        cells = row.get("cells", [])
+                        for cell_idx, cell in enumerate(cells):
+                            if isinstance(cell, dict):
+                                cell_blocks = cell.get("blocks", [])
+                                self._validate_blocks(
+                                    cell_blocks,
+                                    f"{block_path}.rows[{row_idx}].cells[{cell_idx}].blocks",
+                                    chapter_id,
+                                    report,
+                                )
+
+            # 处理 list items 中的 blocks
+            if block_type == "list":
+                items = block.get("items", [])
+                for item_idx, item in enumerate(items):
+                    if isinstance(item, list):
+                        self._validate_blocks(
+                            item,
+                            f"{block_path}.items[{item_idx}]",
+                            chapter_id,
+                            report,
+                        )
+
+    def _validate_chart(
+        self,
+        block: Dict[str, Any],
+        path: str,
+        block_id: str,
+        report: DocumentReport,
+    ):
+        """验证图表"""
+        result = self.chart_validator.validate(block)
+
+        if not result.is_valid or result.warnings:
+            issue = BlockIssue(
+                block_type="chart",
+                block_id=block_id,
+                path=path,
+                errors=result.errors,
+                warnings=result.warnings,
+                is_fixable=result.has_critical_errors(),
+            )
+            report.issues.append(issue)
+
+    def _validate_table(
+        self,
+        block: Dict[str, Any],
+        path: str,
+        block_id: str,
+        report: DocumentReport,
+    ):
+        """验证表格"""
+        result = self.table_validator.validate(block)
+
+        if not result.is_valid or result.warnings or result.nested_cells_detected:
+            issue = BlockIssue(
+                block_type="table",
+                block_id=block_id,
+                path=path,
+                errors=result.errors,
+                warnings=result.warnings,
+                is_fixable=result.nested_cells_detected or result.has_critical_errors(),
+            )
+
+            # 添加嵌套 cells 警告
+            if result.nested_cells_detected:
+                issue.warnings.insert(0, "检测到嵌套 cells 结构（LLM 常见错误）")
+
+            # 添加空单元格信息
+            if result.empty_cells_count > 0:
+                issue.warnings.append(
+                    f"空单元格数量: {result.empty_cells_count}/{result.total_cells_count}"
+                )
+
+            report.issues.append(issue)
+
+    def _validate_wordcloud(
+        self,
+        block: Dict[str, Any],
+        path: str,
+        block_id: str,
+        report: DocumentReport,
+    ):
+        """验证词云"""
+        errors: List[str] = []
+        warnings: List[str] = []
+
+        # 检查数据结构
+        data = block.get("data")
+        props = block.get("props", {})
+
+        words_found = False
+        words_count = 0
+
+        # 检查各种可能的词云数据路径
+        data_paths = [
+            ("data.words", data.get("words") if isinstance(data, dict) else None),
+            ("data.items", data.get("items") if isinstance(data, dict) else None),
+            ("data", data if isinstance(data, list) else None),
+            ("props.words", props.get("words") if isinstance(props, dict) else None),
+            ("props.items", props.get("items") if isinstance(props, dict) else None),
+            ("props.data", props.get("data") if isinstance(props, dict) else None),
+        ]
+
+        for path_name, value in data_paths:
+            if isinstance(value, list) and len(value) > 0:
+                words_found = True
+                words_count = len(value)
+
+                # 验证词云项格式
+                for idx, item in enumerate(value[:5]):  # 只检查前5个
+                    if isinstance(item, dict):
+                        word = item.get("word") or item.get("text") or item.get("label")
+                        weight = item.get("weight") or item.get("value")
+                        if not word:
+                            warnings.append(f"{path_name}[{idx}] 缺少 word/text/label 字段")
+                        if weight is None:
+                            warnings.append(f"{path_name}[{idx}] 缺少 weight/value 字段")
+                    elif not isinstance(item, (str, list, tuple)):
+                        warnings.append(f"{path_name}[{idx}] 格式不正确")
+
+                break
+
+        if not words_found:
+            errors.append("词云数据缺失：未在 data.words, data.items, props.words 等路径找到有效数据")
+        elif words_count == 0:
+            warnings.append("词云数据为空")
+
+        if errors or warnings:
+            issue = BlockIssue(
+                block_type="wordcloud",
+                block_id=block_id,
+                path=path,
+                errors=errors,
+                warnings=warnings,
+                is_fixable=False,  # 词云数据缺失通常无法自动修复
+            )
+            report.issues.append(issue)
+
+    def repair_document(
+        self,
+        document: Dict[str, Any],
+        report: DocumentReport,
+    ) -> Tuple[Dict[str, Any], int]:
+        """
+        修复文档中的问题。
+
+        Args:
+            document: IR 文档数据
+            report: 验证报告
+
+        Returns:
+            Tuple[Dict[str, Any], int]: (修复后的文档, 修复数量)
+        """
+        fixed_count = 0
+
+        # 遍历所有章节
+        chapters = document.get("chapters", [])
+        for chapter in chapters:
+            if not isinstance(chapter, dict):
+                continue
+
+            blocks = chapter.get("blocks", [])
+            chapter["blocks"], chapter_fixed = self._repair_blocks(blocks)
+            fixed_count += chapter_fixed
+
+        return document, fixed_count
+
+    def _repair_blocks(
+        self,
+        blocks: List[Any],
+    ) -> Tuple[List[Any], int]:
+        """递归修复 blocks 列表"""
+        if not isinstance(blocks, list):
+            return blocks, 0
+
+        fixed_count = 0
+        repaired_blocks: List[Any] = []
+
+        for block in blocks:
+            if not isinstance(block, dict):
+                repaired_blocks.append(block)
+                continue
+
+            block_type = block.get("type", "")
+
+            # 修复表格
+            if block_type == "table":
+                result = self.table_repairer.repair(block)
+                if result.has_changes():
+                    block = result.repaired_block
+                    fixed_count += 1
+                    logger.info(f"修复表格: {result.changes}")
+
+            # 修复图表
+            elif block_type == "widget":
+                widget_type = (block.get("widgetType") or "").lower()
+                if "chart.js" in widget_type:
+                    result = self.chart_repairer.repair(block)
+                    if result.has_changes():
+                        block = result.repaired_block
+                        fixed_count += 1
+                        logger.info(f"修复图表: {result.changes}")
+
+            # 递归处理嵌套 blocks
+            nested_blocks = block.get("blocks")
+            if isinstance(nested_blocks, list):
+                block["blocks"], nested_fixed = self._repair_blocks(nested_blocks)
+                fixed_count += nested_fixed
+
+            # 处理 table rows 中的 blocks
+            if block_type == "table":
+                rows = block.get("rows", [])
+                for row in rows:
+                    if isinstance(row, dict):
+                        cells = row.get("cells", [])
+                        for cell in cells:
+                            if isinstance(cell, dict):
+                                cell_blocks = cell.get("blocks", [])
+                                cell["blocks"], cell_fixed = self._repair_blocks(cell_blocks)
+                                fixed_count += cell_fixed
+
+            # 处理 list items 中的 blocks
+            if block_type == "list":
+                items = block.get("items", [])
+                for i, item in enumerate(items):
+                    if isinstance(item, list):
+                        items[i], item_fixed = self._repair_blocks(item)
+                        fixed_count += item_fixed
+
+            repaired_blocks.append(block)
+
+        return repaired_blocks, fixed_count
+
+
+def print_report(report: DocumentReport, verbose: bool = False):
+    """打印验证报告"""
+    print(f"\n{'=' * 60}")
+    print(f"文件: {report.file_path}")
+    print(f"{'=' * 60}")
+
+    print(f"\n📊 统计:")
+    print(f"  - 总 blocks: {report.total_blocks}")
+    print(f"  - 图表数量: {report.chart_count}")
+    print(f"  - 表格数量: {report.table_count}")
+    print(f"  - 词云数量: {report.wordcloud_count}")
+
+    if report.has_issues:
+        print(f"\n⚠️  发现 {len(report.issues)} 个问题:")
+        print(f"  - 错误: {report.error_count}")
+        print(f"  - 警告: {report.warning_count}")
+
+        if verbose:
+            for issue in report.issues:
+                print(f"\n  [{issue.block_type}] {issue.block_id}")
+                print(f"    路径: {issue.path}")
+                if issue.errors:
+                    for error in issue.errors:
+                        print(f"    ❌ {error}")
+                if issue.warnings:
+                    for warning in issue.warnings:
+                        print(f"    ⚠️  {warning}")
+                if issue.is_fixable:
+                    print(f"    🔧 可自动修复")
+    else:
+        print(f"\n✅ 未发现问题")
+
+    if report.fixed_count > 0:
+        print(f"\n🔧 已修复 {report.fixed_count} 个问题")
+
+
+def validate_file(
+    file_path: Path,
+    validator: IRValidator,
+    fix: bool = False,
+    verbose: bool = False,
+) -> DocumentReport:
+    """验证单个文件"""
+    try:
+        with open(file_path, "r", encoding="utf-8") as f:
+            document = json.load(f)
+    except json.JSONDecodeError as e:
+        logger.error(f"JSON 解析错误: {file_path}: {e}")
+        report = DocumentReport(file_path=str(file_path))
+        report.issues.append(BlockIssue(
+            block_type="document",
+            block_id="root",
+            path="",
+            errors=[f"JSON 解析错误: {e}"],
+        ))
+        return report
+    except Exception as e:
+        logger.error(f"读取文件错误: {file_path}: {e}")
+        report = DocumentReport(file_path=str(file_path))
+        report.issues.append(BlockIssue(
+            block_type="document",
+            block_id="root",
+            path="",
+            errors=[f"读取文件错误: {e}"],
+        ))
+        return report
+
+    # 验证文档
+    report = validator.validate_document(document, str(file_path))
+
+    # 修复问题
+    if fix and report.has_issues:
+        fixable_issues = [i for i in report.issues if i.is_fixable]
+        if fixable_issues:
+            logger.info(f"尝试修复 {len(fixable_issues)} 个问题...")
+            document, fixed_count = validator.repair_document(document, report)
+            report.fixed_count = fixed_count
+
+            if fixed_count > 0:
+                # 保存修复后的文件
+                backup_path = file_path.with_suffix(f".bak{file_path.suffix}")
+                try:
+                    # 创建备份
+                    import shutil
+                    shutil.copy(file_path, backup_path)
+                    logger.info(f"已创建备份: {backup_path}")
+
+                    # 保存修复后的文件
+                    with open(file_path, "w", encoding="utf-8") as f:
+                        json.dump(document, f, ensure_ascii=False, indent=2)
+                    logger.info(f"已保存修复后的文件: {file_path}")
+                except Exception as e:
+                    logger.error(f"保存文件失败: {e}")
+
+    return report
+
+
+def main():
+    """主函数"""
+    parser = argparse.ArgumentParser(
+        description="IR 文档验证工具",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        epilog="""
+示例:
+  %(prog)s chapter-030-section-3-0.json
+  %(prog)s *.json --fix
+  %(prog)s ./output/ --recursive --fix --verbose
+        """,
+    )
+    parser.add_argument(
+        "paths",
+        nargs="+",
+        help="要验证的 JSON 文件或目录",
+    )
+    parser.add_argument(
+        "-r", "--recursive",
+        action="store_true",
+        help="递归处理目录",
+    )
+    parser.add_argument(
+        "-f", "--fix",
+        action="store_true",
+        help="自动修复常见问题",
+    )
+    parser.add_argument(
+        "-v", "--verbose",
+        action="store_true",
+        help="显示详细信息",
+    )
+    parser.add_argument(
+        "--no-color",
+        action="store_true",
+        help="禁用彩色输出",
+    )
+
+    args = parser.parse_args()
+
+    # 配置日志
+    logger.remove()
+    if args.verbose:
+        logger.add(sys.stderr, level="DEBUG")
+    else:
+        logger.add(sys.stderr, level="INFO")
+
+    # 收集文件
+    files: List[Path] = []
+    for path_str in args.paths:
+        path = Path(path_str)
+        if path.is_file():
+            if path.suffix.lower() == ".json":
+                files.append(path)
+        elif path.is_dir():
+            if args.recursive:
+                files.extend(path.rglob("*.json"))
+            else:
+                files.extend(path.glob("*.json"))
+        else:
+            # 可能是 glob 模式
+            import glob
+            matched = glob.glob(path_str)
+            for m in matched:
+                mp = Path(m)
+                if mp.is_file() and mp.suffix.lower() == ".json":
+                    files.append(mp)
+
+    if not files:
+        print("未找到 JSON 文件")
+        sys.exit(1)
+
+    print(f"找到 {len(files)} 个文件")
+
+    # 创建验证器
+    validator = IRValidator()
+
+    # 验证文件
+    total_issues = 0
+    total_fixed = 0
+    reports: List[DocumentReport] = []
+
+    for file_path in files:
+        report = validate_file(file_path, validator, args.fix, args.verbose)
+        reports.append(report)
+        total_issues += len(report.issues)
+        total_fixed += report.fixed_count
+
+        if args.verbose or report.has_issues:
+            print_report(report, args.verbose)
+
+    # 打印总结
+    print(f"\n{'=' * 60}")
+    print("总结")
+    print(f"{'=' * 60}")
+    print(f"  - 文件数: {len(files)}")
+    print(f"  - 问题总数: {total_issues}")
+    if args.fix:
+        print(f"  - 已修复: {total_fixed}")
+
+    # 返回适当的退出码
+    if total_issues > 0 and total_fixed < total_issues:
+        sys.exit(1)
+    sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()

@@ -10,8 +10,23 @@ from ReportEngine.utils.chart_review_service import (
     review_document_charts,
 )
 
+from ReportEngine.utils.table_validator import (
+    TableValidator,
+    TableRepairer,
+    TableValidationResult,
+    TableRepairResult,
+    create_table_validator,
+    create_table_repairer,
+)
+
 __all__ = [
     "ChartReviewService",
     "get_chart_review_service",
     "review_document_charts",
+    "TableValidator",
+    "TableRepairer",
+    "TableValidationResult",
+    "TableRepairResult",
+    "create_table_validator",
+    "create_table_repairer",
 ]

@@ -83,6 +83,243 @@ CHART_REPAIR_SYSTEM_PROMPT = """你是一个专业的图表数据修复助手。
 """
 
 
+# 表格修复提示词
+TABLE_REPAIR_SYSTEM_PROMPT = """你是一个专业的表格数据修复助手。你的任务是修复IR表格数据中的格式错误，确保表格能够正常渲染。
+
+**标准表格数据格式：**
+
+```json
+{
+  "type": "table",
+  "rows": [
+    {
+      "cells": [
+        {
+          "header": true,
+          "blocks": [
+            {
+              "type": "paragraph",
+              "inlines": [{"text": "列标题", "marks": []}]
+            }
+          ]
+        },
+        {
+          "header": true,
+          "blocks": [
+            {
+              "type": "paragraph",
+              "inlines": [{"text": "另一列", "marks": []}]
+            }
+          ]
+        }
+      ]
+    },
+    {
+      "cells": [
+        {
+          "blocks": [
+            {
+              "type": "paragraph",
+              "inlines": [{"text": "数据内容", "marks": []}]
+            }
+          ]
+        },
+        {
+          "blocks": [
+            {
+              "type": "paragraph",
+              "inlines": [{"text": "另一数据", "marks": []}]
+            }
+          ]
+        }
+      ]
+    }
+  ]
+}
+```
+
+**⚠️ 常见错误：嵌套 cells 结构**
+
+这是一个非常常见的错误，LLM 经常把同级的 cells 错误地嵌套起来：
+
+❌ **错误示例：**
+```json
+{
+  "cells": [
+    { "blocks": [...], "colspan": 1 },
+    { "cells": [
+        { "blocks": [...] },
+        { "cells": [...] }
+      ]
+    }
+  ]
+}
+```
+
+✅ **正确格式：**
+```json
+{
+  "cells": [
+    { "blocks": [...], "colspan": 1 },
+    { "blocks": [...] },
+    { "blocks": [...] }
+  ]
+}
+```
+
+**修复原则：**
+1. **展平嵌套 cells** - 将错误嵌套的 cells 展平为同级
+2. **确保每个 cell 有 blocks** - 每个单元格必须有 blocks 数组
+3. **blocks 内使用 paragraph** - 文本内容应放在 paragraph block 内
+4. **保持数据完整性** - 不要丢失原始内容
+
+**修复方法：**
+1. 嵌套 cells 结构 → 展平为同级 cells 数组
+2. 缺少 blocks 字段 → 添加包含 paragraph 的 blocks
+3. 空 cells 数组 → 添加默认空单元格
+4. 非法 cell 类型 → 转换为标准格式
+
+请根据错误信息修复表格数据，并返回修复后的完整 table block（JSON格式）。
+"""
+
+
+# 词云修复提示词
+WORDCLOUD_REPAIR_SYSTEM_PROMPT = """你是一个专业的词云数据修复助手。你的任务是修复词云 widget 数据中的格式错误，确保词云能够正常渲染。
+
+**标准词云数据格式：**
+
+```json
+{
+  "type": "widget",
+  "widgetType": "wordcloud",
+  "widgetId": "wordcloud-001",
+  "title": "词云标题",
+  "data": {
+    "words": [
+      {"text": "关键词1", "weight": 10},
+      {"text": "关键词2", "weight": 8},
+      {"text": "关键词3", "weight": 6}
+    ]
+  }
+}
+```
+
+**⚠️ 数据路径说明：**
+
+词云数据可以位于以下路径（按优先级）：
+1. `data.words` - 推荐路径
+2. `data.items` - 备选路径
+3. `props.words` - 备选路径
+4. `props.items` - 备选路径
+5. `props.data` - 备选路径
+
+**词云项目格式：**
+
+每个词云项目应该是一个对象，包含：
+- `text` 或 `word` 或 `label`: 词语文本（必需）
+- `weight` 或 `value`: 权重/频率（必需）
+- `category`: 类别（可选）
+
+**修复原则：**
+1. **规范化数据路径** - 优先使用 `data.words`
+2. **确保必需字段** - 每个词项必须有文本和权重
+3. **转换兼容格式** - 将其他格式转换为标准格式
+4. **保持数据完整性** - 不要丢失原始词语
+
+**常见错误及修复方法：**
+1. 数据位于错误路径 → 移动到 `data.words`
+2. 缺少 weight 字段 → 根据位置生成默认权重
+3. 使用 word 而非 text → 统一为 text 字段
+4. 数组元素是字符串 → 转换为对象格式
+
+请根据错误信息修复词云数据，并返回修复后的完整 widget block（JSON格式）。
+"""
+
+
+def build_table_repair_prompt(
+    table_block: Dict[str, Any],
+    validation_errors: List[str]
+) -> str:
+    """
+    构建表格修复提示词。
+
+    Args:
+        table_block: 原始 table block
+        validation_errors: 验证错误列表
+
+    Returns:
+        str: 提示词
+    """
+    block_json = json.dumps(table_block, ensure_ascii=False, indent=2)
+    errors_text = "\n".join(f"- {error}" for error in validation_errors)
+
+    prompt = f"""请修复以下表格数据中的错误：
+
+**原始数据：**
+```json
+{block_json}
+```
+
+**检测到的错误：**
+{errors_text}
+
+**要求：**
+1. 返回修复后的完整 table block（JSON格式）
+2. 特别注意展平嵌套的 cells 结构
+3. 确保每个 cell 都有 blocks 数组
+4. 如果无法确定如何修复，保持原始数据
+
+**重要的输出格式要求：**
+1. 只返回纯JSON对象，不要添加任何说明文字
+2. 不要使用```json```标记包裹
+3. 确保JSON语法完全正确
+4. 所有字符串使用双引号
+"""
+    return prompt
+
+
+def build_wordcloud_repair_prompt(
+    widget_block: Dict[str, Any],
+    validation_errors: List[str]
+) -> str:
+    """
+    构建词云修复提示词。
+
+    Args:
+        widget_block: 原始 wordcloud widget block
+        validation_errors: 验证错误列表
+
+    Returns:
+        str: 提示词
+    """
+    block_json = json.dumps(widget_block, ensure_ascii=False, indent=2)
+    errors_text = "\n".join(f"- {error}" for error in validation_errors)
+
+    prompt = f"""请修复以下词云数据中的错误：
+
+**原始数据：**
+```json
+{block_json}
+```
+
+**检测到的错误：**
+{errors_text}
+
+**要求：**
+1. 返回修复后的完整 widget block（JSON格式）
+2. 确保词云数据位于 data.words 路径
+3. 每个词项必须有 text 和 weight 字段
+4. 如果无法确定如何修复，保持原始数据
+
+**重要的输出格式要求：**
+1. 只返回纯JSON对象，不要添加任何说明文字
+2. 不要使用```json```标记包裹
+3. 确保JSON语法完全正确
+4. 所有字符串使用双引号
+"""
+    return prompt
+
+
 def build_chart_repair_prompt(
     widget_block: Dict[str, Any],
     validation_errors: List[str]
@@ -283,3 +520,111 @@ def create_llm_repair_functions() -> List:
         logger.info(f"图表API修复功能已启用，共 {len(repair_functions)} 个Engine可用")
 
     return repair_functions
+
+
+def create_table_repair_functions() -> List:
+    """
+    创建表格 LLM 修复函数列表。
+
+    使用与图表修复相同的 Engine 配置。
+
+    Returns:
+        List[Callable]: 修复函数列表
+    """
+    repair_functions = []
+
+    # 使用 ReportEngine 修复表格
+    if settings.REPORT_ENGINE_API_KEY and settings.REPORT_ENGINE_BASE_URL:
+        def repair_table_with_report_engine(table_block: Dict[str, Any], errors: List[str]) -> Optional[Dict[str, Any]]:
+            """使用 ReportEngine 的 LLM 修复表格"""
+            try:
+                from ReportEngine.llms import LLMClient
+
+                client = LLMClient(
+                    api_key=settings.REPORT_ENGINE_API_KEY,
+                    base_url=settings.REPORT_ENGINE_BASE_URL,
+                    model_name=settings.REPORT_ENGINE_MODEL_NAME or "gpt-4",
+                )
+
+                prompt = build_table_repair_prompt(table_block, errors)
+                response = client.invoke(
+                    TABLE_REPAIR_SYSTEM_PROMPT,
+                    prompt,
+                    temperature=0.0,
+                    top_p=0.05
+                )
+
+                if not response:
+                    return None
+
+                # 解析响应
+                repaired = json.loads(response)
+                return repaired
+
+            except Exception as e:
+                logger.exception(f"ReportEngine 表格修复失败: {e}")
+                return None
+
+        repair_functions.append(repair_table_with_report_engine)
+        logger.debug("已添加 ReportEngine 表格修复函数")
+
+    if not repair_functions:
+        logger.warning("未配置任何 Engine API，表格 API 修复功能将不可用")
+    else:
+        logger.info(f"表格 API 修复功能已启用，共 {len(repair_functions)} 个 Engine 可用")
+
+    return repair_functions
+
+
+def create_wordcloud_repair_functions() -> List:
+    """
+    创建词云 LLM 修复函数列表。
+
+    使用与图表修复相同的 Engine 配置。
+
+    Returns:
+        List[Callable]: 修复函数列表
+    """
+    repair_functions = []
+
+    # 使用 ReportEngine 修复词云
+    if settings.REPORT_ENGINE_API_KEY and settings.REPORT_ENGINE_BASE_URL:
+        def repair_wordcloud_with_report_engine(widget_block: Dict[str, Any], errors: List[str]) -> Optional[Dict[str, Any]]:
+            """使用 ReportEngine 的 LLM 修复词云"""
+            try:
+                from ReportEngine.llms import LLMClient
+
+                client = LLMClient(
+                    api_key=settings.REPORT_ENGINE_API_KEY,
+                    base_url=settings.REPORT_ENGINE_BASE_URL,
+                    model_name=settings.REPORT_ENGINE_MODEL_NAME or "gpt-4",
+                )
+
+                prompt = build_wordcloud_repair_prompt(widget_block, errors)
+                response = client.invoke(
+                    WORDCLOUD_REPAIR_SYSTEM_PROMPT,
+                    prompt,
+                    temperature=0.0,
+                    top_p=0.05
+                )
+
+                if not response:
+                    return None
+
+                # 解析响应
+                repaired = json.loads(response)
+                return repaired
+
+            except Exception as e:
+                logger.exception(f"ReportEngine 词云修复失败: {e}")
+                return None
+
+        repair_functions.append(repair_wordcloud_with_report_engine)
+        logger.debug("已添加 ReportEngine 词云修复函数")
+
+    if not repair_functions:
+        logger.warning("未配置任何 Engine API，词云 API 修复功能将不可用")
+    else:
+        logger.info(f"词云 API 修复功能已启用，共 {len(repair_functions)} 个 Engine 可用")
+
+    return repair_functions

+"""
+表格验证和修复工具。
+
+提供对 IR 表格数据的验证和修复能力：
+1. 验证表格数据格式是否符合 IR schema 要求
+2. 检测嵌套 cells 结构问题
+3. 验证 rows/cells 基本格式
+4. 检查数据完整性
+5. 本地规则修复常见问题
+"""
+
+from __future__ import annotations
+
+import copy
+from typing import Any, Dict, List, Optional, Tuple
+from dataclasses import dataclass
+from loguru import logger
+
+
+@dataclass
+class TableValidationResult:
+    """表格验证结果"""
+    is_valid: bool
+    errors: List[str]
+    warnings: List[str]
+    nested_cells_detected: bool = False
+    empty_cells_count: int = 0
+    total_cells_count: int = 0
+
+    def has_critical_errors(self) -> bool:
+        """是否有严重错误（会导致渲染失败）"""
+        return not self.is_valid and len(self.errors) > 0
+
+
+@dataclass
+class TableRepairResult:
+    """表格修复结果"""
+    success: bool
+    repaired_block: Optional[Dict[str, Any]]
+    changes: List[str]
+
+    def has_changes(self) -> bool:
+        """是否有修改"""
+        return len(self.changes) > 0
+
+
+class TableValidator:
+    """
+    表格验证器 - 验证 IR 表格数据格式是否正确。
+
+    验证规则：
+    1. 基本结构验证：type, rows 字段
+    2. 行结构验证：每行必须有 cells 数组
+    3. 单元格结构验证：每个 cell 必须有 blocks 数组
+    4. 嵌套 cells 检测：检测错误的嵌套 cells 结构
+    5. 数据完整性验证：检查空单元格和缺失数据
+    """
+
+    def __init__(self):
+        """初始化验证器"""
+        pass
+
+    def validate(self, table_block: Dict[str, Any]) -> TableValidationResult:
+        """
+        验证表格格式。
+
+        Args:
+            table_block: table 类型的 block，包含 type, rows 等字段
+
+        Returns:
+            TableValidationResult: 验证结果
+        """
+        errors: List[str] = []
+        warnings: List[str] = []
+        nested_cells_detected = False
+        empty_cells_count = 0
+        total_cells_count = 0
+
+        # 1. 基本结构验证
+        if not isinstance(table_block, dict):
+            errors.append("table_block 必须是字典类型")
+            return TableValidationResult(
+                False, errors, warnings, nested_cells_detected,
+                empty_cells_count, total_cells_count
+            )
+
+        # 2. 检查 type
+        block_type = table_block.get('type')
+        if block_type != 'table':
+            errors.append(f"block type 应为 'table'，实际为 '{block_type}'")
+
+        # 3. 验证 rows 字段
+        rows = table_block.get('rows')
+        if rows is None:
+            errors.append("缺少 rows 字段")
+            return TableValidationResult(
+                False, errors, warnings, nested_cells_detected,
+                empty_cells_count, total_cells_count
+            )
+
+        if not isinstance(rows, list):
+            errors.append("rows 必须是数组类型")
+            return TableValidationResult(
+                False, errors, warnings, nested_cells_detected,
+                empty_cells_count, total_cells_count
+            )
+
+        if len(rows) == 0:
+            warnings.append("rows 数组为空，表格可能无法正常显示")
+
+        # 4. 验证每一行
+        for row_idx, row in enumerate(rows):
+            row_result = self._validate_row(row, row_idx)
+            errors.extend(row_result['errors'])
+            warnings.extend(row_result['warnings'])
+            if row_result['nested_cells_detected']:
+                nested_cells_detected = True
+            empty_cells_count += row_result['empty_cells_count']
+            total_cells_count += row_result['total_cells_count']
+
+        # 5. 检查列数一致性
+        column_counts = []
+        for row in rows:
+            if isinstance(row, dict):
+                cells = row.get('cells', [])
+                if isinstance(cells, list):
+                    col_count = 0
+                    for cell in cells:
+                        if isinstance(cell, dict):
+                            col_count += int(cell.get('colspan', 1))
+                        else:
+                            col_count += 1
+                    column_counts.append(col_count)
+
+        if column_counts and len(set(column_counts)) > 1:
+            warnings.append(
+                f"各行列数不一致: {column_counts}，可能导致渲染问题"
+            )
+
+        # 6. 空单元格警告
+        if total_cells_count > 0 and empty_cells_count > total_cells_count * 0.5:
+            warnings.append(
+                f"超过50%的单元格为空 ({empty_cells_count}/{total_cells_count})，"
+                "表格可能缺少数据"
+            )
+
+        is_valid = len(errors) == 0
+        return TableValidationResult(
+            is_valid, errors, warnings, nested_cells_detected,
+            empty_cells_count, total_cells_count
+        )
+
+    def _validate_row(self, row: Any, row_idx: int) -> Dict[str, Any]:
+        """验证单行"""
+        result = {
+            'errors': [],
+            'warnings': [],
+            'nested_cells_detected': False,
+            'empty_cells_count': 0,
+            'total_cells_count': 0,
+        }
+
+        if not isinstance(row, dict):
+            result['errors'].append(f"rows[{row_idx}] 必须是对象类型")
+            return result
+
+        cells = row.get('cells')
+        if cells is None:
+            result['errors'].append(f"rows[{row_idx}] 缺少 cells 字段")
+            return result
+
+        if not isinstance(cells, list):
+            result['errors'].append(f"rows[{row_idx}].cells 必须是数组类型")
+            return result
+
+        if len(cells) == 0:
+            result['warnings'].append(f"rows[{row_idx}].cells 数组为空")
+
+        # 验证每个单元格
+        for cell_idx, cell in enumerate(cells):
+            cell_result = self._validate_cell(cell, row_idx, cell_idx)
+            result['errors'].extend(cell_result['errors'])
+            result['warnings'].extend(cell_result['warnings'])
+            if cell_result['nested_cells_detected']:
+                result['nested_cells_detected'] = True
+            if cell_result['is_empty']:
+                result['empty_cells_count'] += 1
+            result['total_cells_count'] += 1
+
+        return result
+
+    def _validate_cell(self, cell: Any, row_idx: int, cell_idx: int) -> Dict[str, Any]:
+        """验证单个单元格"""
+        result = {
+            'errors': [],
+            'warnings': [],
+            'nested_cells_detected': False,
+            'is_empty': False,
+        }
+
+        if not isinstance(cell, dict):
+            result['errors'].append(
+                f"rows[{row_idx}].cells[{cell_idx}] 必须是对象类型"
+            )
+            return result
+
+        # 检测嵌套 cells 结构（这是常见的 LLM 错误）
+        if 'cells' in cell and 'blocks' not in cell:
+            result['nested_cells_detected'] = True
+            result['errors'].append(
+                f"rows[{row_idx}].cells[{cell_idx}] 检测到错误的嵌套 cells 结构，"
+                "应该是 blocks 而不是 cells"
+            )
+            return result
+
+        # 验证 blocks 字段
+        blocks = cell.get('blocks')
+        if blocks is None:
+            result['errors'].append(
+                f"rows[{row_idx}].cells[{cell_idx}] 缺少 blocks 字段"
+            )
+            return result
+
+        if not isinstance(blocks, list):
+            result['errors'].append(
+                f"rows[{row_idx}].cells[{cell_idx}].blocks 必须是数组类型"
+            )
+            return result
+
+        # 检查是否为空
+        if len(blocks) == 0:
+            result['is_empty'] = True
+        else:
+            # 检查 blocks 内容是否有效
+            has_content = False
+            for block in blocks:
+                if isinstance(block, dict):
+                    # 检查 paragraph 的 inlines
+                    if block.get('type') == 'paragraph':
+                        inlines = block.get('inlines', [])
+                        for inline in inlines:
+                            if isinstance(inline, dict):
+                                text = inline.get('text', '')
+                                if text and text.strip():
+                                    has_content = True
+                                    break
+                    # 检查其他类型的 text/content
+                    elif block.get('text') or block.get('content'):
+                        has_content = True
+                        break
+                if has_content:
+                    break
+
+            if not has_content:
+                result['is_empty'] = True
+
+        # 验证 colspan/rowspan
+        colspan = cell.get('colspan')
+        if colspan is not None:
+            if not isinstance(colspan, int) or colspan < 1:
+                result['warnings'].append(
+                    f"rows[{row_idx}].cells[{cell_idx}].colspan 值无效: {colspan}"
+                )
+
+        rowspan = cell.get('rowspan')
+        if rowspan is not None:
+            if not isinstance(rowspan, int) or rowspan < 1:
+                result['warnings'].append(
+                    f"rows[{row_idx}].cells[{cell_idx}].rowspan 值无效: {rowspan}"
+                )
+
+        return result
+
+    def can_render(self, table_block: Dict[str, Any]) -> bool:
+        """
+        判断表格是否能正常渲染（快速检查）。
+
+        Args:
+            table_block: table 类型的 block
+
+        Returns:
+            bool: 是否能正常渲染
+        """
+        result = self.validate(table_block)
+        return result.is_valid
+
+    def has_nested_cells(self, table_block: Dict[str, Any]) -> bool:
+        """
+        检测表格是否包含嵌套 cells 结构。
+
+        Args:
+            table_block: table 类型的 block
+
+        Returns:
+            bool: 是否包含嵌套 cells
+        """
+        result = self.validate(table_block)
+        return result.nested_cells_detected
+
+
+class TableRepairer:
+    """
+    表格修复器 - 尝试修复表格数据。
+
+    修复策略：
+    1. 展平嵌套 cells 结构
+    2. 补充缺失的 blocks 字段
+    3. 规范化单元格结构
+    4. 验证修复结果
+    """
+
+    def __init__(self, validator: Optional[TableValidator] = None):
+        """
+        初始化修复器。
+
+        Args:
+            validator: 表格验证器实例
+        """
+        self.validator = validator or TableValidator()
+
+    def repair(
+        self,
+        table_block: Dict[str, Any],
+        validation_result: Optional[TableValidationResult] = None
+    ) -> TableRepairResult:
+        """
+        尝试修复表格数据。
+
+        Args:
+            table_block: table 类型的 block
+            validation_result: 验证结果（可选，如果没有会先进行验证）
+
+        Returns:
+            TableRepairResult: 修复结果
+        """
+        # 1. 如果没有验证结果，先验证
+        if validation_result is None:
+            validation_result = self.validator.validate(table_block)
+
+        # 2. 如果已经有效，返回原数据
+        if validation_result.is_valid and not validation_result.nested_cells_detected:
+            return TableRepairResult(True, table_block, [])
+
+        # 3. 尝试修复
+        repaired = copy.deepcopy(table_block)
+        changes: List[str] = []
+
+        # 确保基本结构
+        if 'type' not in repaired:
+            repaired['type'] = 'table'
+            changes.append("添加缺失的 type 字段")
+
+        if 'rows' not in repaired or not isinstance(repaired.get('rows'), list):
+            repaired['rows'] = []
+            changes.append("添加缺失的 rows 字段")
+
+        # 修复每一行
+        repaired_rows: List[Dict[str, Any]] = []
+        for row_idx, row in enumerate(repaired.get('rows', [])):
+            repaired_row, row_changes = self._repair_row(row, row_idx)
+            repaired_rows.append(repaired_row)
+            changes.extend(row_changes)
+
+        repaired['rows'] = repaired_rows
+
+        # 4. 验证修复结果
+        repaired_validation = self.validator.validate(repaired)
+        success = repaired_validation.is_valid
+
+        if not success:
+            logger.warning(
+                f"表格修复后仍有问题: {repaired_validation.errors}"
+            )
+
+        return TableRepairResult(success, repaired, changes)
+
+    def _repair_row(
+        self, row: Any, row_idx: int
+    ) -> Tuple[Dict[str, Any], List[str]]:
+        """修复单行"""
+        changes: List[str] = []
+
+        if not isinstance(row, dict):
+            return {'cells': [self._default_cell()]}, [
+                f"rows[{row_idx}] 类型错误，已重建"
+            ]
+
+        repaired_row = dict(row)
+
+        # 确保有 cells 字段
+        if 'cells' not in repaired_row or not isinstance(repaired_row.get('cells'), list):
+            repaired_row['cells'] = [self._default_cell()]
+            changes.append(f"rows[{row_idx}] 添加缺失的 cells 字段")
+            return repaired_row, changes
+
+        # 修复每个单元格
+        repaired_cells: List[Dict[str, Any]] = []
+        for cell_idx, cell in enumerate(repaired_row.get('cells', [])):
+            if isinstance(cell, dict) and 'cells' in cell and 'blocks' not in cell:
+                # 展平嵌套 cells
+                flattened = self._flatten_nested_cells(cell)
+                repaired_cells.extend(flattened)
+                changes.append(
+                    f"rows[{row_idx}].cells[{cell_idx}] 展平嵌套 cells 结构"
+                )
+            else:
+                repaired_cell, cell_changes = self._repair_cell(cell, row_idx, cell_idx)
+                repaired_cells.append(repaired_cell)
+                changes.extend(cell_changes)
+
+        repaired_row['cells'] = repaired_cells
+        return repaired_row, changes
+
+    def _repair_cell(
+        self, cell: Any, row_idx: int, cell_idx: int
+    ) -> Tuple[Dict[str, Any], List[str]]:
+        """修复单个单元格"""
+        changes: List[str] = []
+
+        if not isinstance(cell, dict):
+            if isinstance(cell, (str, int, float)):
+                return {
+                    'blocks': [self._text_to_paragraph(str(cell))]
+                }, [f"rows[{row_idx}].cells[{cell_idx}] 转换为标准格式"]
+            return self._default_cell(), [
+                f"rows[{row_idx}].cells[{cell_idx}] 类型错误，已重建"
+            ]
+
+        repaired_cell = dict(cell)
+
+        # 确保有 blocks 字段
+        if 'blocks' not in repaired_cell:
+            # 尝试从其他字段提取内容
+            text = ''
+            for key in ('text', 'content', 'value'):
+                if key in repaired_cell and repaired_cell[key]:
+                    text = str(repaired_cell[key])
+                    break
+
+            repaired_cell['blocks'] = [self._text_to_paragraph(text or '')]
+            changes.append(
+                f"rows[{row_idx}].cells[{cell_idx}] 添加缺失的 blocks 字段"
+            )
+        elif not isinstance(repaired_cell['blocks'], list):
+            repaired_cell['blocks'] = [self._text_to_paragraph('')]
+            changes.append(
+                f"rows[{row_idx}].cells[{cell_idx}].blocks 类型错误，已重建"
+            )
+        elif len(repaired_cell['blocks']) == 0:
+            repaired_cell['blocks'] = [self._text_to_paragraph('')]
+            changes.append(
+                f"rows[{row_idx}].cells[{cell_idx}].blocks 为空，添加默认内容"
+            )
+
+        return repaired_cell, changes
+
+    def _flatten_nested_cells(self, cell: Dict[str, Any]) -> List[Dict[str, Any]]:
+        """展平嵌套的 cells 结构"""
+        nested_cells = cell.get('cells', [])
+        if not isinstance(nested_cells, list):
+            return [self._default_cell()]
+
+        result: List[Dict[str, Any]] = []
+        for nested in nested_cells:
+            if isinstance(nested, dict):
+                if 'blocks' in nested and 'cells' not in nested:
+                    # 正常的 cell
+                    result.append(nested)
+                elif 'cells' in nested and 'blocks' not in nested:
+                    # 继续递归展平
+                    result.extend(self._flatten_nested_cells(nested))
+                else:
+                    # 尝试修复
+                    repaired, _ = self._repair_cell(nested, 0, 0)
+                    result.append(repaired)
+            elif isinstance(nested, (str, int, float)):
+                result.append({
+                    'blocks': [self._text_to_paragraph(str(nested))]
+                })
+
+        return result if result else [self._default_cell()]
+
+    def _default_cell(self) -> Dict[str, Any]:
+        """创建默认单元格"""
+        return {
+            'blocks': [self._text_to_paragraph('')]
+        }
+
+    def _text_to_paragraph(self, text: str) -> Dict[str, Any]:
+        """将文本转换为 paragraph block"""
+        return {
+            'type': 'paragraph',
+            'inlines': [{'text': text, 'marks': []}]
+        }
+
+
+def create_table_validator() -> TableValidator:
+    """创建表格验证器实例"""
+    return TableValidator()
+
+
+def create_table_repairer(
+    validator: Optional[TableValidator] = None
+) -> TableRepairer:
+    """创建表格修复器实例"""
+    return TableRepairer(validator)
+
+
+__all__ = [
+    'TableValidator',
+    'TableRepairer',
+    'TableValidationResult',
+    'TableRepairResult',
+    'create_table_validator',
+    'create_table_repairer',
+]