Add File

2025-11-07 09:05:40 +08:00
parent 3229255f9a
commit 6f4841f86d
1 changed files with 351 additions and 0 deletions
--- a/src/summeryanyfile/generators/ppt_generator.py
+++ b/src/summeryanyfile/generators/ppt_generator.py
@@ -0,0 +1,351 @@
+"""
+PPT大纲生成器 - 主要的生成器类
+"""
+
+from typing import Dict, Any, Optional, Callable, List
+import logging
+
+from ..core.models import PPTState, PPTOutline, SlideInfo, ProcessingConfig, ChunkStrategy
+from ..core.document_processor import DocumentProcessor
+from ..core.llm_manager import LLMManager
+from ..generators.chains import ChainManager
+from ..graph.workflow import WorkflowManager, WorkflowExecutor
+from ..utils.logger import LoggerMixin, ProgressLogger
+
+logger = logging.getLogger(__name__)
+
+
+class PPTOutlineGenerator(LoggerMixin):
+    """基于迭代细化的PPT大纲生成器"""
+    
+    def __init__(self, config: ProcessingConfig, save_markdown: bool = False, temp_dir: Optional[str] = None,
+                 use_magic_pdf: bool = True, cache_dir: Optional[str] = None):
+        """
+        初始化PPT大纲生成器
+
+        Args:
+            config: 处理配置
+            save_markdown: 是否保存转换后的Markdown文件
+            temp_dir: 自定义temp目录路径
+            use_magic_pdf: 是否使用Magic-PDF处理PDF文件（本地处理，优先级高于MarkItDown）
+            cache_dir: 缓存目录路径
+        """
+        self.config = config
+        # 根据use_magic_pdf确定处理模式
+        processing_mode = "magic_pdf" if use_magic_pdf else "markitdown"
+
+        self.document_processor = DocumentProcessor(
+            save_markdown=save_markdown,
+            temp_dir=temp_dir,
+            use_magic_pdf=use_magic_pdf,
+            cache_dir=cache_dir,
+            processing_mode=processing_mode
+        )
+        self.llm_manager = LLMManager()
+
+        # 初始化LLM
+        self.llm = self._initialize_llm()
+
+        # 初始化处理链和工作流
+        self.chain_manager = ChainManager(self.llm)
+        self.workflow_manager = WorkflowManager(self.chain_manager, self.config)
+        self.workflow_executor = WorkflowExecutor(self.workflow_manager)
+
+        self.logger.info(f"PPT生成器初始化完成，使用模型: {config.llm_model}，缓存目录: {cache_dir}，处理模式: {processing_mode}")
+    
+    def _initialize_llm(self):
+        """初始化LLM"""
+        try:
+            return self.llm_manager.get_llm(
+                model=self.config.llm_model,
+                provider=self.config.llm_provider,
+                temperature=self.config.temperature,
+                max_tokens=self.config.max_tokens
+            )
+        except Exception as e:
+            self.logger.error(f"LLM初始化失败: {e}")
+            raise
+    
+    async def generate_from_text(
+        self,
+        text: str,
+        title: Optional[str] = None,
+        progress_callback: Optional[Callable[[str, float], None]] = None,
+        project_topic: str = "",
+        project_scenario: str = "general",
+        project_requirements: str = "",
+        target_audience: str = "普通大众",
+        custom_audience: str = "",
+        ppt_style: str = "general",
+        custom_style_prompt: str = "",
+        page_count_mode: str = "ai_decide",
+        min_pages: Optional[int] = None,
+        max_pages: Optional[int] = None,
+        fixed_pages: Optional[int] = None
+    ) -> PPTOutline:
+        """
+        从文本生成PPT大纲
+        
+        Args:
+            text: 输入文本
+            title: 可选的标题
+            progress_callback: 进度回调函数
+            
+        Returns:
+            PPT大纲对象
+        """
+        self.logger.info("开始从文本生成PPT大纲...")
+        
+        if not text.strip():
+            raise ValueError("输入文本不能为空")
+        
+        try:
+            # 文档分块
+            if progress_callback:
+                progress_callback("正在分析文档...", 5)
+            
+            chunks = self.document_processor.chunk_document(
+                text,
+                chunk_size=self.config.chunk_size,
+                chunk_overlap=self.config.chunk_overlap,
+                strategy=self.config.chunk_strategy,
+                max_tokens=self.config.max_tokens if self.config.chunk_strategy == ChunkStrategy.FAST else None
+            )
+            
+            self.logger.info(f"文档分块完成，共 {len(chunks)} 个块")
+            
+            # 准备初始状态
+            initial_state: PPTState = {
+                "document_chunks": chunks,
+                "current_index": 0,
+                "ppt_title": title or "",
+                "slides": [],
+                "total_pages": 0,
+                "page_count_mode": page_count_mode,
+                "document_structure": {},
+                "accumulated_context": "",
+                # 项目信息参数
+                "project_topic": project_topic,
+                "project_scenario": project_scenario,
+                "project_requirements": project_requirements,
+                "target_audience": target_audience,
+                "custom_audience": custom_audience,
+                "ppt_style": ppt_style,
+                "custom_style_prompt": custom_style_prompt,
+                # 页数设置参数
+                "min_pages": min_pages,
+                "max_pages": max_pages,
+                "fixed_pages": fixed_pages
+            }
+            
+            # 执行工作流
+            final_state = await self.workflow_executor.execute_with_monitoring(
+                initial_state,
+                progress_callback
+            )
+            
+            # 转换为PPT大纲对象
+            outline = self._state_to_outline(final_state)
+            
+            self.logger.info(f"PPT大纲生成完成，共 {outline.total_pages} 页")
+            return outline
+            
+        except Exception as e:
+            self.logger.error(f"PPT大纲生成失败: {e}")
+            raise
+    
+    async def generate_from_file(
+        self,
+        file_path: str,
+        encoding: Optional[str] = None,
+        progress_callback: Optional[Callable[[str, float], None]] = None,
+        project_topic: str = "",
+        project_scenario: str = "general",
+        project_requirements: str = "",
+        target_audience: str = "普通大众",
+        custom_audience: str = "",
+        ppt_style: str = "general",
+        custom_style_prompt: str = "",
+        page_count_mode: str = "ai_decide",
+        min_pages: Optional[int] = None,
+        max_pages: Optional[int] = None,
+        fixed_pages: Optional[int] = None
+    ) -> PPTOutline:
+        """
+        从文件生成PPT大纲
+
+        Args:
+            file_path: 文件路径
+            encoding: 文件编码
+            progress_callback: 进度回调函数
+
+        Returns:
+            PPT大纲对象
+        """
+        self.logger.info(f"开始从文件生成PPT大纲: {file_path}")
+
+        try:
+            # 加载文档（在线程池中执行以避免阻塞主服务）
+            if progress_callback:
+                progress_callback("正在加载文档...", 2)
+
+            import asyncio
+            loop = asyncio.get_running_loop()
+            document_info = await loop.run_in_executor(
+                None,
+                self.document_processor.load_document,
+                file_path,
+                encoding
+            )
+
+            self.logger.info(f"文档加载完成: {document_info.title}")
+
+            # 从文本生成
+            return await self.generate_from_text(
+                document_info.content,
+                document_info.title,
+                progress_callback,
+                project_topic,
+                project_scenario,
+                project_requirements,
+                target_audience,
+                custom_audience,
+                ppt_style,
+                custom_style_prompt,
+                page_count_mode,
+                min_pages,
+                max_pages,
+                fixed_pages
+            )
+
+        except Exception as e:
+            self.logger.error(f"从文件生成PPT大纲失败: {e}")
+            raise
+    
+    def _state_to_outline(self, state: Dict[str, Any]) -> PPTOutline:
+        """将状态转换为PPT大纲对象"""
+        slides = []
+        
+        for slide_data in state.get("slides", []):
+            slide = SlideInfo.from_dict(slide_data)
+            slides.append(slide)
+        
+        return PPTOutline(
+            title=state.get("ppt_title", "PPT大纲"),
+            total_pages=state.get("total_pages", len(slides)),
+            page_count_mode=state.get("page_count_mode", "final"),
+            slides=slides
+        )
+    
+    def update_config(self, config: ProcessingConfig):
+        """
+        更新配置
+        
+        Args:
+            config: 新的处理配置
+        """
+        self.logger.info("更新生成器配置...")
+        
+        # 检查是否需要重新初始化LLM
+        if (config.llm_model != self.config.llm_model or 
+            config.llm_provider != self.config.llm_provider or
+            config.temperature != self.config.temperature or
+            config.max_tokens != self.config.max_tokens):
+            
+            self.config = config
+            self.llm = self._initialize_llm()
+            self.chain_manager.update_llm(self.llm)
+            # 重新创建工作流管理器以传递新配置
+            self.workflow_manager = WorkflowManager(self.chain_manager, self.config)
+        else:
+            self.config = config
+    
+    def get_supported_formats(self) -> List[str]:
+        """获取支持的文件格式"""
+        return list(self.document_processor.SUPPORTED_EXTENSIONS.keys())
+    
+    def validate_input(self, text: str) -> bool:
+        """
+        验证输入文本
+        
+        Args:
+            text: 输入文本
+            
+        Returns:
+            是否有效
+        """
+        if not text or not isinstance(text, str):
+            return False
+        
+        # 检查最小长度
+        if len(text.strip()) < 100:
+            self.logger.warning("输入文本太短，可能影响生成质量")
+            return False
+        
+        # 检查最大长度
+        max_length = self.config.chunk_size * 50  # 允许最多50个块
+        if len(text) > max_length:
+            self.logger.warning(f"输入文本太长 ({len(text)} 字符)，建议分段处理")
+        
+        return True
+    
+    async def generate_with_retry(
+        self,
+        text: str,
+        title: Optional[str] = None,
+        max_retries: int = 3,
+        progress_callback: Optional[Callable[[str, float], None]] = None,
+        project_topic: str = "",
+        project_scenario: str = "general",
+        project_requirements: str = "",
+        target_audience: str = "普通大众",
+        custom_audience: str = "",
+        ppt_style: str = "general",
+        custom_style_prompt: str = ""
+    ) -> PPTOutline:
+        """
+        带重试的生成方法
+        
+        Args:
+            text: 输入文本
+            title: 可选标题
+            max_retries: 最大重试次数
+            progress_callback: 进度回调
+            
+        Returns:
+            PPT大纲对象
+        """
+        last_exception = None
+        
+        for attempt in range(max_retries):
+            try:
+                if attempt > 0:
+                    self.logger.info(f"第 {attempt + 1} 次尝试生成PPT大纲...")
+                
+                return await self.generate_from_text(
+                    text, title, progress_callback,
+                    project_topic, project_scenario, project_requirements,
+                    target_audience, custom_audience, ppt_style, custom_style_prompt
+                )
+                
+            except Exception as e:
+                last_exception = e
+                self.logger.warning(f"第 {attempt + 1} 次尝试失败: {e}")
+                
+                if attempt < max_retries - 1:
+                    # 可以在这里添加一些恢复策略
+                    import asyncio
+                    await asyncio.sleep(2 ** attempt)  # 指数退避
+        
+        self.logger.error(f"所有 {max_retries} 次尝试都失败了")
+        raise last_exception
+    
+    def get_generation_stats(self) -> Dict[str, Any]:
+        """获取生成统计信息"""
+        return {
+            "config": self.config.to_dict(),
+            "llm_model": self.config.llm_model,
+            "llm_provider": self.config.llm_provider,
+            "supported_formats": self.get_supported_formats(),
+            "workflow_info": self.workflow_manager.get_workflow_info()
+        }