Add File

src/landppt/services/image/adapters/ppt_prompt_adapter.py

@@ -0,0 +1,345 @@
+"""
+PPT内容到图片生成提示词的适配器
+将PPT幻灯片内容转换为适合AI图片生成的提示词
+"""
+
+import logging
+import re
+from typing import Dict, Any, List, Optional, Tuple
+from dataclasses import dataclass
+
+from ..models import ImageGenerationRequest, ImageProvider
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class PPTSlideContext:
+    """PPT幻灯片上下文信息"""
+    title: str
+    content: str
+    scenario: str
+    topic: str
+    page_number: int
+    total_pages: int
+    slide_type: str = "content"
+    language: str = "zh"
+
+
+class PPTPromptAdapter:
+    """PPT内容到图片生成提示词的适配器"""
+    
+    def __init__(self, config: Dict[str, Any] = None):
+        self.config = config or {}
+        
+        # 场景到视觉风格的映射
+        self.scenario_styles = {
+            "general": "professional, clean, modern",
+            "business": "corporate, professional, sleek, business meeting",
+            "education": "educational, academic, classroom, learning",
+            "technology": "futuristic, digital, high-tech, innovation",
+            "tourism": "scenic, beautiful, travel, destination",
+            "analysis": "data visualization, charts, analytics, research",
+            "history": "historical, vintage, cultural, heritage",
+            "medical": "medical, healthcare, clinical, scientific",
+            "finance": "financial, banking, investment, economic",
+            "marketing": "creative, colorful, engaging, promotional"
+        }
+        
+        # 内容类型到视觉元素的映射
+        self.content_type_visuals = {
+            "title": "title slide, presentation cover, professional header",
+            "overview": "overview diagram, roadmap, structure visualization",
+            "content": "content slide, information display, bullet points",
+            "conclusion": "conclusion slide, summary, final thoughts",
+            "data": "data visualization, charts, graphs, statistics",
+            "process": "process flow, workflow, step-by-step diagram",
+            "comparison": "comparison chart, versus layout, side-by-side",
+            "timeline": "timeline, chronological order, historical progression"
+        }
+        
+        # 语言特定的提示词模板
+        self.prompt_templates = {
+            "zh": {
+                "base": "专业的PPT幻灯片背景图片，{style_desc}，{content_desc}，高质量，商务风格",
+                "title": "PPT标题页背景，{topic}主题，{style_desc}，专业设计",
+                "content": "PPT内容页背景，支持{content_type}展示，{style_desc}，简洁明了",
+                "data": "数据可视化背景，适合图表展示，{style_desc}，专业分析风格",
+                "conclusion": "PPT结论页背景，总结性设计，{style_desc}，完整收尾"
+            },
+            "en": {
+                "base": "Professional PPT slide background, {style_desc}, {content_desc}, high quality, business style",
+                "title": "PPT title slide background, {topic} theme, {style_desc}, professional design",
+                "content": "PPT content slide background, supports {content_type} display, {style_desc}, clean and clear",
+                "data": "Data visualization background, suitable for charts, {style_desc}, professional analysis style",
+                "conclusion": "PPT conclusion slide background, summary design, {style_desc}, complete ending"
+            }
+        }
+    
+    async def generate_image_prompt(self, slide_context: PPTSlideContext) -> str:
+        """为PPT幻灯片生成图片提示词"""
+        try:
+            # 分析幻灯片内容
+            content_analysis = self._analyze_slide_content(slide_context)
+            
+            # 获取场景风格描述
+            style_desc = self._get_scenario_style(slide_context.scenario)
+            
+            # 获取内容类型描述
+            content_desc = self._get_content_description(slide_context, content_analysis)
+            
+            # 选择合适的提示词模板
+            template_key = self._select_template_key(slide_context, content_analysis)
+            template = self.prompt_templates[slide_context.language][template_key]
+            
+            # 构建提示词
+            prompt = template.format(
+                style_desc=style_desc,
+                content_desc=content_desc,
+                topic=slide_context.topic,
+                content_type=content_analysis["type"]
+            )
+            
+            # 添加质量和风格修饰符
+            prompt = self._enhance_prompt_quality(prompt, slide_context)
+            
+            # 添加负面提示词建议
+            negative_prompt = self._generate_negative_prompt(slide_context)
+            
+            logger.info(f"Generated image prompt for slide {slide_context.page_number}: {prompt[:100]}...")
+            
+            return prompt
+            
+        except Exception as e:
+            logger.error(f"Failed to generate image prompt: {e}")
+            # 返回基础提示词
+            return self._get_fallback_prompt(slide_context)
+    
+    async def create_generation_request(self, 
+                                      slide_context: PPTSlideContext,
+                                      provider: ImageProvider = ImageProvider.DALLE,
+                                      **kwargs) -> ImageGenerationRequest:
+        """创建图片生成请求"""
+        # 生成主提示词
+        prompt = await self.generate_image_prompt(slide_context)
+        
+        # 生成负面提示词
+        negative_prompt = self._generate_negative_prompt(slide_context)
+        
+        # 根据幻灯片类型调整参数
+        generation_params = self._get_generation_params(slide_context, provider)
+        generation_params.update(kwargs)
+        
+        return ImageGenerationRequest(
+            prompt=prompt,
+            negative_prompt=negative_prompt,
+            provider=provider,
+            **generation_params
+        )
+    
+    def _analyze_slide_content(self, slide_context: PPTSlideContext) -> Dict[str, Any]:
+        """分析幻灯片内容"""
+        content = slide_context.content.lower()
+        title = slide_context.title.lower()
+        
+        analysis = {
+            "type": "content",
+            "keywords": [],
+            "themes": [],
+            "visual_elements": [],
+            "complexity": "medium"
+        }
+        
+        # 识别内容类型
+        if slide_context.page_number == 1:
+            analysis["type"] = "title"
+        elif slide_context.page_number == slide_context.total_pages:
+            analysis["type"] = "conclusion"
+        elif any(word in content for word in ["数据", "统计", "图表", "分析", "data", "chart", "graph"]):
+            analysis["type"] = "data"
+        elif any(word in content for word in ["流程", "步骤", "过程", "process", "workflow", "step"]):
+            analysis["type"] = "process"
+        elif any(word in content for word in ["对比", "比较", "versus", "comparison", "vs"]):
+            analysis["type"] = "comparison"
+        elif any(word in content for word in ["时间", "历史", "发展", "timeline", "history", "evolution"]):
+            analysis["type"] = "timeline"
+        elif any(word in content for word in ["概述", "总览", "overview", "summary", "outline"]):
+            analysis["type"] = "overview"
+        
+        # 提取关键词
+        keywords = self._extract_content_keywords(content + " " + title)
+        analysis["keywords"] = keywords[:10]  # 最多10个关键词
+        
+        # 识别主题
+        themes = self._identify_content_themes(content + " " + title)
+        analysis["themes"] = themes
+        
+        # 评估复杂度
+        if len(content.split()) > 100:
+            analysis["complexity"] = "high"
+        elif len(content.split()) < 30:
+            analysis["complexity"] = "low"
+        
+        return analysis
+    
+    def _extract_content_keywords(self, text: str) -> List[str]:
+        """从内容中提取关键词"""
+        # 移除标点符号和特殊字符
+        text = re.sub(r'[^\w\s]', ' ', text)
+        words = text.split()
+        
+        # 停用词列表
+        stop_words = {
+            '的', '了', '在', '是', '我', '有', '和', '就', '不', '人', '都', '一', '上', '也', '很', '到',
+            'the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with', 'by'
+        }
+        
+        # 过滤停用词和短词
+        keywords = [word for word in words if word not in stop_words and len(word) > 2]
+        
+        # 统计词频并返回最常见的词
+        from collections import Counter
+        word_counts = Counter(keywords)
+        
+        return [word for word, count in word_counts.most_common(15)]
+    
+    def _identify_content_themes(self, text: str) -> List[str]:
+        """识别内容主题"""
+        themes = []
+        text_lower = text.lower()
+        
+        theme_keywords = {
+            "success": ["成功", "成就", "胜利", "优秀", "success", "achievement", "victory", "excellent"],
+            "innovation": ["创新", "创意", "新颖", "突破", "innovation", "creative", "novel", "breakthrough"],
+            "growth": ["增长", "发展", "提升", "进步", "growth", "development", "improvement", "progress"],
+            "teamwork": ["团队", "合作", "协作", "配合", "team", "cooperation", "collaboration", "partnership"],
+            "technology": ["技术", "科技", "数字", "智能", "technology", "digital", "smart", "tech"],
+            "business": ["商业", "业务", "市场", "经济", "business", "market", "economic", "commercial"],
+            "education": ["教育", "学习", "培训", "知识", "education", "learning", "training", "knowledge"],
+            "future": ["未来", "前景", "展望", "趋势", "future", "prospect", "outlook", "trend"]
+        }
+        
+        for theme, keywords in theme_keywords.items():
+            if any(keyword in text_lower for keyword in keywords):
+                themes.append(theme)
+        
+        return themes[:3]  # 最多3个主题
+    
+    def _get_scenario_style(self, scenario: str) -> str:
+        """获取场景对应的视觉风格"""
+        return self.scenario_styles.get(scenario, self.scenario_styles["general"])
+    
+    def _get_content_description(self, slide_context: PPTSlideContext, content_analysis: Dict[str, Any]) -> str:
+        """获取内容描述"""
+        content_type = content_analysis["type"]
+        base_desc = self.content_type_visuals.get(content_type, "professional presentation slide")
+        
+        # 添加关键词增强
+        if content_analysis["keywords"]:
+            keywords_str = ", ".join(content_analysis["keywords"][:3])
+            base_desc += f", related to {keywords_str}"
+        
+        # 添加主题增强
+        if content_analysis["themes"]:
+            themes_str = ", ".join(content_analysis["themes"])
+            base_desc += f", {themes_str} theme"
+        
+        return base_desc
+    
+    def _select_template_key(self, slide_context: PPTSlideContext, content_analysis: Dict[str, Any]) -> str:
+        """选择合适的提示词模板"""
+        content_type = content_analysis["type"]
+        
+        if content_type == "title":
+            return "title"
+        elif content_type == "data":
+            return "data"
+        elif content_type == "conclusion":
+            return "conclusion"
+        else:
+            return "content"
+    
+    def _enhance_prompt_quality(self, prompt: str, slide_context: PPTSlideContext) -> str:
+        """增强提示词质量"""
+        # 添加质量修饰符
+        quality_modifiers = [
+            "high resolution",
+            "professional quality",
+            "clean design",
+            "modern aesthetic",
+            "suitable for presentation"
+        ]
+        
+        # 根据场景添加特定修饰符
+        if slide_context.scenario == "business":
+            quality_modifiers.extend(["corporate style", "executive presentation"])
+        elif slide_context.scenario == "technology":
+            quality_modifiers.extend(["futuristic design", "digital aesthetic"])
+        elif slide_context.scenario == "education":
+            quality_modifiers.extend(["educational design", "academic style"])
+        
+        enhanced_prompt = prompt + ", " + ", ".join(quality_modifiers[:3])
+        
+        return enhanced_prompt
+    
+    def _generate_negative_prompt(self, slide_context: PPTSlideContext) -> str:
+        """生成负面提示词"""
+        negative_elements = [
+            "cluttered",
+            "messy",
+            "unprofessional",
+            "low quality",
+            "blurry",
+            "distorted",
+            "inappropriate",
+            "distracting elements",
+            "poor composition",
+            "amateur design"
+        ]
+        
+        # 根据场景添加特定的负面元素
+        if slide_context.scenario == "business":
+            negative_elements.extend(["casual", "informal", "playful"])
+        elif slide_context.scenario == "education":
+            negative_elements.extend(["complex", "overwhelming", "confusing"])
+        
+        return ", ".join(negative_elements[:8])
+    
+    def _get_generation_params(self, slide_context: PPTSlideContext, provider: ImageProvider) -> Dict[str, Any]:
+        """获取生成参数"""
+        params = {
+            "width": 1920,  # PPT标准宽度
+            "height": 1080,  # PPT标准高度
+            "quality": "standard"
+        }
+        
+        # 根据提供者调整参数
+        if provider == ImageProvider.DALLE:
+            params.update({
+                "size": "1792x1024",  # DALL-E 3支持的最接近16:9的尺寸
+                "quality": "standard",
+                "style": "natural"  # 更适合PPT背景
+            })
+        elif provider == ImageProvider.STABLE_DIFFUSION:
+            params.update({
+                "width": 1024,
+                "height": 576,  # 16:9比例
+                "steps": 30,
+                "cfg_scale": 7.0,
+                "sampler": "K_DPM_2_ANCESTRAL"
+            })
+        
+        # 根据幻灯片类型调整
+        if slide_context.page_number == 1:  # 标题页
+            params["quality"] = "hd" if provider == ImageProvider.DALLE else params.get("quality")
+        
+        return params
+    
+    def _get_fallback_prompt(self, slide_context: PPTSlideContext) -> str:
+        """获取回退提示词"""
+        scenario_style = self._get_scenario_style(slide_context.scenario)
+        
+        if slide_context.language == "zh":
+            return f"专业的PPT幻灯片背景，{scenario_style}风格，适合{slide_context.topic}主题，高质量，商务设计"
+        else:
+            return f"Professional PPT slide background, {scenario_style} style, suitable for {slide_context.topic} theme, high quality, business design"