ai应用落地尝试，软文生成功能

原创已于 2025-06-27 18:21:21 修改 · 294 阅读

1 ·

CC 4.0 BY-SA版权

文章标签：

#人工智能 #python

于 2025-06-27 16:24:06 首次发布

ai 专栏收录该内容

1 篇文章

订阅专栏

Python脚本驱动的大模型软文批量生成方案

以下是一个基于Python的高效软文批量生成解决方案，通过API集成主流大模型（如ChatGPT、Claude、文心一言等），实现全自动化的内容创作流程：

核心架构设计

graph LR
    A[关键词输入] --> B(提示词工程模块)
    C[行业模板库] --> B
    B --> D{大模型API}
    D --> E[内容生成]
    E --> F(后处理引擎)
    F --> G[敏感词过滤]
    F --> H[SEO优化]
    F --> I[改写降重]
    G & H & I --> J[批量输出]
    J --> K[格式转换]

完整实现方案

1. 依赖安装

pip install openai anthropic requests pandas numpy python-docx markdown

2. API集成模块（支持多模型热切换）

# api_integration.py
import openai
import anthropic
import requests

class LLM_Generator:
    def __init__(self, provider="openai", model="gpt-4-turbo"):
        self.provider = provider
        self.model = model
        self.api_keys = {
            "openai": "sk-your-openai-key",
            "anthropic": "your-claude-key",
            "baidu": "your-wenxin-key"
        }
        
    def generate(self, prompt, max_tokens=1200):
        if self.provider == "openai":
            return self._openai_generation(prompt, max_tokens)
        elif self.provider == "anthropic":
            return self._claude_generation(prompt, max_tokens)
        elif self.provider == "baidu":
            return self._wenxin_generation(prompt, max_tokens)
            
    def _openai_generation(self, prompt, max_tokens):
        client = openai.OpenAI(api_key=self.api_keys["openai"])
        response = client.chat.completions.create(
            model=self.model,
            messages=[{"role": "user", "content": prompt}],
            max_tokens=max_tokens,
            temperature=0.7
        )
        return response.choices[0].message.content
    
    def _claude_generation(self, prompt, max_tokens):
        client = anthropic.Anthropic(api_key=self.api_keys["anthropic"])
        response = client.messages.create(
            model="claude-3-opus-20240229",
            max_tokens=max_tokens,
            messages=[{"role": "user", "content": prompt}]
        )
        return response.content[0].text
    
    def _wenxin_generation(self, prompt, max_tokens):
        url = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions?access_token=" + self.api_keys["baidu"]
        payload = {
            "messages": [{"role":"user", "content": prompt}],
            "max_output_tokens": max_tokens
        }
        response = requests.post(url, json=payload)
        return response.json()["result"]

3. 智能提示词工程模块

# prompt_engineering.py
import pandas as pd

class PromptEngine:
    def __init__(self, template_file="templates.csv"):
        self.templates = pd.read_csv(template_file)
        
    def craft_prompt(self, keywords, template_type="marketing", word_count=800):
        # 智能选择模板
        template = self.templates[
            (self.templates["type"] == template_type) &
            (self.templates["word_count"] == word_count)
        ].sample().iloc[0]["prompt"]
        
        # 动态注入关键词
        prompt = template.replace("{keywords}", ", ".join(keywords))
        
        # 添加格式要求
        prompt += f"""
        请严格遵循以下要求生成内容：
        - 内容结构：标题+5个段落+行动号召
        - 插入3个相关关键词：{','.join(keywords[-3:])}
        - 自然过渡段落
        - 包含行业术语但保持通俗易懂
        - 最后添加3个常见用户问答
        """
        return prompt

4. 批量化生成引擎

# batch_generator.py
import time
import pandas as pd
from concurrent.futures import ThreadPoolExecutor
from api_integration import LLM_Generator
from prompt_engineering import PromptEngine

class ContentFactory:
    def __init__(self, keywords_list, output_format="markdown"):
        self.keywords_list = keywords_list
        self.output_format = output_format
        self.generator = LLM_Generator(provider="anthropic")  # 默认使用Claude
        self.prompt_engine = PromptEngine()
        
    def process_keywords(self, keywords):
        try:
            prompt = self.prompt_engine.craft_prompt(keywords)
            content = self.generator.generate(prompt)
            return self.post_process(content, keywords)
        except Exception as e:
            print(f"生成失败: {e}")
            return None
            
    def post_process(self, content, keywords):
        # 内容优化处理（简化版）
        content = "\n".join([line for line in content.split("\n") if line.strip()])
        content = content.replace("人工智能", "AI")  # 术语统一
        content += f"\n\n相关关键词: {', '.join(keywords)}"
        return content
    
    def run_batch(self, max_workers=3):
        results = []
        with ThreadPoolExecutor(max_workers=max_workers) as executor:
            futures = [executor.submit(self.process_keywords, kw) for kw in self.keywords_list]
            for future in futures:
                result = future.result()
                if result:
                    results.append(result)
                    # 遵守API限速
                    time.sleep(3)  
        return self.export(results)
    
    def export(self, contents):
        timestamp = time.strftime("%Y%m%d_%H%M")
        if self.output_format == "markdown":
            file_name = f"soft_articles_{timestamp}.md"
            with open(file_name, "w", encoding="utf-8") as f:
                for i, content in enumerate(contents):
                    f.write(f"## 文章 {i+1}\n\n")
                    f.write(content)
                    f.write("\n\n---\n\n")
            return file_name
        # 支持Word、Excel等格式导出

5. 行业模板数据库 (templates.csv)

type	word_count	prompt
marketing	800	作为资深数字营销专家，请创作一篇关于{keywords}的推广软文。重点突出产品的创新性，解决目标用户的痛点，并包含3个真实应用场景...
tech	1200	作为科技专栏作者，撰写一篇{keywords}的深度解析文章。从技术原理出发，分析行业现状，预测未来发展趋势，适合专业读者...
finance	1000	撰写面向普通投资者的理财软文，主题围绕{keywords}。用通俗语言解释复杂概念，提供实用建议，避免夸大收益...

6. 主程序入口

# main.py
from batch_generator import ContentFactory

if __name__ == "__main__":
    # 批量关键词输入 (实际应用中可从Excel或数据库读取)
    topics = [
        ["云服务器", "企业数字化转型", "成本优化"],
        ["AI客服", "客户体验", "智能化升级"],
        ["区块链", "供应链金融", "去中心化"]
    ]
    
    # 启动生成工厂
    factory = ContentFactory(
        keywords_list=topics,
        output_format="markdown"
    )
    
    # 批量生成50篇软文
    output_file = factory.run_batch(max_workers=5)
    print(f"生成完成! 文件已保存至: {output_file}")

优化进阶功能

SEO增强模块

def seo_optimize(content, keywords):
    """自动SEO优化处理"""
    from bs4 import BeautifulSoup  # 针对HTML格式
    
    # 关键词密度分析
    total_words = len(content.split())
    for kw in keywords:
        kw_density = content.lower().count(kw.lower()) / total_words * 100
        if kw_density < 1.5:  # 增加关键词密度
            paragraphs = content.split("\n\n")
            for i, para in enumerate(paragraphs):
                if i % 2 == 0 and kw_density < 1.5:
                    paragraphs[i] = f"{kw.capitalize()}应用领域广泛的背景下，{para}"
                    kw_density = paragraphs[i].count(kw) / len(paragraphs[i].split()) * 100
            content = "\n\n".join(paragraphs)
    
    # 自动生成Meta描述
    meta_desc = f"本文深入探讨{','.join(keywords)}的技术创新与应用价值。"
    
    return f"<meta name='description' content='{meta_desc}'>\n" + content

敏感词过滤系统

# 加载敏感词库 (约5万词)
with open("sensitive_words.txt", encoding="utf-8") as f:
    sensitive_words = set(line.strip() for line in f)

def filter_content(content):
    """过滤违规内容并替换"""
    import re
    for word in sensitive_words:
        if word in content:
            # 保留首尾字符的模糊处理
            content = re.sub(rf"\b{word}\b", word[0] + "*"*(len(word)-2) + word[-1], content)
    return content

智能改写降重

def rewrite_content(content, similarity=0.3):
    """AI辅助降重改写，保留原意改变表述"""
    from difflib import SequenceMatcher
    from api_integration import LLM_Generator
    
    prompt = f"""请以专业编辑身份改写以下内容，要求：
    - 保持核心信息不变
    - 原文相似度低于{int(similarity*100)}%
    - 采用新的表达结构和词汇
    - 增强可读性和专业性
    
    原文：
    {content[:1500]}..."""
    
    llm = LLM_Generator(provider="openai")
    rewritten = llm.generate(prompt)
    
    # 相似度验证
    ratio = SequenceMatcher(None, content, rewritten).ratio()
    if ratio < similarity:
        return rewritten
    else:
        return rewrite_content(content, similarity-0.05)  # 递归降重