lt
325 lines
13 KiB
Python
325 lines
13 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
撰写阶段：基于大纲和选题生成完整文章（三平台版本）
|
||
"""
|
||
|
||
import json, datetime, logging, sys, re, subprocess
|
||
from pathlib import Path
|
||
from typing import Dict, List
|
||
import base64
|
||
from io import BytesIO
|
||
|
||
PROJECT_ROOT = Path(__file__).parent.parent
|
||
# 添加项目根和 backend 路径，以导入 app.core.llm_client
|
||
sys.path.insert(0, str(PROJECT_ROOT))
|
||
sys.path.insert(0, str(PROJECT_ROOT / 'platform' / 'backend'))
|
||
|
||
# 导入 LLM 客户端（NVIDIA）
|
||
try:
|
||
from app.core.qnaigc_client import expand_content_with_llm # type: ignore
|
||
HAVE_LLM = True
|
||
except ImportError as e:
|
||
logging.warning(f"LLM client unavailable: {e}")
|
||
HAVE_LLM = False
|
||
|
||
PROJECT_ROOT = Path(__file__).parent.parent
|
||
sys.path.insert(0, str(PROJECT_ROOT))
|
||
# 导入数据库模型
|
||
from app.database import SessionLocal
|
||
from app.models import Topic
|
||
|
||
DATA_DIR = PROJECT_ROOT / "automation" / "data"
|
||
TOPICS_FILE = DATA_DIR / "sustainability_topics.json"
|
||
OUTLINE_DIR = DATA_DIR / "outlines"
|
||
RELEASE_DIR = DATA_DIR / "releases"
|
||
TEMPLATES_DIR = PROJECT_ROOT / "automation" / "templates"
|
||
LOGS_DIR = PROJECT_ROOT / "automation" / "logs"
|
||
TODAY = datetime.datetime.now().strftime("%Y-%m-%d")
|
||
|
||
logging.basicConfig(
|
||
level=logging.INFO,
|
||
format='%(asctime)s - %(levelname)s - %(message)s',
|
||
handlers=[
|
||
logging.FileHandler(LOGS_DIR / f"writer_{TODAY}.log"),
|
||
logging.StreamHandler()
|
||
]
|
||
)
|
||
logger = logging.getLogger(__name__)
|
||
|
||
class Writer:
|
||
def __init__(self, topic_id: str):
|
||
self.topic_id = topic_id
|
||
self.topic = self._load_topic()
|
||
outline_file = OUTLINE_DIR / TODAY / f"{topic_id}_outline.md"
|
||
if not outline_file.exists():
|
||
raise FileNotFoundError(f"Outline not found: {outline_file}")
|
||
self.outline_content = outline_file.read_text(encoding='utf-8')
|
||
self.release_dir = RELEASE_DIR / TODAY
|
||
self.release_dir.mkdir(parents=True, exist_ok=True)
|
||
# 加载研究笔记（作为 LLM 上下文）
|
||
research_file = DATA_DIR / "research" / TODAY / f"{topic_id}_research.md"
|
||
self.research_notes = research_file.read_text(encoding='utf-8') if research_file.exists() else ""
|
||
|
||
def _load_topic(self) -> Dict:
|
||
topics = json.loads(TOPICS_FILE.read_text(encoding='utf-8'))
|
||
for t in topics:
|
||
if t['id'] == self.topic_id:
|
||
return t
|
||
raise ValueError(f"Topic {self.topic_id} not found")
|
||
|
||
def _clean_title(self, title: str) -> str:
|
||
"""去除标题中的指导性文字（如字数说明、MVP标记等）"""
|
||
import re
|
||
# 去掉括号中的说明：约200字、约300字、MVP、试行等
|
||
title = re.sub(r'[（(]约\s*\d+字[）)]', '', title)
|
||
title = re.sub(r'[（(]MVP[）)]', '', title)
|
||
title = re.sub(r'[（(][^）)]*?[）)]', '', title) # 保守移除任意括号内容（可能误伤，但大纲通常不包含重要括号信息）
|
||
return title.strip()
|
||
|
||
def _parse_outline_sections(self) -> List[Dict]:
|
||
"""将大纲 Markdown 解析为结构化列表，保留层级和内容"""
|
||
sections = []
|
||
current = None
|
||
for line in self.outline_content.splitlines():
|
||
if line.startswith("# "):
|
||
if current:
|
||
sections.append(current)
|
||
current = {"level": 1, "title": line[2:].strip(), "content": ""}
|
||
elif line.startswith("## "):
|
||
if current:
|
||
sections.append(current)
|
||
current = {"level": 2, "title": line[3:].strip(), "content": ""}
|
||
elif line.startswith("### "):
|
||
if current:
|
||
sections.append(current)
|
||
current = {"level": 3, "title": line[4:].strip(), "content": ""}
|
||
else:
|
||
if current and line.strip():
|
||
current['content'] = current.get('content', '') + line + "\n"
|
||
if current:
|
||
sections.append(current)
|
||
return sections
|
||
|
||
def _expand_section(self, section: Dict) -> str:
|
||
"""将大纲中的简短描述扩展为完整段落"""
|
||
content = section.get('content', '').strip()
|
||
# 如果有足够内容（>200字），直接返回
|
||
if len(content) > 200:
|
||
return content
|
||
# 如果内容极少，需要 LLM 扩写
|
||
if HAVE_LLM and len(content) < 150:
|
||
logger.info(f"使用 LLM 扩写章节: {section['title']}")
|
||
try:
|
||
expanded = expand_content_with_llm(
|
||
topic=self.topic,
|
||
section_title=section['title'],
|
||
section_content=content,
|
||
context=self.research_notes
|
||
)
|
||
if expanded and len(expanded.strip()) > len(content):
|
||
return expanded.strip()
|
||
else:
|
||
logger.warning("LLM 扩写结果为空或过短，使用占位")
|
||
raise ValueError("Empty expansion")
|
||
except Exception as e:
|
||
logger.warning(f"LLM 扩写失败: {e}，使用占位内容")
|
||
# 返回占位内容，保持流程继续
|
||
return f"{content}\n\n（本段内容需要人工补充：当前模型调用失败或未配置）"
|
||
# 否则返回原内容
|
||
return content
|
||
|
||
def generate_full_markdown(self) -> str:
|
||
"""根据大纲生成完整 Markdown 正文（不用原标题，全部由 LLM 扩写生成）"""
|
||
sections = self._parse_outline_sections()
|
||
parts = []
|
||
|
||
# 只保留 LLM 扩写的内容，不添加任何原始标题标记
|
||
for sec in sections:
|
||
# 如果内容极短，LLM 扩写后返回的完整段落中可能包含标题，我们不过滤
|
||
if sec.get('content'):
|
||
expanded = self._expand_section(sec)
|
||
parts.append(expanded + "\n\n")
|
||
|
||
full_md = "\n".join(parts).strip()
|
||
|
||
# 添加文末声明
|
||
full_md += f"\n<p>（本文由宇之然AI助手生成，数据来源可靠，内容经合规审查）</p>\n"
|
||
full_md += f"<p><em>生成时间：{TODAY}</em></p>\n"
|
||
return full_md
|
||
|
||
def generate_platform_html(self, markdown: str, platform: str) -> str:
|
||
"""将 Markdown 转换为平台 HTML（基于模板）"""
|
||
title = self.topic['title']
|
||
|
||
# 加载模板
|
||
tpl_path = TEMPLATES_DIR / f"{platform}.html"
|
||
if tpl_path.exists():
|
||
template = tpl_path.read_text(encoding='utf-8')
|
||
else:
|
||
template = "<!DOCTYPE html><html><head><meta charset='UTF-8'><title>{{TITLE}}</title></head><body><h1>{{TITLE}}</h1><!-- CONTENT --></body></html>"
|
||
|
||
# 替换变量
|
||
html = template.replace("{{TITLE}}", title).replace("{{DATE}}", TODAY)
|
||
|
||
# 注入内容 (简单处理：markdown 转 HTML 可以用 marked.js 或 simple转换，这里暂时用 <pre> 包裹或简单段落化)
|
||
# 为了快速展示，我们将 markdown 的段落转换为 <p> 标签
|
||
# 实际中建议使用 markdown 库（如 python-markdown）转换
|
||
html_content = self._markdown_to_html(markdown)
|
||
html = html.replace("<!-- CONTENT -->", html_content)
|
||
|
||
# 平台特定标签补充
|
||
if platform == "zhihu":
|
||
tags = '<div class="tags">#科技 #职场</div>'
|
||
html = html.replace("<!-- TAGS -->", tags)
|
||
elif platform == "xiaohongshu":
|
||
hashtags = '<div class="hashtags">#AI #可持续 #生活方式</div>'
|
||
html = html.replace("<!-- HASHTAGS -->", hashtags)
|
||
elif platform == "wechat":
|
||
# 微信公众号可能还需要摘要等，模板已处理
|
||
pass
|
||
|
||
if platform == "xiaohongshu":
|
||
html = self._fill_image_placeholders(html, platform, title)
|
||
return html
|
||
|
||
def _markdown_to_html(self, md: str) -> str:
|
||
"""极简 markdown 转换（仅本场景使用）"""
|
||
lines = md.split('\n')
|
||
html_parts = []
|
||
for line in lines:
|
||
if line.startswith('# '):
|
||
html_parts.append(f"<h1>{line[2:]}</h1>")
|
||
elif line.startswith('## '):
|
||
html_parts.append(f"<h2>{line[3:]}</h2>")
|
||
elif line.startswith('### '):
|
||
html_parts.append(f"<h3>{line[4:]}</h3>")
|
||
elif line.strip().startswith('- '):
|
||
html_parts.append(f"<li>{line[2:]}</li>")
|
||
elif re.match(r'^\d+\. ', line):
|
||
content = re.sub(r'^\d+\. ', '', line)
|
||
html_parts.append(f"<li>{content}</li>")
|
||
elif line.strip():
|
||
html_parts.append(f"<p>{line}</p>")
|
||
else:
|
||
html_parts.append("") # 空行
|
||
return "\n".join(html_parts)
|
||
|
||
def save_html(self, html: str, platform: str) -> Path:
|
||
out_dir = self.release_dir / platform
|
||
out_dir.mkdir(parents=True, exist_ok=True)
|
||
filename = f"{platform}_{self.topic_id}_{platform}.html"
|
||
out_path = out_dir / filename
|
||
out_path.write_text(html, encoding='utf-8')
|
||
logger.info(f"HTML 生成: {out_path}")
|
||
return out_path
|
||
|
||
def mark_draft(self):
|
||
"""标记选题为「待审查」，同时更新数据库"""
|
||
"""标记选题为「待发布」，同时更新数据库"""
|
||
# 更新 JSON 文件
|
||
with open(TOPICS_FILE, 'r', encoding='utf-8') as f:
|
||
topics = json.load(f)
|
||
updated = False
|
||
for t in topics:
|
||
if t.get('id') == self.topic_id:
|
||
t['status'] = '待审查'
|
||
updated = True
|
||
break
|
||
if updated:
|
||
with open(TOPICS_FILE, 'w', encoding='utf-8') as f:
|
||
json.dump(topics, f, ensure_ascii=False, indent=2)
|
||
|
||
# 更新数据库
|
||
db = SessionLocal()
|
||
try:
|
||
topic_db = db.query(Topic).filter(Topic.id == self.topic_id).first()
|
||
if topic_db:
|
||
topic_db.status = '待审查'
|
||
db.commit()
|
||
logger.info(f"选题 {self.topic_id} 状态已更新为 draft（数据库）")
|
||
else:
|
||
logger.warning(f"数据库中未找到选题 {self.topic_id}")
|
||
except Exception as e:
|
||
logger.error(f"更新数据库失败: {e}")
|
||
db.rollback()
|
||
finally:
|
||
db.close()
|
||
|
||
logger.info(f"选题 {self.topic_id} 状态更新为「待审查」（JSON）")
|
||
"""标记选题为「待发布」"""
|
||
with open(TOPICS_FILE, 'r', encoding='utf-8') as f:
|
||
topics = json.load(f)
|
||
for t in topics:
|
||
if t.get('id') == self.topic_id:
|
||
t['status'] = '待审查'
|
||
# ready_at 留空，待合规审核通过后设置
|
||
break
|
||
with open(TOPICS_FILE, 'w', encoding='utf-8') as f:
|
||
json.dump(topics, f, ensure_ascii=False, indent=2)
|
||
logger.info(f"选题 {self.topic_id} 状态更新为「待审查」")
|
||
|
||
def run(self):
|
||
logger.info("开始撰写阶段")
|
||
markdown = self.generate_full_markdown()
|
||
results = {}
|
||
for platform in ["zhihu", "wechat", "xiaohongshu"]:
|
||
html = self.generate_platform_html(markdown, platform)
|
||
results[platform] = str(self.save_html(html, platform))
|
||
self.mark_draft()
|
||
logger.info(f"撰写完成，状态改为 draft，待合规审核")
|
||
return {"ok": True, "files": results}
|
||
|
||
|
||
def _image_to_data_url(self, img_path: Path, fmt: str = None) -> str:
|
||
data = img_path.read_bytes()
|
||
if fmt is None:
|
||
fmt = img_path.suffix.lstrip('.').lower()
|
||
b64 = base64.b64encode(data).decode('ascii')
|
||
return f"data:image/{fmt};base64,{b64}"
|
||
|
||
def _generate_and_inline_images(self, platform: str, title: str) -> dict:
|
||
from scripts.image_generator import ImageGenerator
|
||
gen = ImageGenerator()
|
||
files = gen.generate_all_placeholders(title, platform)
|
||
mapping = {}
|
||
cover = files.get('cover')
|
||
if cover and cover.exists():
|
||
mapping['main-image-src'] = self._image_to_data_url(cover)
|
||
thumbs = []
|
||
for k, p in files.items():
|
||
if k != 'cover' and p.exists():
|
||
thumbs.append(self._image_to_data_url(p))
|
||
mapping['thumbnail-srcs'] = thumbs
|
||
return mapping
|
||
|
||
def _fill_image_placeholders(self, html: str, platform: str, title: str) -> str:
|
||
if platform != 'xiaohongshu':
|
||
return html
|
||
mapping = self._generate_and_inline_images(platform, title)
|
||
# Replace main image placeholder
|
||
main_ph = '<img src="" alt="封面图" class="main-image">'
|
||
if 'main-image-src' in mapping:
|
||
new_main = f'<img src="{mapping["main-image-src"]}" alt="封面图" class="main-image">'
|
||
html = html.replace(main_ph, new_main)
|
||
# Replace thumbnail placeholders (6)
|
||
thumbs = mapping.get('thumbnail-srcs', [])
|
||
for idx, src in enumerate(thumbs[:6], start=1):
|
||
ph = f'<img src="" alt="图{idx}" class="thumbnail">'
|
||
new_thumb = f'<img src="{src}" alt="图{idx}" class="thumbnail">'
|
||
html = html.replace(ph, new_thumb)
|
||
return html
|
||
|
||
def main():
|
||
import argparse
|
||
parser = argparse.ArgumentParser()
|
||
parser.add_argument('--topic-id', required=True, help='选题ID')
|
||
args = parser.parse_args()
|
||
|
||
w = Writer(args.topic_id)
|
||
result = w.run()
|
||
print(json.dumps(result, ensure_ascii=False))
|
||
sys.exit(0 if result['ok'] else 1)
|
||
|
||
if __name__ == "__main__":
|
||
main()
|