9b44a07a44
주요 추가 기능: - bots/shorts/ 서브모듈 7개: tts_engine, script_extractor, asset_resolver, stock_fetcher, caption_renderer, video_assembler, youtube_uploader - bots/shorts_bot.py: 6단계 Shorts 파이프라인 오케스트레이터 (auto/semi_auto 두 가지 생산 모드, CLI 지원) - bots/writer_bot.py: 독립 실행형 AI 글쓰기 봇 (대시보드 연동) - bots/assist_bot.py: URL 기반 수동 어시스트 파이프라인 - config/shorts_config.json: Shorts 전체 설정 - templates/shorts/extract_prompt.txt: LLM 스크립트 추출 프롬프트 - scheduler.py에 shorts 잡(10:35/16:00) + /shorts Telegram 명령 추가 보안 개선: - .env 파일 외부 경로 참조로 변경 (load_dotenv dotenv_path, 24개 파일) - .gitignore에 민감 파일/내부 문서/런타임 데이터 항목 추가 문서: - README.md 전면 재작성 (상세 한글 설명, 설치/설정/사용법 포함) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
534 lines
20 KiB
Python
534 lines
20 KiB
Python
"""
|
|
수집봇 (collector_bot.py)
|
|
역할: 트렌드/도구/사례 수집 + 품질 점수 계산 + 폐기 규칙 적용
|
|
실행: 매일 07:00 (스케줄러 호출)
|
|
"""
|
|
import json
|
|
import logging
|
|
import os
|
|
import re
|
|
import hashlib
|
|
from datetime import datetime, timedelta, timezone
|
|
from difflib import SequenceMatcher
|
|
from pathlib import Path
|
|
|
|
import feedparser
|
|
import requests
|
|
from bs4 import BeautifulSoup
|
|
from dotenv import load_dotenv
|
|
|
|
load_dotenv(dotenv_path='D:/key/blog-writer.env.env')
|
|
|
|
BASE_DIR = Path(__file__).parent.parent
|
|
CONFIG_DIR = BASE_DIR / 'config'
|
|
DATA_DIR = BASE_DIR / 'data'
|
|
LOG_DIR = BASE_DIR / 'logs'
|
|
LOG_DIR.mkdir(exist_ok=True)
|
|
|
|
logging.basicConfig(
|
|
level=logging.INFO,
|
|
format='%(asctime)s [%(levelname)s] %(message)s',
|
|
handlers=[
|
|
logging.FileHandler(LOG_DIR / 'collector.log', encoding='utf-8'),
|
|
logging.StreamHandler(),
|
|
]
|
|
)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
# 코너별 타입
|
|
CORNER_TYPES = {
|
|
'easy_guide': '쉬운세상',
|
|
'hidden_gems': '숨은보물',
|
|
'vibe_report': '바이브리포트',
|
|
'fact_check': '팩트체크',
|
|
'one_cut': '한컷',
|
|
}
|
|
|
|
# 글감 타입 비율: 에버그린 50%, 트렌드 30%, 개성 20%
|
|
TOPIC_RATIO = {'evergreen': 0.5, 'trending': 0.3, 'personality': 0.2}
|
|
|
|
|
|
def load_config(filename: str) -> dict:
|
|
with open(CONFIG_DIR / filename, 'r', encoding='utf-8') as f:
|
|
return json.load(f)
|
|
|
|
|
|
def load_published_titles() -> list[str]:
|
|
"""발행 이력에서 제목 목록을 불러옴 (유사도 비교용)"""
|
|
titles = []
|
|
published_dir = DATA_DIR / 'published'
|
|
for f in published_dir.glob('*.json'):
|
|
try:
|
|
data = json.loads(f.read_text(encoding='utf-8'))
|
|
if 'title' in data:
|
|
titles.append(data['title'])
|
|
except Exception:
|
|
pass
|
|
return titles
|
|
|
|
|
|
def title_similarity(a: str, b: str) -> float:
|
|
return SequenceMatcher(None, a, b).ratio()
|
|
|
|
|
|
def is_duplicate(title: str, published_titles: list[str], threshold: float = 0.8) -> bool:
|
|
for pub_title in published_titles:
|
|
if title_similarity(title, pub_title) >= threshold:
|
|
return True
|
|
return False
|
|
|
|
|
|
def calc_freshness_score(published_at: datetime | None, max_score: int = 20) -> int:
|
|
"""발행 시간 기준 신선도 점수 (24h 이내 만점, 7일 초과 0점)"""
|
|
if published_at is None:
|
|
return max_score // 2
|
|
now = datetime.now(timezone.utc)
|
|
if published_at.tzinfo is None:
|
|
published_at = published_at.replace(tzinfo=timezone.utc)
|
|
age_hours = (now - published_at).total_seconds() / 3600
|
|
if age_hours <= 24:
|
|
return max_score
|
|
elif age_hours >= 168:
|
|
return 0
|
|
else:
|
|
ratio = 1 - (age_hours - 24) / (168 - 24)
|
|
return int(max_score * ratio)
|
|
|
|
|
|
def calc_korean_relevance(text: str, rules: dict) -> int:
|
|
"""한국 독자 관련성 점수"""
|
|
max_score = rules['scoring']['korean_relevance']['max']
|
|
keywords = rules['scoring']['korean_relevance']['keywords']
|
|
|
|
# 한국어 문자(가-힣) 비율 체크 — 한국어 콘텐츠 자체에 기본점수 부여
|
|
korean_chars = sum(1 for c in text if '\uac00' <= c <= '\ud7a3')
|
|
korean_ratio = korean_chars / max(len(text), 1)
|
|
if korean_ratio >= 0.15:
|
|
base = 15 # 한국어 텍스트면 기본 15점
|
|
elif korean_ratio >= 0.05:
|
|
base = 8
|
|
else:
|
|
base = 0
|
|
|
|
# 브랜드/지역 키워드 보너스
|
|
matched = sum(1 for kw in keywords if kw in text)
|
|
bonus = min(matched * 5, max_score - base)
|
|
|
|
return min(base + bonus, max_score)
|
|
|
|
|
|
def calc_source_trust(source_url: str, rules: dict) -> tuple[int, str]:
|
|
"""출처 신뢰도 점수 + 레벨"""
|
|
trust_cfg = rules['scoring']['source_trust']
|
|
high_src = trust_cfg.get('high_sources', [])
|
|
low_src = trust_cfg.get('low_sources', [])
|
|
url_lower = source_url.lower()
|
|
for s in low_src:
|
|
if s in url_lower:
|
|
return trust_cfg['levels']['low'], 'low'
|
|
for s in high_src:
|
|
if s in url_lower:
|
|
return trust_cfg['levels']['high'], 'high'
|
|
return trust_cfg['levels']['medium'], 'medium'
|
|
|
|
|
|
def calc_monetization(text: str, rules: dict) -> int:
|
|
"""수익 연결 가능성 점수"""
|
|
keywords = rules['scoring']['monetization']['keywords']
|
|
matched = sum(1 for kw in keywords if kw in text)
|
|
return min(matched * 5, rules['scoring']['monetization']['max'])
|
|
|
|
|
|
def is_evergreen(title: str, rules: dict) -> bool:
|
|
evergreen_kws = rules.get('evergreen_keywords', [])
|
|
return any(kw in title for kw in evergreen_kws)
|
|
|
|
|
|
def apply_discard_rules(item: dict, rules: dict, published_titles: list[str]) -> str | None:
|
|
"""
|
|
폐기 규칙 적용. 폐기 사유 반환(None이면 통과).
|
|
"""
|
|
title = item.get('topic', '')
|
|
text = title + ' ' + item.get('description', '')
|
|
discard_rules = rules.get('discard_rules', [])
|
|
|
|
for rule in discard_rules:
|
|
rule_id = rule['id']
|
|
|
|
if rule_id == 'no_korean_relevance':
|
|
if item.get('korean_relevance_score', 0) == 0:
|
|
return '한국 독자 관련성 없음'
|
|
|
|
elif rule_id == 'unverified_source':
|
|
if item.get('source_trust_level') == 'unknown':
|
|
return '출처 불명'
|
|
|
|
elif rule_id == 'duplicate_topic':
|
|
threshold = rule.get('similarity_threshold', 0.8)
|
|
if is_duplicate(title, published_titles, threshold):
|
|
return f'기발행 주제와 유사도 {threshold*100:.0f}% 이상'
|
|
|
|
elif rule_id == 'stale_trend':
|
|
if not item.get('is_evergreen', False):
|
|
max_days = rule.get('max_age_days', 7)
|
|
pub_at = item.get('published_at')
|
|
if pub_at:
|
|
if isinstance(pub_at, str):
|
|
try:
|
|
pub_at = datetime.fromisoformat(pub_at)
|
|
except Exception:
|
|
pub_at = None
|
|
if pub_at:
|
|
if pub_at.tzinfo is None:
|
|
pub_at = pub_at.replace(tzinfo=timezone.utc)
|
|
age_days = (datetime.now(timezone.utc) - pub_at).days
|
|
if age_days > max_days:
|
|
return f'{age_days}일 지난 트렌드'
|
|
|
|
elif rule_id == 'promotional':
|
|
kws = rule.get('keywords', [])
|
|
if any(kw in text for kw in kws):
|
|
return '광고성/홍보성 콘텐츠'
|
|
|
|
elif rule_id == 'clickbait':
|
|
patterns = rule.get('patterns', [])
|
|
if any(p in text for p in patterns):
|
|
return '클릭베이트성 주제'
|
|
|
|
return None
|
|
|
|
|
|
def assign_corner(item: dict, topic_type: str) -> str:
|
|
"""글감에 코너 배정"""
|
|
title = item.get('topic', '').lower()
|
|
source = item.get('source', 'rss').lower()
|
|
|
|
if topic_type == 'evergreen':
|
|
if any(kw in title for kw in ['가이드', '방법', '사용법', '입문', '튜토리얼', '기초']):
|
|
return '쉬운세상'
|
|
return '숨은보물'
|
|
elif topic_type == 'trending':
|
|
if source in ['github', 'product_hunt']:
|
|
return '숨은보물'
|
|
return '쉬운세상'
|
|
else: # personality
|
|
return '바이브리포트'
|
|
|
|
|
|
def calculate_quality_score(item: dict, rules: dict) -> int:
|
|
"""0-100점 품질 점수 계산"""
|
|
text = item.get('topic', '') + ' ' + item.get('description', '')
|
|
source_url = item.get('source_url', '')
|
|
pub_at_str = item.get('published_at')
|
|
pub_at = None
|
|
if pub_at_str:
|
|
try:
|
|
pub_at = datetime.fromisoformat(pub_at_str)
|
|
except Exception:
|
|
pass
|
|
|
|
kr_score = calc_korean_relevance(text, rules)
|
|
fresh_score = calc_freshness_score(pub_at)
|
|
# search_demand: pytrends 연동 후 실제값 사용 (RSS 기본값 12)
|
|
search_score = item.get('search_demand_score', 12)
|
|
# 신뢰도: _trust_override 이미 설정된 경우 우선 사용
|
|
if '_trust_score' in item:
|
|
trust_score = item['_trust_score']
|
|
trust_level = item.get('source_trust_level', 'medium')
|
|
else:
|
|
trust_score, trust_level = calc_source_trust(source_url, rules)
|
|
mono_score = calc_monetization(text, rules)
|
|
|
|
item['korean_relevance_score'] = kr_score
|
|
item['source_trust_level'] = trust_level
|
|
item['is_evergreen'] = is_evergreen(item.get('topic', ''), rules)
|
|
|
|
total = kr_score + fresh_score + search_score + trust_score + mono_score
|
|
return min(total, 100)
|
|
|
|
|
|
# ─── 수집 소스별 함수 ─────────────────────────────────
|
|
|
|
def collect_google_trends() -> list[dict]:
|
|
"""Google Trends (pytrends) — 한국 일간 트렌딩"""
|
|
items = []
|
|
try:
|
|
from pytrends.request import TrendReq
|
|
pytrends = TrendReq(hl='ko', tz=540, timeout=(10, 30))
|
|
trending_df = pytrends.trending_searches(pn='south_korea')
|
|
for keyword in trending_df[0].tolist()[:20]:
|
|
items.append({
|
|
'topic': keyword,
|
|
'description': f'Google Trends 한국 트렌딩 키워드: {keyword}',
|
|
'source': 'google_trends',
|
|
'source_url': f'https://trends.google.co.kr/trends/explore?q={keyword}&geo=KR',
|
|
'published_at': datetime.now(timezone.utc).isoformat(),
|
|
'search_demand_score': 15,
|
|
'topic_type': 'trending',
|
|
})
|
|
except Exception as e:
|
|
logger.warning(f"Google Trends 수집 실패: {e}")
|
|
return items
|
|
|
|
|
|
def collect_github_trending(sources_cfg: dict) -> list[dict]:
|
|
"""GitHub Trending 크롤링"""
|
|
items = []
|
|
cfg = sources_cfg.get('github_trending', {})
|
|
languages = cfg.get('languages', [''])
|
|
since = cfg.get('since', 'daily')
|
|
|
|
for lang in languages:
|
|
url = f"https://github.com/trending/{lang}?since={since}"
|
|
try:
|
|
resp = requests.get(url, timeout=15, headers={'User-Agent': 'Mozilla/5.0'})
|
|
soup = BeautifulSoup(resp.text, 'lxml')
|
|
repos = soup.select('article.Box-row')
|
|
for repo in repos[:10]:
|
|
name_el = repo.select_one('h2 a')
|
|
desc_el = repo.select_one('p')
|
|
stars_el = repo.select_one('a[href*="stargazers"]')
|
|
if not name_el:
|
|
continue
|
|
repo_path = name_el.get('href', '').strip('/')
|
|
topic = repo_path.replace('/', ' / ')
|
|
desc = desc_el.get_text(strip=True) if desc_el else ''
|
|
stars = stars_el.get_text(strip=True) if stars_el else '0'
|
|
items.append({
|
|
'topic': topic,
|
|
'description': desc,
|
|
'source': 'github',
|
|
'source_url': f'https://github.com/{repo_path}',
|
|
'published_at': datetime.now(timezone.utc).isoformat(),
|
|
'search_demand_score': 12,
|
|
'topic_type': 'trending',
|
|
'extra': {'stars': stars},
|
|
})
|
|
except Exception as e:
|
|
logger.warning(f"GitHub Trending 수집 실패 ({lang}): {e}")
|
|
return items
|
|
|
|
|
|
def collect_hacker_news(sources_cfg: dict) -> list[dict]:
|
|
"""Hacker News API 상위 스토리"""
|
|
items = []
|
|
cfg = sources_cfg.get('hacker_news', {})
|
|
api_url = cfg.get('url', 'https://hacker-news.firebaseio.com/v0/topstories.json')
|
|
top_n = cfg.get('top_n', 30)
|
|
try:
|
|
resp = requests.get(api_url, timeout=10)
|
|
story_ids = resp.json()[:top_n]
|
|
for sid in story_ids:
|
|
story_resp = requests.get(
|
|
f'https://hacker-news.firebaseio.com/v0/item/{sid}.json', timeout=5
|
|
)
|
|
story = story_resp.json()
|
|
if not story or story.get('type') != 'story':
|
|
continue
|
|
pub_ts = story.get('time')
|
|
pub_at = datetime.fromtimestamp(pub_ts, tz=timezone.utc).isoformat() if pub_ts else None
|
|
items.append({
|
|
'topic': story.get('title', ''),
|
|
'description': story.get('url', ''),
|
|
'source': 'hacker_news',
|
|
'source_url': story.get('url', f'https://news.ycombinator.com/item?id={sid}'),
|
|
'published_at': pub_at,
|
|
'search_demand_score': 8,
|
|
'topic_type': 'trending',
|
|
})
|
|
except Exception as e:
|
|
logger.warning(f"Hacker News 수집 실패: {e}")
|
|
return items
|
|
|
|
|
|
def collect_product_hunt(sources_cfg: dict) -> list[dict]:
|
|
"""Product Hunt RSS"""
|
|
items = []
|
|
cfg = sources_cfg.get('product_hunt', {})
|
|
rss_url = cfg.get('rss_url', 'https://www.producthunt.com/feed')
|
|
try:
|
|
feed = feedparser.parse(rss_url)
|
|
for entry in feed.entries[:15]:
|
|
pub_at = None
|
|
if hasattr(entry, 'published_parsed') and entry.published_parsed:
|
|
pub_at = datetime(*entry.published_parsed[:6], tzinfo=timezone.utc).isoformat()
|
|
items.append({
|
|
'topic': entry.get('title', ''),
|
|
'description': entry.get('summary', ''),
|
|
'source': 'product_hunt',
|
|
'source_url': entry.get('link', ''),
|
|
'published_at': pub_at,
|
|
'search_demand_score': 10,
|
|
'topic_type': 'trending',
|
|
})
|
|
except Exception as e:
|
|
logger.warning(f"Product Hunt 수집 실패: {e}")
|
|
return items
|
|
|
|
|
|
def collect_rss_feeds(sources_cfg: dict) -> list[dict]:
|
|
"""설정된 RSS 피드 수집"""
|
|
items = []
|
|
feeds = sources_cfg.get('rss_feeds', [])
|
|
for feed_cfg in feeds:
|
|
url = feed_cfg.get('url', '')
|
|
trust = feed_cfg.get('trust_level', 'medium')
|
|
try:
|
|
feed = feedparser.parse(url)
|
|
for entry in feed.entries[:10]:
|
|
pub_at = None
|
|
if hasattr(entry, 'published_parsed') and entry.published_parsed:
|
|
pub_at = datetime(*entry.published_parsed[:6], tzinfo=timezone.utc).isoformat()
|
|
items.append({
|
|
'topic': entry.get('title', ''),
|
|
'description': entry.get('summary', '') or entry.get('description', ''),
|
|
'source': 'rss',
|
|
'source_name': feed_cfg.get('name', ''),
|
|
'source_url': entry.get('link', ''),
|
|
'published_at': pub_at,
|
|
'search_demand_score': 8,
|
|
'topic_type': 'trending',
|
|
'_trust_override': trust,
|
|
})
|
|
except Exception as e:
|
|
logger.warning(f"RSS 수집 실패 ({url}): {e}")
|
|
return items
|
|
|
|
|
|
def extract_coupang_keywords(topic: str, description: str) -> list[str]:
|
|
"""글감에서 쿠팡 검색 키워드 추출"""
|
|
product_keywords = [
|
|
'마이크', '웹캠', '키보드', '마우스', '모니터', '노트북', '이어폰',
|
|
'헤드셋', '외장하드', 'USB허브', '책상', '의자', '서적', '책', '스피커',
|
|
]
|
|
text = topic + ' ' + description
|
|
found = [kw for kw in product_keywords if kw in text]
|
|
if not found:
|
|
# IT 기기 류 글이면 기본 키워드
|
|
if any(kw in text for kw in ['도구', '앱', '툴', '소프트웨어', '서비스']):
|
|
found = ['키보드', '마우스']
|
|
return found
|
|
|
|
|
|
def save_discarded(item: dict, reason: str):
|
|
"""폐기된 글감 로그 저장"""
|
|
discard_dir = DATA_DIR / 'discarded'
|
|
discard_dir.mkdir(exist_ok=True)
|
|
today = datetime.now().strftime('%Y%m%d')
|
|
log_file = discard_dir / f'{today}_discarded.jsonl'
|
|
record = {**item, 'discard_reason': reason, 'discarded_at': datetime.now().isoformat()}
|
|
with open(log_file, 'a', encoding='utf-8') as f:
|
|
f.write(json.dumps(record, ensure_ascii=False) + '\n')
|
|
|
|
|
|
def save_topic(item: dict):
|
|
"""합격한 글감을 data/topics/에 저장"""
|
|
topics_dir = DATA_DIR / 'topics'
|
|
topics_dir.mkdir(exist_ok=True)
|
|
topic_id = hashlib.md5(item['topic'].encode()).hexdigest()[:8]
|
|
filename = f"{datetime.now().strftime('%Y%m%d')}_{topic_id}.json"
|
|
with open(topics_dir / filename, 'w', encoding='utf-8') as f:
|
|
json.dump(item, f, ensure_ascii=False, indent=2)
|
|
|
|
|
|
def run():
|
|
logger.info("=== 수집봇 시작 ===")
|
|
rules = load_config('quality_rules.json')
|
|
sources_cfg = load_config('sources.json')
|
|
published_titles = load_published_titles()
|
|
min_score = rules.get('min_score', 70)
|
|
|
|
# 수집
|
|
all_items = []
|
|
all_items += collect_google_trends()
|
|
all_items += collect_github_trending(sources_cfg)
|
|
all_items += collect_product_hunt(sources_cfg)
|
|
all_items += collect_hacker_news(sources_cfg)
|
|
all_items += collect_rss_feeds(sources_cfg)
|
|
|
|
logger.info(f"수집 완료: {len(all_items)}개")
|
|
|
|
passed = []
|
|
discarded_count = 0
|
|
|
|
for item in all_items:
|
|
if not item.get('topic'):
|
|
continue
|
|
|
|
# 신뢰도 오버라이드 (RSS 피드별 설정)
|
|
trust_override = item.pop('_trust_override', None)
|
|
if trust_override:
|
|
trust_levels = rules['scoring']['source_trust']['levels']
|
|
item['source_trust_level'] = trust_override
|
|
item['_trust_score'] = trust_levels.get(trust_override, trust_levels['medium'])
|
|
|
|
# 품질 점수 계산
|
|
score = calculate_quality_score(item, rules)
|
|
item['quality_score'] = score
|
|
|
|
# 폐기 규칙 검사
|
|
discard_reason = apply_discard_rules(item, rules, published_titles)
|
|
if discard_reason:
|
|
save_discarded(item, discard_reason)
|
|
discarded_count += 1
|
|
logger.debug(f"폐기: [{score}점] {item['topic']} — {discard_reason}")
|
|
continue
|
|
|
|
if score < min_score:
|
|
save_discarded(item, f'품질 점수 미달 ({score}점 < {min_score}점)')
|
|
discarded_count += 1
|
|
logger.debug(f"폐기: [{score}점] {item['topic']}")
|
|
continue
|
|
|
|
# 코너 배정
|
|
topic_type = item.get('topic_type', 'trending')
|
|
corner = assign_corner(item, topic_type)
|
|
item['corner'] = corner
|
|
|
|
# 쿠팡 키워드 추출
|
|
item['coupang_keywords'] = extract_coupang_keywords(
|
|
item.get('topic', ''), item.get('description', '')
|
|
)
|
|
|
|
# 트렌딩 경과 시간 표시
|
|
pub_at_str = item.get('published_at')
|
|
if pub_at_str:
|
|
try:
|
|
pub_at = datetime.fromisoformat(pub_at_str)
|
|
if pub_at.tzinfo is None:
|
|
pub_at = pub_at.replace(tzinfo=timezone.utc)
|
|
hours_ago = int((datetime.now(timezone.utc) - pub_at).total_seconds() / 3600)
|
|
item['trending_since'] = f'{hours_ago}시간 전' if hours_ago < 24 else f'{hours_ago // 24}일 전'
|
|
except Exception:
|
|
item['trending_since'] = '알 수 없음'
|
|
|
|
# sources 필드 정리
|
|
item['sources'] = [{'url': item.get('source_url', ''), 'title': item.get('topic', ''),
|
|
'date': item.get('published_at', '')}]
|
|
item['related_keywords'] = item.get('topic', '').split()[:5]
|
|
|
|
passed.append(item)
|
|
|
|
# 에버그린/트렌드/개성 비율 맞추기
|
|
total_target = len(passed)
|
|
evergreen = [i for i in passed if i.get('is_evergreen')]
|
|
trending = [i for i in passed if not i.get('is_evergreen') and i.get('topic_type') == 'trending']
|
|
personality = [i for i in passed if i.get('topic_type') == 'personality']
|
|
|
|
logger.info(
|
|
f"합격: {len(passed)}개 (에버그린 {len(evergreen)}, 트렌드 {len(trending)}, "
|
|
f"개성 {len(personality)}) / 폐기: {discarded_count}개"
|
|
)
|
|
|
|
# 글감 저장
|
|
for item in passed:
|
|
save_topic(item)
|
|
logger.info(f"[{item['quality_score']}점][{item['corner']}] {item['topic']}")
|
|
|
|
logger.info("=== 수집봇 완료 ===")
|
|
return passed
|
|
|
|
|
|
if __name__ == '__main__':
|
|
run()
|