c8c9c592cf
All checks were successful
news-summary-bot-cicd / build_push_deploy (push) Successful in 15m23s
extract_info에 process=False 적용하여 포맷 선택 단계를 완전히 스킵. 자막 정보는 YouTube extractor에서 직접 반환되므로 포맷 처리 불필요. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
68 lines
1.9 KiB
Python
68 lines
1.9 KiB
Python
import os
|
|
import shutil
|
|
import tempfile
|
|
|
|
import httpx
|
|
import yt_dlp
|
|
|
|
COOKIES_SRC = "/app/cookies.txt"
|
|
|
|
|
|
def extract_video_id(url: str) -> str:
|
|
"""YouTube URL에서 video ID 추출."""
|
|
if "youtu.be/" in url:
|
|
return url.split("youtu.be/")[1].split("?")[0]
|
|
if "v=" in url:
|
|
return url.split("v=")[1].split("&")[0]
|
|
raise ValueError(f"유효하지 않은 YouTube URL: {url}")
|
|
|
|
|
|
def fetch_transcript(video_id: str) -> str:
|
|
"""yt-dlp로 YouTube 자동생성 자막을 텍스트로 추출."""
|
|
url = f"https://www.youtube.com/watch?v={video_id}"
|
|
|
|
ydl_opts = {
|
|
"skip_download": True,
|
|
"quiet": True,
|
|
"no_warnings": True,
|
|
}
|
|
|
|
if os.path.isfile(COOKIES_SRC):
|
|
tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".txt")
|
|
shutil.copy2(COOKIES_SRC, tmp.name)
|
|
ydl_opts["cookiefile"] = tmp.name
|
|
|
|
try:
|
|
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
|
info = ydl.extract_info(url, ie_key="Youtube", download=False, process=False)
|
|
finally:
|
|
if "cookiefile" in ydl_opts:
|
|
os.unlink(ydl_opts["cookiefile"])
|
|
|
|
subs = info.get("automatic_captions", {})
|
|
lang = "ko" if "ko" in subs else "en" if "en" in subs else None
|
|
if not lang:
|
|
raise ValueError(f"자막을 찾을 수 없습니다: {video_id}")
|
|
|
|
sub_url = None
|
|
for fmt in subs[lang]:
|
|
if fmt["ext"] == "json3":
|
|
sub_url = fmt["url"]
|
|
break
|
|
|
|
if not sub_url:
|
|
raise ValueError(f"json3 자막 포맷을 찾을 수 없습니다: {video_id}")
|
|
|
|
resp = httpx.get(sub_url)
|
|
resp.raise_for_status()
|
|
data = resp.json()
|
|
|
|
texts = []
|
|
for event in data.get("events", []):
|
|
for seg in event.get("segs", []):
|
|
text = seg.get("utf8", "").strip()
|
|
if text and text != "\n":
|
|
texts.append(text)
|
|
|
|
return " ".join(texts)
|