Spaces:
Sleeping
Sleeping
File size: 4,375 Bytes
a316308 bc0284f a316308 bc0284f a316308 bc0284f a316308 bc0284f a316308 bc0284f a316308 bc0284f 17a7db3 a316308 bc0284f a316308 17a7db3 a316308 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | import os
import re
from urllib.parse import urlparse, parse_qs
import requests
import spaces
import gradio as gr
from groq import Groq
# Pull key from Hugging Face Settings -> Secret
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
def extract_video_id(url_or_topic: str):
parsed = urlparse(url_or_topic)
if parsed.hostname in ("youtu.be", "www.youtu.be"):
return parsed.path.lstrip("/")
if parsed.hostname in ("youtube.com", "www.youtube.com"):
if parsed.path == "/watch":
return parse_qs(parsed.query).get("v", [None])[0]
if parsed.path.startswith("/shorts/"):
parts = parsed.path.split("/")
return parts[2] if len(parts) > 2 else None
match = re.search(r"(?:v=|\/|youtu\.be\/)([0-9A-Za-z_-]{11})", url_or_topic)
return match.group(1) if match else None
def get_clean_transcript(video_id: str) -> str:
"""Fetches transcript via edge proxy to bypass cloud datacenter IP blocks."""
# Attempt 1: Free Edge-cached proxy
try:
url = f"https://youtube-transcript.ai/transcript/{video_id}.txt"
resp = requests.get(url, timeout=10, headers={"User-Agent": "Mozilla/5.0"})
if resp.status_code == 200 and len(resp.text) > 100 and "Error" not in resp.text[:50]:
# Clean markdown headers if present
lines = [l for l in resp.text.split("\n") if not l.startswith("#")]
return "\n".join(lines).strip()
except Exception:
pass
# Attempt 2: Public Invidious mirror
instances = ["https://inv.tux.pizza", "https://invidious.nerdvpn.de"]
for inst in instances:
try:
res = requests.get(f"{inst}/api/v1/captions/{video_id}", timeout=6)
if res.status_code == 200:
captions = res.json().get("captions", [])
if captions:
cap_url = inst + captions[0].get("url")
cap_res = requests.get(cap_url, timeout=6)
clean = re.sub(r'<[^>]+>', ' ', cap_res.text)
clean = re.sub(r'\s+', ' ', clean).strip()
if len(clean) > 100:
return clean
except Exception:
continue
raise ValueError(
"Could not retrieve video captions. Please test with a public video that has captions enabled."
)
@spaces.GPU
def summarize(video_url: str):
if not video_url.strip():
return "Please provide a valid YouTube URL."
video_id = extract_video_id(video_url)
if not video_id:
return "Invalid YouTube URL format. Please provide a standard watch link or shorts URL."
try:
transcript_text = get_clean_transcript(video_id)
except Exception as e:
return f"Extraction Notice: {str(e)}"
try:
completion = client.chat.completions.create(
model="llama-3.3-70b-versatile",
messages=[
{
"role": "system",
"content": (
"You are BriefTube AI, an enterprise-grade video intelligence engine. "
"Synthesize the provided transcript strictly and objectively from the source text. "
"Format cleanly into:\n"
"### 1. Executive Summary\n"
"### 2. Key Highlights & Actionable Takeaways\n"
"### 3. Critical Facts, Entities & Mentions"
)
},
{
"role": "user",
"content": f"Transcript:\n{transcript_text[:14000]}"
}
],
temperature=0.2
)
return completion.choices[0].message.content
except Exception as e:
return f"Groq LLM Synthesis Error: {str(e)}"
demo = gr.Interface(
fn=summarize,
inputs=gr.Textbox(
label="YouTube Video URL",
placeholder="https://www.youtube.com/watch?v=aircAruvnKk",
lines=1
),
outputs=gr.Markdown(label="BriefTube Synthesis & Intelligence"),
title="BriefTube AI — Video Intelligence Platform",
description="Asynchronous video summarization pipeline powered by Meta Llama 3.3 70B via Groq LPUs.",
examples=[
["https://www.youtube.com/watch?v=aircAruvnKk"]
]
)
if __name__ == "__main__":
demo.launch() |