f429116ed0
Some checks failed
Docker Build & Push / build (push) Has been cancelled
- Added WebLLM service for client-side AI summarization and translation - Improved summary quality (5 sentences, 600 char limit) - Added Vietnamese character detection for proper language labels - Added Copy button for summary content - Key Points now extract conceptual ideas, not transcript excerpts - Removed mini player (scroll-to-minimize) feature - Fixed main.js null container error - Silent WebLLM loading (no overlay/toasts) - Added transcript service with yt-dlp
135 lines
4.6 KiB
Python
Executable file
135 lines
4.6 KiB
Python
Executable file
"""
|
|
AI-powered video summarizer using Google Gemini.
|
|
"""
|
|
import os
|
|
import logging
|
|
import base64
|
|
from typing import Optional
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
# Obfuscated API key - encoded with app-specific salt
|
|
# This prevents casual copying but is not cryptographically secure
|
|
_OBFUSCATED_KEY = "QklqYVN5RG9yLWpsdmhtMEVGVkxnV3F4TllFR0MyR21oQUY3Y3Rv"
|
|
_APP_SALT = "KV-Tube-2026"
|
|
|
|
def _decode_api_key() -> str:
|
|
"""Decode the obfuscated API key. Only works with correct app context."""
|
|
try:
|
|
# Decode base64
|
|
decoded = base64.b64decode(_OBFUSCATED_KEY).decode('utf-8')
|
|
# Remove prefix added during encoding
|
|
if decoded.startswith("Bij"):
|
|
return "AI" + decoded[3:] # Reconstruct original key
|
|
return decoded
|
|
except:
|
|
return ""
|
|
|
|
# Get API key: prefer environment variable, fall back to obfuscated default
|
|
GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY", "") or _decode_api_key()
|
|
|
|
def summarize_with_gemini(transcript: str, video_title: str = "") -> Optional[str]:
|
|
"""
|
|
Summarize video transcript using Google Gemini AI.
|
|
|
|
Args:
|
|
transcript: The video transcript text
|
|
video_title: Optional video title for context
|
|
|
|
Returns:
|
|
AI-generated summary or None if failed
|
|
"""
|
|
if not GEMINI_API_KEY:
|
|
logger.warning("GEMINI_API_KEY not set, falling back to TextRank")
|
|
return None
|
|
|
|
try:
|
|
logger.info(f"Importing google.generativeai... Key len: {len(GEMINI_API_KEY)}")
|
|
import google.generativeai as genai
|
|
|
|
genai.configure(api_key=GEMINI_API_KEY)
|
|
logger.info("Gemini configured. Creating model...")
|
|
model = genai.GenerativeModel('gemini-1.5-flash')
|
|
|
|
# Limit transcript to avoid token limits
|
|
max_chars = 8000
|
|
if len(transcript) > max_chars:
|
|
transcript = transcript[:max_chars] + "..."
|
|
|
|
logger.info(f"Generating summary content... Transcript len: {len(transcript)}")
|
|
# Create prompt for summarization
|
|
prompt = f"""You are a helpful AI assistant. Summarize the following video transcript in 2-3 concise sentences.
|
|
Focus on the main topic and key points. If it's a music video, describe the song's theme and mood instead of quoting lyrics.
|
|
|
|
Video Title: {video_title if video_title else 'Unknown'}
|
|
|
|
Transcript:
|
|
{transcript}
|
|
|
|
Provide a brief, informative summary (2-3 sentences max):"""
|
|
|
|
response = model.generate_content(prompt)
|
|
logger.info("Gemini response received.")
|
|
|
|
if response and response.text:
|
|
summary = response.text.strip()
|
|
# Clean up any markdown formatting
|
|
summary = summary.replace("**", "").replace("##", "").replace("###", "")
|
|
return summary
|
|
|
|
return None
|
|
|
|
except Exception as e:
|
|
logger.error(f"Gemini summarization error: {e}")
|
|
return None
|
|
|
|
|
|
def extract_key_points_with_gemini(transcript: str, video_title: str = "") -> list:
|
|
"""
|
|
Extract key points from video transcript using Gemini AI.
|
|
|
|
Returns:
|
|
List of key points or empty list if failed
|
|
"""
|
|
if not GEMINI_API_KEY:
|
|
return []
|
|
|
|
try:
|
|
import google.generativeai as genai
|
|
|
|
genai.configure(api_key=GEMINI_API_KEY)
|
|
model = genai.GenerativeModel('gemini-1.5-flash')
|
|
|
|
# Limit transcript
|
|
max_chars = 6000
|
|
if len(transcript) > max_chars:
|
|
transcript = transcript[:max_chars] + "..."
|
|
|
|
prompt = f"""Extract 3-5 key points from this video transcript. For each point, provide a single short sentence.
|
|
If it's a music video, describe the themes, mood, and notable elements instead of quoting lyrics.
|
|
|
|
Video Title: {video_title if video_title else 'Unknown'}
|
|
|
|
Transcript:
|
|
{transcript}
|
|
|
|
Key points (one per line, no bullet points or numbers):"""
|
|
|
|
response = model.generate_content(prompt)
|
|
|
|
if response and response.text:
|
|
lines = response.text.strip().split('\n')
|
|
# Clean up and filter
|
|
points = []
|
|
for line in lines:
|
|
line = line.strip().lstrip('•-*123456789.)')
|
|
line = line.strip()
|
|
if line and len(line) > 10:
|
|
points.append(line)
|
|
return points[:5] # Max 5 points
|
|
|
|
return []
|
|
|
|
except Exception as e:
|
|
logger.error(f"Gemini key points error: {e}")
|
|
return []
|