core/utils.py

# code/utils.py
import os
import urllib.request
import re
from core.constants import FONTS_FOLDER
import subprocess

def load_api_keys(prefix):
    try:
        prefix_lower = prefix.lower()
        keys = []
        for k, v in os.environ.items():
            if k.lower().startswith(prefix_lower):
                clean_key = v.strip().strip('"').strip("'")
                if clean_key:
                    keys.append(clean_key)
        if keys:
            print(f"✅ API Key Check: '{prefix}' के लिए {len(keys)} कीज़ मिलीं।")
        else:
            print(f"❌ API Key Check: '{prefix}' के लिए कोई कीज़ नहीं मिलीं!")
        return keys
    except Exception as e:
        print(f"🚨 एनवायरनमेंट वेरिएबल्स लोड करते समय त्रुटि: {e}")
        return []

def ensure_hindi_font():
    font_path = os.path.join(FONTS_FOLDER, 'NotoSansDevanagari-Bold.ttf')
    if not os.path.exists(font_path):
        print("-> 📥 Hindi Subtitle Font (Noto Sans) डाउनलोड किया जा रहा है...")
        url = "https://raw.githubusercontent.com/googlefonts/noto-fonts/main/hinted/ttf/NotoSansDevanagari/NotoSansDevanagari-Bold.ttf"
        try:
            urllib.request.urlretrieve(url, font_path)
            print("-> ✅ फ़ॉन्ट सफलतापूर्वक डाउनलोड हो गया!")
        except Exception as e:
            print(f"🚨 फ़ॉन्ट डाउनलोड एरर: {e}")
            return FONTS_FOLDER, "sans-serif" # Fallback
    return FONTS_FOLDER, "Noto Sans Devanagari"

def smart_text_chunker(text, max_words=150):
    """
    Splits the script into logical chunks of roughly 'max_words' length.
    It ensures splits only happen at sentence boundaries (. or । or ? or !).
    """
    # Clean up multiple spaces or newlines
    clean_text = re.sub(r'\s+', ' ', text.strip())
    
    # Split by Hindi and English sentence terminators, keeping the delimiters
    # We use a regex that splits after the punctuation mark and optional spaces.
    sentences = re.split(r'(?<=[।\.!\?])\s+', clean_text)
    
    chunks = []
    current_chunk = []
    current_word_count = 0
    
    for sentence in sentences:
        if not sentence.strip():
            continue
            
        words = sentence.split()
        word_count = len(words)
        
        # If adding this sentence exceeds the limit AND we already have content, cut the chunk here
        if current_word_count + word_count > max_words and current_chunk:
            chunks.append(" ".join(current_chunk))
            current_chunk = [sentence]
            current_word_count = word_count
        else:
            current_chunk.append(sentence)
            current_word_count += word_count
            
    # Add the last remaining chunk
    if current_chunk:
        chunks.append(" ".join(current_chunk))
        
    return chunks

# ==============================================================================
# 🕵️‍♂️ THE MATH TEACHER (FFmpeg Pause Detector)
# ==============================================================================
def get_audio_silences(audio_path, min_silence_len=0.4, silence_thresh=-35):
    """FFmpeg का उपयोग करके ऑडियो के 100% सटीक सन्नाटे (Pauses) निकालता है।"""
    cmd = ['ffmpeg', '-i', audio_path, '-af', f'silencedetect=n={silence_thresh}dB:d={min_silence_len}', '-f', 'null', '-']
    result = subprocess.run(cmd, capture_output=True, text=True)
    
    silences = []
    starts = re.findall(r'silence_start:\s+([\d\.]+)', result.stderr)
    ends = re.findall(r'silence_end:\s+([\d\.]+)', result.stderr)
    
    for s, e in zip(starts, ends):
        silences.append({'start': float(s), 'end': float(e)})
    return silences