.env.example

# ============================================================
# Audio-Subtitle Mismatch Detector — Environment Variables
# Copy this file to .env and fill in values before running
# ============================================================

# ----- Optional: OpenAI Cloud Whisper (leave blank to use local Whisper) -----
# Using local Whisper is free; cloud is faster but costs money.
OPENAI_API_KEY=

# ----- Optional: YouTube Data API v3 (yt-dlp works WITHOUT this key) -----
# Only needed for advanced metadata queries beyond what yt-dlp provides.
YOUTUBE_API_KEY=

# ----- Backend server configuration -----
BACKEND_HOST=0.0.0.0
BACKEND_PORT=8000

# ----- Frontend URL (used for CORS) -----
FRONTEND_URL=http://localhost:3000

# ----- Processing configuration -----
# Whisper model size: tiny | base | small | medium | large | large-v2 | large-v3
# Recommended: medium (good accuracy, reasonable speed). large-v3 for best Indic accuracy.
WHISPER_MODEL=base

# Max number of jobs running concurrently
MAX_CONCURRENT_JOBS=2

# Directory for temporary job files (cleaned up after 24h)
TEMP_DIR=./temp

# ----- Mismatch detection thresholds -----
# Score >= HIGH_THRESHOLD  => OK (green)
# Score >= LOW_THRESHOLD   => MARGINAL (yellow)
# Score <  LOW_THRESHOLD   => REVIEW (red)
HIGH_THRESHOLD=0.85
LOW_THRESHOLD=0.65