LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
VITE_BACKEND_URL=http://localhost:8000
# Analytics - PostHog (set to false to disable tracking)
# Default: false for local development
VITE_ENABLE_POSTHOG=false
# For development
NODE_ENV=development
# Enable file watching via polling instead of filesystem events
# These are critical for hot reloading to work properly in Docker containers
# as the normal filesystem event notification doesn't work reliably with mounted volumes
CHOKIDAR_USEPOLLING=true
# HMR WebSocket port - must match the externally accessible port (Traefik = 80)
# NOT the internal Vite port (3000), since the browser connects through the proxy
WDS_SOCKET_PORT=80