This project processes arbitrary CSV data provided via HTTP or MCP tools. To limit resource usage:

• CSV files are capped at a configurable size (default 50 MB, max 1 GB).
• Only HTTP/HTTPS URLs are allowed for remote datasets.
• Uploaded datasets and validation results are stored ephemeral in-memory by default. Use --storage-backend sqlite:///path/to/gx.db to persist them.

• Place the HTTP server behind a reverse proxy (e.g., Nginx, Caddy, cloud LB) to terminate TLS/HTTPS.
• Monitor resource usage; the SQLite backend can grow unbounded if not pruned.
• Anonymous validation sessions use random UUIDv4 handles. For persistent user sessions:
  • Generate handles with secrets.token_urlsafe(32)
  • Compare them using hmac.compare_digest to prevent timing attacks.

Open a GitHub issue or email dfront@gmail.com