Why docgate exists

Docling, Unstructured, Marker, and pdfplumber are excellent choices for broad conversion, layout/structure extraction, built-in OCR, and rich document representations. docgate deliberately does not compete on that surface.

It is the provider-neutral control plane for extraction calls:

• plan locally before paid work;
• use any OCR or vision model behind a small protocol;
• bound raster memory even when PDF metadata is pathological;
• enforce a hard budget before each paid page;
• cache only successful non-empty results by exact provider bytes;
• stop retry storms with self-closing breakers; and
• return receipts that distinguish completed work from unknown work.

That makes docgate useful as a small standalone reader and as a safety wrapper around a larger conversion pipeline. Its usefulness depends on keeping this scope sharp: no RAG, no hosted service, no telemetry, no extraction-to-business decision layer.