Skip to content

“The bill grows faster than the usage.”

A document pipeline that costs more every month

OCR, embeddings, a vector database and the servers between them, each billed separately and running whether anyone searches or not. DocSlurp charges for pages, storage, searches and answers.

Why it happens

  • Always-on vector databases and GPU workers bill for idle hours.
  • Every page goes through the most expensive model, even when it already has a text layer.
  • Nobody can say what one customer’s documents cost.

What changes with DocSlurp

  • $0.10 per 1,000 text pages

    Pages with a text layer are read directly, not by a model. Scans, photos and complex layouts are $2.50 per 1,000.

    Details
  • From $0.005 per 1,000 searches

    Keyword search; hybrid and deep search cost more (docslurp.io/docs/http#search-modes). Storage is $0.25 per GB a month. There is no subscription, seat or minimum; credits don’t expire.

    Details
  • Cost on every response

    Each run and search reports its tokens and cost, and any API key can carry its own spend cap.

  • Or run it yourself

    The server and workers are open source and start with one Docker Compose file.

    Details

usage.ts

const { usage } = await client.search({ q: 'parental leave' });
console.log(usage.inputTokens, usage.costCents);

Sound like your setup?

Tell us what you run today and what it costs. We’ll say plainly whether DocSlurp fits.

Jump to