Skip to content

Taking on now

RAG systems

Model answers based on your documents, with a link to the source.

What it is

Document preparation, indexing, search and answer generation with a citation. Quality is tested on your own set of questions, not by eye.

Typical problem

The model does not know your data and confidently makes things up, and there are too many documents to paste into a chat.

What you get

  • Data preparation and indexing
  • Search and answers with sources
  • Set of test questions

Who it is for

  • Startups and founders
  • Small businesses
  • Medium-sized businesses
  • Large companies

Technologies

  • pgvector
  • Embeddings

Examples. The stack is chosen for the task.

Rough guideline

$500–$2,500

All prices here are approximate planning guides, not an offer. The final cost is set after the project is assessed and fixed in the contract.

Questions

Will the AI make mistakes?

Yes, any model sometimes gets things wrong. That is why we measure quality on your examples, show the sources of each answer and pass disputed cases to a person.

Where does our data go?

This is decided before work starts: an external API with suitable terms, a private cloud or a model on your own server. The choice depends on how sensitive the data is.

Which model do you use?

We choose for the task and budget, and build the architecture so the provider can be changed.

Tell us what needs to be built

We will tell you whether it can be done, how and roughly for how much. No obligation.