About XEYE

What is XEYE?

A hybrid search engine (text and meaning) for small catalogues of short texts: products, help articles, FAQs. You upload a list in the console, train it with one click and query a single HTTP endpoint with an API key. It is meant for people who want results by meaning without building or running their own search infrastructure. What it does and does not do, with figures, is in measured results.

Who is behind it?

Joan Martorell Coll, a Computer Engineering student at the Universitat de les Illes Balears. XEYE began as his bachelor's thesis: design a complete semantic search platform (console, model training and API) and measure how much search by meaning improves on keyword search on real Spanish catalogues. The figures from that evaluation are the ones this site publishes.

It is run by one person. That means direct answers from whoever wrote the code, and also that there is no on-call support team: see the terms of service for what you can expect.

How is it built?

• Console and website: Nuxt (Vue), served from Cloudflare.
• Backend: Java with Spring Boot and a MariaDB database; it manages accounts, lists, trainings and credit.
• Search engine: a Python (FastAPI) service that combines fuzzy text matching with sentence-transformers embeddings.
• Training: one GPU container per training (RunPod) that computes the embeddings and, if you enable it, generates descriptions with a language model.

All four components are in-house software, with automated tests and continuous deployment. The architecture and design decisions are described in the bachelor's thesis report.

Where is the data?

The main servers (database, backend and search engine) are in the European Union. Trainings run on GPUs rented by the minute, and the texts of your elements only leave for an external language model if AI descriptions are enabled for that list. All the details, provider by provider, are in the privacy policy.

Principles

• Pay per use, no fees: you only pay for what you consume, and the rates are published on the pricing page.
• Measured figures, not promises: we publish accuracy and latency with their methodology and limits.
• Your data is yours: you can disable AI per list, delete lists and elements whenever you want and remove the account with everything in it from the console.
• Honest about status: XEYE is at an early stage. The status page shows whether the service is responding.

Contact

Write to info@xeye.es for questions, proposals or incidents. Enquiries are usually answered within one or two working days. The owner's identification details are in the legal notice.