How it works

Three verbs. One product.

iRelate is built on three primitives that map to the way memory actually works. You collect. The system connects. The privacy guarantees protect.

Collect

Ingest

You bring your artifacts to iRelate — photographs, videos, voice notes, documents, anything with a story. Behind the scenes, iRelate extracts the meaningful structure: faces, places, dates, language, the relationships between artifacts.

Indexing by people, places, and moments.
Voice notes are transcribed. Documents are full-text indexed.
Visual similarity is computed for photographs.
Metadata is preserved, not overwritten.

Connect

Infer

iRelate builds a graph of your memories. Each artifact is a node. The edges are the relationships — a person across many photographs, a place across many years, a moment that ties a voice note to a photograph to a document.

Graph queries answer questions in seconds.
Vector search finds conceptually similar artifacts.
Clue fusion combines weak signals to surface strong ones.
You can ask in plain language. iRelate answers with evidence.

Protect

Preserve

Your vault is yours. iRelate is encrypted at rest, encrypted in transit, and architected so that you can export your entire memory at any time. There is no second product. There is no third-party training pipeline. There is no model that needs your data more than you do.

Full export as a portable archive.
No third-party analytics, no third-party fonts, no third-party scripts.
No training on your data. Inference is local where possible.
Open foundations. The architecture is described, not hidden.