Load your catalogue

Recommendations are similarity: “more like this” means “records whose meaning sits closest to this one”. So the setup is any semantic list — records plus an embedding model.

Create a list with a model

In the console choose Create List, name it books and pick an embedding model from the Embedding Model dropdown. The model reads each record's descriptive text into meaning, so two books about the same thing land near each other. Without a model there is nothing to measure similarity against.

The Create List dialog with the name 'books' entered and an embedding model selected

Load the records

The list's Import button takes a JSON, NDJSON or CSV file and creates the fields from your data. The description field is the one that matters — it is what the model turns into similarity. Genre, rating and year make good filters, but the recommendations come from the description.

The books list in the console, loaded with the catalogue: each record has a description plus genre, rating and year filters

From code, the same import is one request:

POST https://api.searchstack.dev/search-result/Demo/books/with-fields
X-API-Key: {your key}
Content-Type: application/json

[
  { "name": "The Glass Hours", "description": "a slow-burn mystery set in a coastal town…" },
  { "name": "North Light",     "description": "two strangers, one lighthouse, a long winter…" }
]

Go deeper: Embedding models and similarity in the reference.