# Importing data

The import form quietly does three things, and the last one is a checkbox most
people scroll past. Here is all three.

```mermaid
flowchart LR
  I["Upload · paste · URL"] --> F["Project file<br/>data/products.csv"]
  F --> M["Model<br/>existing · inferred · none"]
  F --> R["Resource entry<br/>in dcupl.lc.json"]
```

## 1. Your rows become a file

An **upload** or a **paste** is written to the path shown under **Upload Path** —
the field starts at `data/`, and picking `products.csv` makes it
`data/products.csv` — and is a project file from then on. You can open
it in the file editor, replace it, or push and pull it with the CLI like any
other file.

An **external URL** is different: it is not copied. The import points at the data
where it already lives, and every load re-fetches it from there. Use it for a
feed you do not own; use an upload when you want the rows versioned with the
project.

## 2. The rows are loaded into a model

**Model handling** decides which model the file feeds:

- **Existing Model** — point the file at a model key that is already in the
  project.
- **Create New Model** — the importer reads a sample of the file, infers the
  attributes, and hands you straight into the model creation form. For a file
  over 10 MB it infers from the first 2 MB instead of reading the whole file;
  that only works for CSV, so a large JSON file falls back to **Skip Model
  Creation**.
- **Skip Model Creation** — no model file is created or assigned.

Skipping does not cancel the import. The resource entry is still written, with
`"model": "UNDEFINED_MODEL"` when you gave no model key. The rows still load:
with the loader's default options a data resource whose model is not declared
gets a model of that key, inferred from the rows. Set the entry's `model` to the
key you want before you build on it — the console will not remind you.

## 3. Nothing loads until a resource entry exists

The checkbox at the bottom of the form, **Create a new resource in the loader
configuration**, is **on by default**. It adds one `data`
entry to `dcupl.lc.json`: this URL, the model it feeds, and any tags you gave it.
That entry is what makes the file part of an application — a file sitting in the
project with no resource entry is never fetched by anything.

Untick it only when you deliberately want to stage a file that no application
fetches yet. If you do, remember to add the resource entry by hand later, either
in the file editor or by re-running the import.

Tags you enter here are the same tags applications select on. A file tagged
`catalog` reaches every application that selects `catalog` and every application
with no tags, and stays out of applications that select only other tags.

## Next steps

- [Models in the Console](/console/models) — the shape the rows land in
- [Loader configuration](/console/loader) — what the resource entry looks like
- [Applications](/console/applications) — run the app and see the rows