# Input and portability

Choose an input route for the material you have. Manual input creates records you describe. CSV maps a table into entities. AI proposes records from pasted text. Network JSON transfers a complete validated Network. These routes have different effects.

## Enter an international company manually

**Prerequisites:** an open Network and a source you are allowed to use.

1. Choose **Create entity**, set **Entity type** to Company and enter **Name or label**. Choose **Create**.
2. Select the company, choose **Edit**, and fill only fields supported by the source. Company fields include jurisdiction, registration number, legal form, status, incorporation date, registered address, trading names and tax ID.
3. Preserve the original legal name and source spelling. Registration numbers and jurisdiction are separate fields; the same-looking number from another register is not automatically the same company.
4. Add a citation and explain any interpretation in the claim-specific note.

A field's presence does not validate a company's legal status. There are no native company-register connectors, automatic translations, document uploads or OCR in this workflow. The [Resources directory](https://lagebureau.de/resources) is a set of external research entry points.

## Map and review CSV entities

**Prerequisites:** a local UTF-8 `.csv` with a header row. The current file limit is 5 MiB and the row limit is 10,000; begin with a small sample.

1. Choose **Import CSV**, then **Choose CSV file**. Use **Download CSV template** for the current interface language if helpful.
2. Review **Delimiter**. Comma and semicolon are supported; an ambiguous header needs a deliberate choice.
3. In **Map columns**, map exactly one column to **Name / label**. Set **Default entity type**. A mapped **Entity type** column can override it per row.
4. Map other columns to registered fields, an explicitly typed **Custom imported attribute**, or **Ignore**. Do not leave a source column mapped to an unrelated field just because its name looks similar.
5. Choose **Preview import**. Inspect warnings and blocked rows. **Back to mapping** lets you correct the mapping; **Go to first blocked row** identifies a problem.
6. Choose **Import valid rows** or, when applicable, **Import valid rows only** after reviewing what will be skipped. Check the created, skipped and blocked counts.

**Expected result:** entities only, applied as one undoable batch. No relationships, Sources or Citations are created by CSV import. Similar labels produce warnings; they are not automatically merged.

Synthetic sample, saved as UTF-8:

```csv
name;kind;confidence;jurisdiction;legalForm
LB Docs — Sample DE GmbH;company;unconfirmed;Germany;GmbH
LB Docs — Sample FR SAS;company;unconfirmed;France;SAS
LB Docs — Sample IE Ltd;company;unconfirmed;Ireland;Ltd
```

Verify the jurisdiction and legal-form column destinations before applying.

## Export or transfer a Network

Choose **Export Network JSON** in the editor. The export includes the Network's records, Sources, Citations and layout; it excludes the OpenRouter key and account session. It can therefore include exact pasted source text. Store it with the same care as the working material.

In the catalog, choose **Import Network JSON** and select the exported file. Import validates the document and rejects unsupported versions, invalid references and IDs already present in local storage. It does not overwrite or merge an existing Network. Use the app's duplicate action for an independent working copy; do not hand-edit one ID while leaving references inconsistent.