Datasets

Overview

A dataset is a table stored in Veilus: accounts, posts, keywords, links. You import it once, assign it to profiles, and every run of an approved script on those profiles receives that profile’s share as environment variables. Schedules and batch runs need no new setting.

Each profile has two dataset slots:

Slot	Dataset kind	What a run gets
Identity	Fixed: one row per profile, kept for good	Each column as VEILUS_VAR_<COLUMN>
Content	Consume: each run takes new rows	VEILUS_VAR_ROWS, a JSON array of rows (plus each column, when the dataset takes 1 row per run)

The slot follows from the kind of dataset, so a list of accounts is never “used up” by mistake and a list of posts is never pinned to one profile. A profile has at most one dataset per slot. VEILUS_VAR_ROW_INDEX holds the row’s number (starting at 1), so a script can print it next to its result: the identity row when the profile has one, otherwise the first content row of the run.

Create a dataset

1. Open Datasets in the sidebar and click New from file.
2. Choose a .csv or .txt file. Columns may be separated by ,, | or ;: the app detects the Delimiter from the first line (| first, then ;, then ,) and you can change it by hand. First line is column names is on by default; turn it off and the columns are named COL_1, COL_2… with the first line read as data. Quoted values may contain the delimiter and line breaks. A file with no delimiter is one value per line, in a column named VALUE.
3. Check the preview of the first 20 rows. Lines that cannot be read are listed by line number.
4. Enter a Dataset name (the file name is filled in).
5. Under Kind, pick Fixed or Consume. For Consume, set Rows per run (1 to 50).
6. Tick the Secret columns, such as passwords.
7. Click Import.

Column names become upper case, with any character other than A-Z, 0-9 and _ turned into _ (First name becomes FIRST_NAME). The preview shows the final names. ROWS, ROW_INDEX, PROFILE_ID, RUN_ID and DEBUG_PORT are reserved and cannot be column names.

Assign by profile name: if a Fixed dataset has a PROFILE_NAME column, you can tick Assign rows to profiles by the PROFILE_NAME column. Each row then goes to the profile with that name. Names that match no profile, or more than one, are listed after the import.

On the dataset’s page

Click a dataset in the list to see its rows, with the profile each row belongs to and its state: Available, In use, Used or Unassigned. From there:

• Add rows from file appends rows from another .csv or .txt file.
• Export saves the dataset as .csv. Secret columns are included only if you tick Include secret columns, and then they are written as plain text.
• Return used rows (Consume datasets) makes used rows available again.

Deleting a dataset from the Datasets list deletes its rows, and the profiles using it lose that data. This cannot be undone.

Assign it to profiles

Select profiles in the profile list and click Assign dataset in the bulk action bar, then choose the dataset. It goes into the Identity or Content slot according to its kind.

• Shortage: a fixed dataset gives each selected profile the next unassigned row. If rows run out, the profiles that got no row are listed by name. Two profiles never share one row.
• Replace: if a profile already has a different dataset in that slot, the assignment is refused unless you tick Replace a dataset already in this slot.
• Column clash: if the dataset has a column with the same name as one in the profile’s other slot, the assignment is refused and the clashing column is named.

When you create profiles with Batch create, the Organize step has Identity dataset and Content dataset selects, so new profiles are assigned as they are created.

Open a profile’s panel and its Data tab to see the Dataset section: the profile’s Identity row (secret values masked) and, for Content, rows per run and how many rows are left. Each slot has a Remove button. A profile assigned a Fixed dataset but given no row fails early when it runs.

The Automation tab lists, under Variables from datasets, the variable names a script will receive from each slot. They are read-only and override a manual variable of the same name.

How consume rows are used

Before a run starts, the profile reserves up to rows per run unused rows, lowest row number first. If fewer remain, it gets what is left; the script can read the length of ROWS.

• A run that succeeds marks its rows as used.
• A run that fails returns all of its rows to the dataset, so nothing is lost.
• Profiles running at the same time never get the same row.
• When no unused row is left, the profile fails before the browser starts, with an error saying the dataset is out of unused rows. Add rows, or use Return used rows on the dataset’s page to make used rows available again.

Read the data in a script

Values arrive as environment variables. Fixed (identity) columns and, when a content dataset takes 1 row per run, its columns too:

const user = process.env.VEILUS_VAR_USERNAME;
const password = process.env.VEILUS_VAR_PASSWORD;

With more than one row per run, parse ROWS:

const rows: Array<Record<string, string>> = JSON.parse(
  process.env.VEILUS_VAR_ROWS ?? '[]',
);
for (const row of rows) {
  console.log(row.KEYWORD);
}
console.log('row', process.env.VEILUS_VAR_ROW_INDEX);

If the same name is set in several places, run variables win over dataset rows, and dataset rows win over a profile’s stored variables.

Approved scripts only

Dataset values reach approved scripts only. A trial run of an unapproved script gets none of them and reserves no rows; pass the values it needs as run variables instead.

Security

• Secret columns are masked in the app and are never returned over the local API or MCP. They reach approved scripts, because the script needs them.
• Dataset values are not encrypted at rest. They are stored on your computer in plain text, the same way profile variables are.
• An approved script can still print a value it received. Review a script before you approve it.

Use datasets through Claude (MCP)

An agent connected over MCP can manage datasets with create_dataset, append_dataset_rows, list_datasets, get_dataset_rows, assign_dataset, unassign_dataset and reset_dataset_rows. create_profiles also accepts an identity and a content dataset. Listing and reading rows never return secret column values. For example, give Claude a spreadsheet of accounts and ask it to create a fixed dataset and assign it to your profiles.