Data Store

What a Data Store is

A Data Store holds structured records — a customer list, a price table, a queue of things to process — that live in your workspace and persist between process runs.

You define the columns once, then add rows through the interface, a CSV import, a process, a form, or the API. Any process in the same workspace can read those rows, add to them, change them or remove them.

Data Store or Data Model?

They solve different problems. A Data Model describes the shape of data moving through a single process run — it holds no records of its own. A Data Store is where records actually live and stay. If you need the information to still be there tomorrow, or to be shared between runs, use a Data Store.

Creating a Data Store

Open Data Stores from the sidebar and choose Create. Give it a name, then define your columns.

The Data Store list. Each store shows its name and description; Create is top right.

Columns

Columns you get for free

Every Data Store automatically tracks four system columns. You cannot create, rename or write to them, but you can filter and sort by them:

Only those four exact names are reserved. A near-miss like created_on is accepted as an ordinary column, which is rarely what you want — it will sit alongside the real CreatedOn and confuse everyone later.

Column types

Nine types are available. There is no separate Date, Time, Decimal or File type — use DateTime or String instead.

Working with rows

Open a Data Store to see its rows. From here you can add records, filter, sort and page through the data, and — with the Read-Only switch flipped to Edit Mode — change it in place.

The Data Store viewer. Each column header carries its type, the key column is marked, and required columns show an asterisk. Import and Export handle CSV; the Read-Only switch protects the data while you are just looking.

Filtering and sorting

Every column header has a small filter icon. Click it to open the column's menu: sort the grid A→Z or Z→A from the top, or build a filter below — pick an operator, give a value, and press Add. Each active filter appears as a chip above the grid; remove one chip, or Clear all, to widen the view again.

Which operators the menu offers depends on the column's type — text columns offer contains and starts with, number and date columns offer greater than, less than and between, and so on.

The filter menu on a DateTime column with its operator list open. Date columns take their value from a date picker plus a time picker — both matter, as explained below.

Text matching ignores case. Searching acme finds ACME and Acme. Sorting is case-insensitive too, so apple and Apple sort together rather than in separate blocks.

Filtering dates: pick a date and a time

On a date column, equals compares the exact instant — the date and the time together. If you pick only a date, the time part defaults to midnight (00:00), so the filter matches only rows stored at exactly midnight — against real data that is usually no rows at all, with no error to tell you why.

A date column's filter takes its value from two pickers — Date and Time. Leaving Time empty means 00:00, not "any time that day".

• Looking for one exact record? Pick the date and the time with equals.
• Looking for "everything on that day"? Use between — from the day at 00:00 to the day at 23:59 — or a pair of greater than / less than conditions.

Paging

Rows come back a page at a time. Two behaviours are worth knowing if you page through data from a process or the API:

• Asking for a page beyond the end returns the last page, not an empty one. Don't loop "until the page comes back empty" — it never will. Work out the number of pages from the total count instead.
• A page size of 0 means every row, unpaged. Handy for small tables, risky for large ones.

Editing rows in the grid

The grid opens in Read-Only mode so nothing changes while you browse. Flip the switch to Edit Mode and the rows become editable in place: type into any cell, add new rows with Add Row, and remove a row with the bin icon at its end.

Edit Mode. Cells become inputs, Add Row appends a record, and the bin icon on each row deletes it. Nothing is saved yet — the edits are held until you commit.

Your edits are collected, not saved one by one. The Commit Changes button counts them as you go. When you are done, press it and a Confirm changes dialog tells you how many rows will be written and reminds you the write cannot be undone. Yes, commit changes writes the whole batch together; Discard Changes throws it away and puts the grid back the way it was; Cancel returns to editing with the batch intact.

The Confirm changes dialog guards every batch of edits: Yes, commit changes writes them, Discard Changes abandons them, Cancel keeps editing.

Two details to know: a new row must respect the schema — required columns need a value, and the primary key must be unique — and a cell edit that violates a rule is reported when you commit, not while you type. If the Edit Mode switch is greyed out, your account lacks the Data Store rows permission — see Who can see what.

Importing and exporting CSV

Both run as background jobs — you start them, and the file becomes available when the job finishes. Large files take a moment; nothing is lost if you navigate away.

Preparing a file to import

• Comma-separated, UTF-8. A byte-order mark is tolerated.
• The first row is a header of column names. They are matched by name, ignoring case and surrounding spaces, so column order doesn't matter.
• Booleans as true / false in lower case.
• Dates as 2026-07-17T14:30:00.

What happens to bad data

That last one differs from adding rows through a process or the API, where a duplicate key is an error rather than an update.

Exporting

Export produces the user columns only — the four system columns are not included. If you need CreatedOn in a spreadsheet, copy it into a column of your own.

Using a Data Store in a process

The Data Store action reads and writes rows as part of a process. Find it in the toolbox by searching for "Data Store" and drop it onto the canvas.

Search the toolbox for "Data Store" to find the action. If the toolbox shows shapes instead of actions, switch the toolbox source back to Actions — only that side has a search box.

Setting it up

1. Click the action to open its panel.
2. Pick the store under Select Data Store.
3. Choose an Operation — Select, Insert, Update or Delete.
4. Open Configure Operation to set up the details. The fields shown change with the operation you chose.

The action panel: pick the store, then the operation. Configure Operation opens the fields for the operation you chose.

What each operation does

Reading rows

Page Item Count defaults to 0, which returns every matching row. Set a number to page through instead. Total Count always reports how many rows matched your filter, not how many are on the current page — use it to work out how many pages there are. Like the viewer, a page number past the end returns the last page, never an empty one.

Configure Operation for a Select — paging, the Where builder, sorting, and the outputs.

Sort By lists your columns and the four system columns, so you can sort by CreatedOn to process the oldest records first. Tick Desc to reverse the order.

Writing rows

Set Values maps a column to the value you want written. Each entry can be a fixed value or a process variable, so an Insert can write data collected earlier in the run. Columns you leave out of the mapping are untouched by an Update.

Set Values for an Insert — one entry per column you want to fill.

A primary key column cannot be changed by an Update — put it in the Where filter to choose which row to change instead. Trying to set it produces a clear error telling you the same.

Using the results downstream

The outputs are ordinary process data. Bind Result Rows to a variable and a later action can loop over it, and Total Count can drive a Decisional — for example, branch when a lookup found nothing.

Filtering with Where

The Where builder decides which rows a Select, Update or Delete applies to. Add a condition, pick the column, pick the operator, and give a value. Add more conditions and chain them with and / or. Picking the column also decides which operators the condition offers, so always pick the column first.

You can filter on the system columns too — CreatedOn after a given date is a common way to pick up recent records.

A condition on Qty — operator "greater than or equal", value 40. Once picked, the column is not redisplayed in the collapsed row (only the operator and value show) — click the left box when you need to check or change which column a condition points at. "Add condition" chains another; the badge between two conditions toggles and/or. Sort By, Desc and the outputs sit below.

Chaining conditions: and binds before or

With three or more conditions mixing and with or, the and pairs are evaluated first — exactly as the builder's own tooltip says: "AND has priority over OR in evaluating conditions." So

City = Cluj or Qty > 50 and Active is false

means City is Cluj, or (Qty over 50 and not active) — not "one of the first two, and also not active". If you mean the other grouping, put the or branch in its own group; a group is evaluated as one bracket.

Which operators you get

The list depends on the column's type:

Operators that need a particular kind of value

• Between and Is not between need two values — a lower and an upper bound. Both bounds are included in the match; leaving the second one out fails the run with a clear message.
• Like and Not like match text with SQL wildcards: % stands for any run of characters, so %Pop% finds "Ana Pop" and "Ioana Popescu", and %Pop only values that end in "Pop". Without a wildcard, Like is an exact match — Pop alone matches only the value "Pop". If you just want "text appears anywhere", Contains is the simpler operator.
• Belongs and Does not belong match against a list. Provide it as a JSON array, for example ["Draft","Pending"]. A plain comma-separated list is not accepted.
• Is empty matches rows where the column has no value at all. A column holding an empty piece of text is not "empty" in this sense.
• On a date column, Equals compares date and time — the same rule as the viewer's filter menu; see Filtering dates. Use Between for a whole day.

Using a Data Store in a form

A form can read from and write to a Data Store through an event. The usual pattern is to show stored records in a table when the form opens; the same mechanism can add, update and delete records from the form's fields.

Showing records in a table

1. Add a table element to the form.
2. On the form, add an event with the trigger Form load and the action Run Data Store operation.
3. Choose the store and the operation Read.
4. Map the returned rows to the table's value.

A Read loads page 1 with 1,000 records per page by default. For a bigger store, raise Records per page, or better, add row filters so the form only loads what it shows.

A public form whose Form load event reads a Data Store into a table. The pager reports the number of records the read returned.

Filtering what the form reads

A Read event doesn't have to load the whole store. Open the event's Map data and filter rows panel — the row carries a Filters configured / not configured badge — and use Row filters at the bottom: + Add condition adds one condition (column, operator, value), + Add group adds a bracketed sub-group, and the badge between conditions toggles and/or.

The Row filters editor inside the event's mapping panel. It is the same condition builder the process designer uses, so the operators per column type match the Where tables above.

Writing from a form

The same event action offers Add, Update and Delete, typically on a button click, with the form's fields mapped to columns. All four operations work end to end: Add writes a new row (the primary key must be unique, exactly as everywhere else), Update changes the rows its filter matches, and Delete removes them.

Public forms

A form shared publicly reads and writes its Data Store without the visitor signing in. That access is granted workspace-wide, so the safe way to think about it is: any store in the same workspace as a public form may be reachable from the outside. Keep public forms — and the stores they use — in a workspace of their own, holding only data you are comfortable exposing, and never a store with personal or internal records.

Who can see what

Access to Data Stores is governed by workspace permissions, and viewing is separate from editing: an account can be allowed to see stores and rows yet not change them — for such an account the viewer's Edit Mode switch is disabled. A member with no Data Store permission at all cannot list stores, view their structure, or read their rows — those requests are refused rather than returning an empty result.

Troubleshooting

My filter returns nothing, but I can see matching rows

Check the operator and the value shape. On a date column, equals with only a date means midnight exactly — use between for a whole day. With Like, no wildcard means an exact match — add % (%Pop%), or use Contains. With Belongs, the value must be a JSON array such as ["A","B"]. And on a yes/no column in a process Where, use Is true / Is false rather than Equals.

My edits in the grid disappeared

Edits made in Edit Mode are held until you press Commit changes and confirm with Yes, commit changes. Navigating away — or answering the dialog with Discard Changes — abandons the whole batch.

My process fails saying a column doesn't exist

The action is pointing at a different store than the column belongs to — usually after changing Select Data Store on an action that was already configured. Re-open Configure Operation and re-pick Sort By and each Where column.

The process saved fine but does nothing when it runs

A Where condition whose operator was never chosen is saved without complaint. Open the condition and confirm an operator is selected.

My form's table only shows one record

The table's row template is static. Change it to a dynamic row — see Using a Data Store in a form.

My form filter on a list of values shows nothing (or everything)

That's the Belongs / Does not belong limitation in form Row filters — replace the list with several equals conditions joined by or. See Filtering what the form reads.

My CSV import didn't add anything

If a header doesn't match a column name, or a required column is absent, the entire import is rejected before any row is written. Compare your header row against the column names, then download the failures file for per-row reasons.

Dates are being rejected

Use 2026-07-17T14:30:00. A trailing Z, or a format like 17/07/2026, is not accepted when writing rows.