Variables in the dashboard

An environment page is where most of the day to day work happens. It lists every variable in one environment, and every value on it is ciphertext until your browser decrypts it locally. Names are stored with the casing you give them and are unique inside an environment without regard to case, so api_key collides with an existing API_KEY instead of sitting beside it. The page lives at https://envless.cloud/dashboard/products/<product-slug>/projects/<project-slug>/environments/<environment-slug>, and the same URL with /raw on the end opens the whole environment as one editable .env document.

Two views of the same environment#

The tabs sit above the search box and carry the current query string across, so a filtered table keeps its filter when you switch back.

Add variables#

Creating variables and importing a .env are one modal with two tabs, not two separate flows. Open it with the Add variables button in the page header, or by pressing c anywhere on the page. The shortcut is ignored while focus is in a text field, while a dropdown or dialog is already open, and while Ctrl, Cmd or Alt is held, so it never fights with copy. It is only armed on the table view, for members who hold variables:create on an unlocked project.

The modal's own subtitle states the rule the whole page rests on: values are encrypted in your browser before they leave.

The Editor tab#

One row per variable. Each row carries:

Add another appends a row, and the trash button removes one. Removing the only row clears it rather than leaving you with none. The submit button counts what it will do, so it reads Create for one row and Create 4 variables for four, and Enter submits it (only on this tab, the Raw tab has no Enter shortcut).

Two duplicate checks run as you type. A name that already exists in this environment is flagged with A variable with this name already exists in this environment., and a name used twice in the batch is flagged on the second row with This name is used by another row above. Both compare without regard to case, so typing api_key where API_KEY already lives is flagged as a duplicate rather than accepted as a second variable. Either one disables the submit button, as does any row with an empty name. An empty value is allowed here and stores an empty value, which the table later renders as empty.

Paste to expand#

Pasting a block of KEY=value lines into a name or value field on the Editor tab expands it into rows instead of dropping the text into one field. The paste is only intercepted when it looks unambiguously like an env block: at least 2 non blank, non comment lines, and every such line matching ^(export\s+)?[A-Za-z_][A-Za-z0-9_]*\s*=. One stray prose line and the paste is treated as ordinary text.

This is the one path that infers a type from the value:

At most 200 rows are kept. If the paste held more, the extras are dropped with Only the first 200 were added. Use the Raw .env tab for larger files.

The Raw .env tab#

Three ways to get a file in front of the importer, all landing on the same tab:

Drag and drop accepts one file per drop. A file is taken when its name matches \.env(\..+)?$ case insensitively, so .env, .env.production and .env.local are all fine, or when its MIME type starts with text/, which quietly lets a plain .txt through as well. Anything else is refused with an Unsupported file toast naming what was skipped. Dropping is only enabled for members who can create variables on an unlocked project.

What the importer accepts#

The importer parses the text on every keystroke and rebuilds the preview. Its rules are deliberately forgiving:

The per row preview#

Every parsed entry becomes a preview row showing Name, Value, Status and an Action select, checked against the variables already in this environment:

Set any row to Skip and it is left alone entirely. The header counts the three buckets (3 add, 2 replace, 1 skip) and the submit button spells out the batch: Add 3, Replace 2, or Add 3, replace 2. With everything skipped the count is zero and the button is disabled.

Adds go out as one bulk create. Replaces go out as one update per row, in parallel, and each one is independent: if some fail you get a Variables not replaced toast reading Failed to replace 2 variables., while the ones that succeeded stay saved.

Editing the whole environment as text#

The Raw view builds one .env document out of the environment, decrypting every value to do it, so it needs the workspace key. You edit the text, and Envless diffs your document against the current variables to produce a plan of creates, updates and deletes, which a confirmation modal shows before anything is written. A name in the document that is not in the environment is a create, a name whose value differs from the decrypted one is an update, and a variable whose name is no longer anywhere in the document is a delete. Removing a line therefore removes the variable. The diff matches names without regard to case, so API_KEY and api_key are the same variable to it.

The six parser diagnostics#

This view uses a strict parser, and every problem it finds is reported as Line <n> plus a message, under a heading that counts them (3 lines need fixing before you can apply, or 1 line needs fixing before you can apply). While any diagnostic is present the apply is blocked: the editor says so under the text area, and the confirmation modal keeps its apply button disabled.

A line that produces a diagnostic is not guessed at, it is left out of the parse entirely. Blank lines and # comment lines are ignored and never produce one. If your edits parse cleanly but change nothing, you get Your edits do not change any values, so there is nothing to apply. instead.

Reading values in the table#

Values are masked as •••••••••••• until you ask for them. The eye button on a row decrypts that one value in your browser and caches the plaintext against the ciphertext it came from, so re-revealing is free until the value changes. Three things a cell can say instead of a value:

Hovering a row exposes a copy button next to the name (copies the name) and next to the value (decrypts first, then copies the plaintext). Neither the API nor the server is involved in that decryption.

Inline edit#

Double click a name or a value cell to edit it in place. It is available to members with variables:update on an unlocked project, and the cell advertises itself with a double-click to edit tooltip. The input selects its contents on open.

An empty name is refused, and so is an empty value: the input turns red and nothing is sent. (Note the asymmetry with the create form, which does accept an empty value.) Saving a value that matches the current plaintext, or a name that matches the current name, closes the editor without a request. Everything else encrypts the new value in the browser and sends a single update, then patches the row in place rather than refetching the list.

The row menu#

A member who can read but not update or delete gets a single history button instead of the menu.

Search#

The search box is labelled Search names and values... and it means both. Names are matched client side, case insensitively, on a substring. Values are matched the same way, but only after they have been decrypted, which is the part worth understanding.

When a search is active, Envless decrypts every non deleted variable in the environment in your browser, 8 at a time, and caches each plaintext against its ciphertext so typing another character does not redo the work. Two notices sit above the table when that scan finishes with something to report:

The input is debounced by 250 ms into the store the URL mirrors, then debounced again by the same amount before it filters the table and starts the value scan, so a search takes about half a second to land. The spinner in the search icon means the input has not settled yet, or the list is revalidating in the background.

Search and sort are mirrored into the URL as q and sort, so a filtered view is a link you can share or reload. Sort accepts name (the default, ascending), updated (newest first, falling back to the created time) and created (newest first). The table is not paginated: every matching variable is rendered.

Bulk actions#

The checkbox column appears for members with variables:delete on an unlocked project. Selecting anything opens the selection bar above the table, which also appears on its own when values are revealed, so you always have one click to put them back.

The kebab button on the bar holds the bulk actions:

Reveal decrypts the selection as one batch, 8 values at a time, and re-reads the current state before writing the results back, so revealing one group while another is still decrypting does not clobber it. Any value that fails raises The value could not be decrypted. Is this the right workspace key? and the rest still reveal.

Export#

The same three export actions appear in two places. In the page header they sit behind the kebab attached to the Add variables button (or behind a plain Actions button when you cannot add variables), and there they cover the whole environment. In the selection bar's kebab they cover only what you selected, and the labels gain a "selected". They are hidden on a locked project and when the environment has no variables.

.env.example, no key required#

Copy example writes one bare NAME= line per variable, with a trailing newline. No value is read, nothing is decrypted, and the workspace key is never touched, so this works on a locked workspace and is the safe thing to paste into a repo or a ticket.

DATABASE_URL=
STRIPE_KEY=
PORT=

The real .env, key required#

Copy and Download produce the actual file, so both decrypt every variable first and both refuse without an unlocked key, prompting Unlock your workspace key first to continue. with an Unlock action on the toast.

Downloads are named .env for a whole environment with no name available, and .env.<environment> otherwise, where the environment name is lower cased, anything outside a-z0-9._- becomes -, and leading and trailing dashes are trimmed. A name that sanitizes to nothing falls back to .env.export.

Quoting#

A value is written bare unless it needs quoting. It is wrapped in double quotes when it is empty, or when it contains any of:

Inside the quotes, \ becomes \\, " becomes \", a newline becomes \n, a carriage return becomes \r and a tab becomes \t. A single quote is left as it is, since it is harmless inside double quotes. Those are exactly the escapes the importer decodes, so a multi line value survives an export and re-import intact.

Deleting#

Both delete paths are type to confirm, and they ask for different things:

A selection of more than 200 is refused before the dialog opens, with At most 200 variables can be deleted at once. Clear some of the selection and try again. Clear part of the selection and run it twice.

Limits and things that will stop you#

Permissions#

Every one of them is additionally gated on the project being unlocked. Roles are managed under Role and Access.