The task editor

The task editor is the full-page form at #/dag/<id>/task/<task_id> where you configure one task of a DAG — its command, template variables, project code, dependencies, and retry behavior. Like everywhere else in cronova's web console, there is no save button: every change auto-saves.

You reach the task editor from a DAG's Structure tab (see Working with a DAG) — click a task row, or add a new task. The breadcrumb and the ← back button return you to the DAG page.

Task ID and type

Field	Behavior
Task ID	The rename applies when the input loses focus (blur), not on every keystroke. IDs must be non-empty, unique within the DAG, and match [A-Za-z0-9][A-Za-z0-9_.-]*. An invalid or duplicate ID shows a warning toast and the input snaps back to the old ID. A successful rename also rewrites every sibling task's dependency that pointed at the old ID.
Type	One of shell, python, sql, jar, http. Switching the type immediately re-renders the form below it with that type's fields — the stored command is kept, so you can switch back without losing work.

The visual command editor

For shell, python, and sql tasks (and the URL/headers/body of http tasks), the command field is a visual template editor. Plain text stays plain text, but every {{ name }} template variable renders as a color-coded pill — an atomic token you can move around but not accidentally edit character-by-character. Pill colors follow the variable kind:

Pill kind	Pattern	Meaning
Built-in	{{ logical_date }} etc.	Per-run values injected by the scheduler
Variable	{{ var.KEY }}	A shared variable from the store
Connection	{{ conn.id.field }}	One field of a stored connection
Param	{{ params.key }}	A manual-trigger parameter

Working with pills:

• Insert a pill by clicking a chip in the palette, by dragging a chip into the editor at the exact spot you want, or with the keyboard — chips are focusable, and Enter / Space inserts at the caret.
• Move a pill by dragging it to a new position inside the editor; a caret marker shows where it will land.
• Delete a pill with its × button, or place the caret next to it and press Backspace / Del.
• Multi-line commands work naturally: Enter inserts a newline, and pasted code is inserted as plain text with its indentation intact. Single-line fields (like the HTTP URL) collapse pasted newlines to spaces.
• Hover a pill for a tooltip describing what the variable resolves to.

Below the editor, the Will run: preview shows the exact stored string — your command with literal {{ }} placeholders highlighted. Pills are only a view; what cronova persists and renders at run time is that plain template string, identical to what you would write in DAG YAML.

Tip

The editor pill-ifies any {{ dotted.name }} token — exactly the set the scheduler substitutes. If you paste a command that already contains template placeholders, they become pills automatically. See Template variables for what each one resolves to.

The variable palette

Above the editor sits a grouped palette of insertable variables:

Group	Contents	How it inserts
built-in	The six run variables: logical_date, logical_datetime, run_id, dag_id, task_id, try_number	Click, drag, or keyboard
variables	One chip per key in the variable store (var.KEY)	Click, drag, or keyboard
connections	One chip per stored connection (conn.id)	Click opens a field menu — pick host, port, login, or password to insert {{ conn.id.field }}. Dragging the chip inserts .host.
params	A free-form key input	Type a key (e.g. day) and press Enter to insert {{ params.day }}

When no variables or connections exist yet, the group shows a set up link straight to the Resources page. See Variables, connections & params for how the values are defined and injected.

Per-type forms

Each task type renders its own fields (full details in Task types):

Type	Fields
shell	Command pill editor + Will run: preview, plus the Project section below.
python	Python code pill editor. The code runs inline via python3 -c; CRONOVA_* variables are readable from the environment, and a non-zero exit means failure.
sql	Connection — the id of a configured connection (its type picks the driver: postgres/mysql/sqlite) — and the SQL query pill editor.
jar	A structured form: Jar path, Main class (optional), Arguments. The form composes java -jar app.jar … (or java -cp jar main …) and shows it in the Will run: preview. An edit raw command link is the escape hatch to free-form editing; use form switches back, but only if the raw command still parses into the form's shape.
http	Method (GET/POST/PUT/PATCH/DELETE/HEAD), URL, Headers (one per line, Key: Value), Body, and Expected status (comma-separated, e.g. 200,201; empty accepts any 2xx). URL, headers, and body are all pill editors, so {{ var. }} and {{ conn. }} work in each.

Project (shell tasks)

A shell task can attach an uploaded project — a directory of your scripts and data files. The Project section lets you either pick an existing project from the dropdown or open the Upload / new project panel without leaving the editor:

• Upload files / folder — choose files, choose a whole folder, or drag files/folders onto the drop zone. Dropping a .zip auto-extracts it.
• Write a script — type a filename (e.g. main.py) and its content inline for quick one-file projects.

Give the project a name (letters, digits, . _ -) and hit Upload; the task is pointed at it automatically. At run time every attempt gets a fresh, isolated copy of the project as its working directory, with the copy's path exported as CRONOVA_PROJECT_DIR — see Run your own scripts & projects for the full staging model.

Note

Uploads are additive: re-uploading under the same name upserts files, so pushing one changed script does not require re-sending the whole folder. Running attempts keep the copy they started with; the change takes effect on the next run.

Dependencies and trigger rule

Every other task in the DAG appears as a chip under Depends on — click a chip to toggle it as an upstream of this task. The editor rejects any toggle that would create a cycle: the chip flips back and a "Dependency cycle detected" toast appears, because a DAG must stay acyclic.

The Trigger rule select decides when the task may run relative to its upstreams; a one-line description of the selected rule is shown right under the field:

Rule	Runs when
all_success	All upstreams succeeded (default)
all_done	All upstreams finished, success or not — good for cleanup/summary
one_success	Any one upstream succeeds
one_failed	Any one upstream fails — good for alerts
all_failed	All upstreams failed
none_failed	No upstream failed (succeeded or skipped)

See Dependencies & trigger rules for worked examples.

Advanced options

The Advanced options group is collapsed by default and opens automatically when any of its values is set:

Field	Meaning
Pool	Named concurrency slots shared across all DAGs; tasks in the same pool compete for its slots. Pools are managed on the admin pages.
Priority	Higher runs first when tasks contend for the same pool.
Retries	Attempts after a failure; empty inherits the DAG's default retries.
Retry delay (s)	Seconds to wait between attempts; empty inherits the default.
Timeout (s)	Kill a single execution after this many seconds. 0 = none.
Task SLA (sec)	Measured from run start; alert if the task hasn't finished in time. 0 = off.

Tuning guidance lives in Retries, timeouts & pools.

Autosave

The badge next to the task title reports the save state of the whole DAG:

Badge	Meaning
Saved	Everything is persisted.
Saving…	A debounced save is in flight (edits are batched over ~half a second).
Fix errors to save	Validation failed — the errors are listed at the bottom of the form, and nothing is written until they're fixed.
Save failed	The server rejected the save; the badge shows the error message.

Validation covers empty or duplicate task IDs, an empty command (for non-HTTP types), a missing URL on http tasks, a missing connection on sql tasks, and dependency cycles.

Warning

Edits save against the live DAG the moment they validate — there is no draft mode. If the scheduler starts a run while you are mid-edit, that run uses whatever was last saved.

Common questions

Where is the save button? There isn't one. The workflow scheduler UI saves the whole DAG automatically a moment after each valid change — watch the badge flip from Saving… to Saved.

How do I use a secret in a command? Don't paste it. Store it as a connection or variable, then insert a {{ conn.id.password }} or {{ var.KEY }} pill — the value is substituted at run time and never lives in the DAG definition.

Can I type {{ }} templates by hand instead of using pills? Yes. Pills are just a visual layer over the stored template string; anything you paste or type that matches {{ name }} becomes a pill, and the Will run: preview always shows the raw string that gets stored.

Why didn't my task rename take effect? The rename applies on blur — click or tab out of the ID field. If the new ID is empty, invalid, or already used by a sibling task, the editor shows a toast and keeps the old ID.