Everything Hydra can do, in one panel

01Discover the palette

Objective

See that there are two catalogues, not one, and read what each offers.

Actions

1. Open the Workflow configuration tab and read the four categories.
2. Switch to Jobs configuration and read them again.
3. Note the number beside each category name.

Workflow configuration ⇄ Jobs configuration

The palette follows the scene. Orchestration offers what a workflow can do; a job offers what data can go through. Nothing is hidden — what a scene does not show, it cannot do.

Workflow    Actions (9)  Control Flow (2)  Containers (3)  Jobs (2)
Job         Sources (7)  Transformations (18)  Destinations (6)  Containers (3)

02Read what a workflow can do

Objective

Discover that a workflow is not limited to running jobs.

Actions

1. Expand Actions.
2. Read the nine entries from top to bottom.
3. Leave it expanded for the next steps.

Actions (9)

This is where a data pipeline becomes an operational one. A job moves rows; an action tells someone, calls something, runs a command, or remembers a value.

What you should see

Nine entries, one of them dimmed. That dimming is the subject of the next step.

03Tell someone, or tell something

Objective

Meet the three actions that carry news out of the workflow.

Actions

1. Open Log and read its subtitle.
2. Open Webhook and read the method list.
3. Open Email and note which fields carry a red asterisk.

Actions → Log · Webhook · Email

These three answer the same question — who needs to know? — at three distances. Log speaks to whoever reads the run afterwards. Webhook speaks to another system. Email speaks to a person who is not watching.

Log       ACTION   Writes a message to the run logs (always succeeds)
Webhook   ACTION   Outgoing HTTP call
Email     ACTION   Send email

04The palette knows which machine it runs on

Objective

Read the platform badges, and understand why one shell is always unavailable.

Actions

1. Find Bash and PowerShell in the list.
2. Read the badge on the right of each: Unix or Win.
3. Note which of the two is dimmed.

Actions → Bash · PowerShell

Hydra runs the command on the machine that hosts it, so only one of the two shells can ever work. The palette does not hide the other — it shows it, dimmed, with the platform it needs. You always know the whole vocabulary, and what your host is missing.

What you should see

Both dialogs are the same, only the command differs. The subtitle states the platform outright.

Bash        ACTION   Runs a Bash script (Linux / macOS)
PowerShell  ACTION   Runs a PowerShell script (Windows)

05Run Python, on any host

Objective

Find the one interpreter that never depends on the platform, and its two ways of receiving code.

Actions

1. Open the Python action.
2. Select Inline script and read the editor.
3. Select .py file and read the fields that replace it.

Actions → Python → Inline script | .py file

Python carries no platform badge, because there is nothing to warn about: it runs on Windows, Linux and macOS alike. Where a shell action ties a workflow to one host, this one keeps it portable — which is why it is the answer whenever a workflow has to travel.

Python  ACTION  Runs a Python script
  Inline script   →  PYTHON SCRIPT *
  .py file        →  .PY FILE PATH *   WORKING DIRECTORY

06Reach outside the machine, and carry values across steps

Objective

Meet the three actions that make a workflow more than a sequence of loads.

Actions

1. Open SSH and read the fields it requires.
2. Open Set Param, then Assign Param.
3. Compare the two subtitles word for word.

Actions → SSH · Set Param · Assign Param

SSH runs a command on another machine — restart a service, move a file, trigger a job on a server Hydra does not live on.

Set Param creates a value; Assign Param changes one that already exists. The distinction is deliberate: creating twice is a mistake, and Hydra refuses it rather than overwriting something you forgot about. Later steps read the value with {{ param:name }}, and shell and Python actions also receive it as an environment variable.

Set Param      Creates a runtime parameter (fails if it already exists)
Assign Param   Assigns a value to an existing runtime parameter

07Steer the graph, and group what it holds

Objective

Find the two categories that shape execution rather than perform it.

Actions

1. Expand Control Flow and read its two entries.
2. Expand Containers in the workflow scene.
3. Switch to Jobs configuration and expand Containers there too.

Control Flow (2) · Containers (3)

Condition evaluates an expression and lets downstream steps decide whether to run; Delay simply waits. Containers do not execute at all — they group nodes and hand a policy down to whatever they enclose.

Control Flow   Condition   Delay
Containers     Sequence Container   Error Scope   Retry Scope

08Your own jobs become palette entries

Objective

See that the palette is not fixed — it grows with the project.

Actions

1. Expand Jobs in the workflow scene.
2. Read the names listed there.
3. Look at the tab bar at the bottom of the job scene.

Jobs (2)

Every job you design in this project appears here, ready to be dropped into any workflow. That is what makes a job reusable: written once in Jobs configuration, orchestrated as many times as needed. The count is the size of your own library, not a product limit.

Jobs (2)
  first-job
  second-job

09Read the connectors, in and out

Objective

Learn what Hydra can read from and write to, and spot the one asymmetry.

Actions

1. In Jobs configuration, expand Sources.
2. Expand Destinations.
3. Compare the two lists entry by entry.

Sources (7) · Destinations (6)

Six connectors work in both directions: three file formats and three databases. The seventh, Web API, reads only — Hydra pulls from an HTTP endpoint but does not push to one. To send data out over HTTP, use the Webhook action in the workflow scene instead.

Sources        CSV  JSON  Parquet  MySQL  PostgreSQL  MongoDB  Web API
Destinations   CSV  JSON  Parquet  MySQL  PostgreSQL  MongoDB

10Eighteen ways to reshape data

Objective

Read the full transformation list once, so you stop reinventing what already exists.

Actions

1. Expand Transformations.
2. Scroll to the end of the list.
3. Note the last entry, Script.

Transformations (18)

They fall into four families. Knowing which family you need is usually faster than remembering a name.

Filter  Select  Rename  Cast  Aggregate  Sort  Deduplicate  Derive  Join
Fill Null  Trim  Clean  Pivot  Unpivot  Transpose  Merge  Union  Script

11Write the transformation that does not exist

Objective

Find the way out when none of the seventeen operations fits.

Actions

1. Drop a Script node onto the job canvas.
2. Read the two execution modes.
3. Use Preview to run it against real rows before saving.

Transformations → Script

Every declarative tool eventually meets a rule its authors did not foresee. Hydra answers with Script: you declare which columns come in and which go out, then write Python between the two. The contract stays readable in the YAML even though the logic is code.

What you should see

np, pd, math and re are available, and Ctrl Space completes as you type. Preview runs the code on real rows so you find out here, not during a run.

12Give the canvas room

Objective

Recover screen space once you know where everything is.

Actions

1. Select the < control at the top right of the palette.
2. Work on the canvas with the palette collapsed.
3. Select it again to bring the panel back.

NODE PALETTE → ‹

Collapsed, the palette keeps its icons: you can still drag a node onto the canvas without giving up the width. The vertical toolbar beside it stays put in both states — pan, select, undo, redo, minimap and terminal.

What you should see

The panel narrows to a strip of icons. Nothing is lost, and the canvas gains a third of the window.