Action Nodes · DeutschlandGPT

Action nodes perform discrete operations: fetching data, running code, generating images, branching on conditions, and looping over arrays. Each node has typed inputs (wired from previous nodes or typed directly) and named outputs (available to downstream nodes).

HTTP Request

Sends an HTTP request to any URL and returns the response.

Input	Required	Description
URL	Yes	The full URL to send the request to
Method	Yes	HTTP method: GET, POST, PUT, PATCH, DELETE
Headers	No	Key-value pairs added as request headers
Body	No	Request body (JSON, form data, or raw text)
Response Type	No	How to parse the response: `json` (default), `text`, or `file`

Output	Type	Description
`response`	any	Parsed response body
`statusCode`	number	HTTP status code
`file`	WorkflowFile	The downloaded file, only when Response Type is `file`

Web Search

Searches the web and returns a list of results.

Input	Required	Description
Query	Yes	The search query
Max Results	No	Number of results to return (1-100, default 10)

Output	Type	Description
`results`	array	Array of result objects, each with `title`, `url`, `snippet`
`resultCount`	number	Number of results returned

RSS Feed

Fetches and parses an RSS or Atom feed.

Input	Required	Description
URL	Yes	The RSS/Atom feed URL
Max Items	No	Maximum number of feed items to return (1-100, default 20)

Output	Type	Description
`items`	array	Array of feed items, each with `title`, `link`, `description`, `pubDate`, `author`
`feedTitle`	string	The feed's title
`feedDescription`	string	The feed's description
`itemCount`	number	Number of items returned

Custom Code

Executes a JavaScript snippet. Use this for data transformation, string manipulation, math, or any deterministic logic that doesn't require an AI call.

Input	Required	Description
Code	Yes	JavaScript code to execute. Previous node outputs are available as global variables named after the node (e.g. `trigger.outputs.formData`). Use `dg.log()`, `dg.warn()`, `dg.error()` for debug output visible in the execution log.

Output	Type	Description
`result`	any	The value returned by the last expression in your code, or the value of `return`

Custom Code nodes run in a sandboxed Node.js environment. They have no access to the file system, network (use HTTP Request for that), or external modules. Execution is synchronous and has a timeout.

Example:

const items = trigger.outputs.formData.items;
return items.filter((i) => i.status === 'open').map((i) => i.title);

Image Generation

Generates an image from a text prompt using an AI image model.

Input	Required	Description
Model	Yes	The image generation model to use
Prompt	Yes	Text description of the image to generate

Output	Type	Description
`image`	WorkflowFile	The generated image as a workflow file
`revisedPrompt`	string	The prompt as interpreted by the model (some models rewrite prompts)

Data Table Nodes

Data table nodes read from and write to persistent Data Tables. Unlike node outputs, data table content persists across workflow runs.

Get Rows

Retrieves rows matching optional filter conditions.

Input	Required	Description
Table	Yes	The data table to read from
Filters	No	Column conditions to filter rows
Sort By	No	Column to sort results
Sort Direction	No	`asc` or `desc`
Limit	No	Max rows to return

Output	Type	Description
`rows`	array	Matching rows as objects
`rowCount`	number	Number of rows returned

Insert Row

Adds a new row to a data table.

Input	Required	Description
Table	Yes	The target data table
Row Data	Yes	Column → value mapping for the new row

Output	Type	Description
`rowId`	string	ID of the newly inserted row
`row`	object	The full inserted row

Update Rows

Updates all rows that match a filter.

Input	Required	Description
Table	Yes	The target data table
Filters	Yes	Conditions identifying rows to update
Updates	Yes	Column values to set on matched rows

Output	Type	Description
`updatedCount`	number	Number of rows modified

Delete Rows

Removes all rows that match a filter.

Input	Required	Description
Table	Yes	The target data table
Filters	Yes	Conditions identifying rows to delete

Output	Type	Description
`deletedCount`	number	Number of rows removed

Flow Control Nodes

Condition

Evaluates a JavaScript expression and routes execution to one of multiple branches depending on the result.

Each branch has a name and a condition expression. The first branch whose expression evaluates to true is taken. An optional default branch runs if no other branch matches.

Example conditions:

// Branch "High priority"
trigger.outputs.priority === 'high';

// Branch "Needs review"
agent1.outputs.score < 0.7;

Loop Start / Loop End

Iterates over an array, running the enclosed nodes once for each item.

Loop Start inputs:

Input	Required	Description
Array	Yes	The array to iterate. Reference a previous node output that is an array.

Within the loop body, the Loop Start node exposes:

item: the current iteration's value
index: zero-based index of the current item

Connect the last node in the loop body back to Loop End to close the loop.

Loop End outputs:

Output	Type	Description
`results`	array	Array of each iteration's output (collected from the node connected to Loop End)
`iterationCount`	number	Total number of iterations

Subworkflow Call

Calls another workflow (that has a Subworkflow Trigger) as if it were a function. Execution pauses until the subworkflow completes, then resumes with its outputs.

Input	Required	Description
Workflow	Yes	The subworkflow to call
Parameters	Depends	Input values expected by the subworkflow's trigger

Outputs are whatever the subworkflow's trigger defines as its return values.