Batch-Transformation

Was es tut

Batch Transform is a built-in tool that processes a CSV file row by row and adds AI-generated columns to it. You give it a source CSV, a task that describes what to do for each row, and the columns you want to add. It returns a new file with those columns populated.

You attach Batch Transform to an Autonomous Agent as a tool. The agent calls the tool while reasoning, passing it the source file and the task, and uses the transformed file in the rest of its work.

The node is labeled Batch Transform in the palette, in the built-in tools category.

When to use this

Use Batch Transform when an agent needs to enrich or classify the rows of a CSV in bulk, and each row needs the same AI-driven treatment.

Typical tasks:

• Classify each row into a category and write the label to a new column.
• Extract a value from a free-text field into its own column.
• Score, summarize, or tag each row based on its existing columns.

For pulling structured data out of a single document rather than transforming tabular rows, connect a Document extraction resource instead. To analyze the content of files that are not row-oriented CSVs, use the Analyze Files built-in tool.

Adding the tool

Add the tool to the canvas, then connect it to an agent.

1. Select an Autonomous Agent node, or add one if your flow doesn't have it yet.
2. Open the node palette and select the Built-in tools category.
3. Select Batch Transform. Flow adds the node to the canvas.
4. Draw an edge from the Batch Transform node to the agent's tool handle.
5. Select the Batch Transform node to open its properties panel and configure it.

Result: The Batch Transform tool is connected to the agent. The agent can now call it while reasoning.

Configuration reference

Configure these fields in the Batch Transform Settings section of the properties panel.

Feld	Erforderlich	Standard	Beschreibung
Source (file)	Ja	Keine	CSV file to transform. Reference a file produced upstream in the flow, or one the agent supplies when it calls the tool.
Batch Transform Task	Ja	Keine	Instructions the tool applies to each row. Describe what to analyze, what to extract, and how to populate the output columns.
Output columns	Ja	Keine	Columns the tool adds to the file. Define at least one column with a name and description, and add up to 10 columns.
Spaltenname	Ja	Keine	Header for the new column in the output file.
Beschreibung	Ja	Keine	Content the column should contain. The tool uses this value to decide what to write for each row, so be specific about the expected format.
Web search grounding	Nein	Aktiviert	Whether the tool can use web search to ground or enrich its results. Select Disabled to use only the source file and task.

Eingaben und Ausgaben

Eingabe

Batch Transform takes its input from the fields you configure: the source CSV, the task, the output columns, and the web search grounding setting. Because it runs as a tool, the calling agent supplies these values when it invokes the tool during a run.

Ausgabe

Batch Transform returns a file: the source CSV with the output columns appended and populated. Access it at $vars.<nodeName>.output, where <nodeName> is the name of the Batch Transform node.

$vars.batchTransform1.output

If the tool fails, the node populates an error object instead. Refer to Error handling for the error shape and how to route on it.

{
  "output": "file. The source CSV with the configured output columns added and populated.",
  "error": "object. Populated only on failure. See Error handling."
}

Beispiel

Triage a CSV of support tickets

An agent receives a CSV of support tickets and needs each ticket categorized and prioritized before it routes them.

Connect a Batch Transform tool to the agent and configure it as follows.

Source (file): the tickets CSV, for example a file passed in from the start of the flow.

Batch Transform Task:

For each ticket, read the Subject and Body columns. Classify the ticket into exactly one category: billing, technical, or account. Then assign a priority of high, medium, or low based on the urgency expressed in the text.

Output columns:

• Category: one of billing, technical, or account, based on the ticket's subject and body.
• Priority: one of high, medium, or low, reflecting the urgency of the ticket.

When the agent calls the tool, Batch Transform returns the original CSV with a Category and a Priority column added for every row. The agent then uses that enriched file, available at $vars.batchTransform1.output, in the rest of its task.

Zugehörige Seiten

• Mitarbeiter
• Dokumentextrahierung
• Fehlerbearbeitung