Creating and Testing Actions in Requestador

Actions are the core building blocks of Requestador. An action defines what task should be executed, which endpoint should process it, which inputs are required, and how the response is validated and transformed. 

An action contains:

• a name

• a selected AI endpoint

• a system prompt

• an action prompt

• input parameters

• an optional validation script

• an optional transform script

• optional catalogs

In Requestador, navigate to Actions and click Add Action.

Action Name - choose a clear and descriptive name, for example:

• pimcore generate category

• pimcore general translate

• pimcore generate short description

Endpoint - select the AI endpoint created earlier.

Description - enter an optional description of the action.

Labels - add labels to organize and filter actions. Labels can represent language, use case, content type, or business group.

Examples:

• english

• translation

• camping-and-hobby

The System Prompt defines the general role and behavior of the AI model.

This should contain stable instructions such as:

• output structure requirements

• tone and style rules

• restrictions

• formatting expectations

Use the system prompt for rules that should always apply.

The Action Prompt defines the specific task the action should perform.

Typical examples:

• generate product descriptions

• assign product category

• translate content into a target language

• return structured JSON output

The action prompt may refer to parameters that will be passed at runtime.

Define all input parameters needed by the action.

Each parameter includes:

• Name

• Value (optional default value)

• Required flag

Examples:

• description

• technical

• language

• name

Required parameters must be provided by the calling system or configuration

The validation script checks whether the AI response is valid.

Typical validation goals:

• required fields exist

• values are within allowed options

• the returned structure matches expectations

The validation script should return a response indicating:

• whether the output is valid

• an optional error message

 For more detailed explanation please check here

The transform script converts the raw AI response into the final format expected by the calling system.

Typical uses:

• map fields

• rename keys

• normalize values

• convert raw text into structured output

 For more detailed explanation please check here

Catalogs can be used as controlled reference data for classification or structured outputs.

Use catalogs when the action must select from predefined values rather than generate completely free-form results.

Click Save.

The action is now available for use in Pimcore configuration.

Do not continue to Pimcore configuration until the action returns the expected output consistently in Requestador. Testing the action first helps avoid unnecessary troubleshooting later in the Pimcore mapping stage.

This step is important to confirm that:

• the prompt logic works as expected

• all required parameters are handled correctly

• the response format is correct

• validation passes successfully

• the transform script produces the desired final output

Use representative test input values and verify that the returned result matches the structure you intend to map in Pimcore.

• use descriptive names

• keep system prompt and action prompt clearly separated

• define only necessary parameters

• validate outputs whenever structured responses are expected

• transform outputs into a Pimcore-friendly format

• test the action in Requestador before mapping it in Pimcore