Text Generation with Prompts via OpenAI

This interaction type can be executed on an object context (events) and lets you create texts for data objects and refine the result by interacting with the OpenAI ChatGPT model.

To refine the result, enter a prompt in the text field below the result and click the enter button. The result will be refined based on your prompt.

Example prompts:

• Translate to Klingon
• Shorten the text to less than 70 words
• Add a short text about Naboo

Available Context

For more information on context limitations refer to the Available Context section.

Multiselect: No Elements: Data Objects

Configuration Options

Type to be used: 'Pimcore\Bundle\CopilotBundle\Samples\InteractionAction\OpenAiTextPrompt'

Type specific configuration options:

# Template for initial prompt automatically sent to ChatGPT when action is started.
template: |
    We have an event with title {{ subject.title }} that starts at {{ subject.fromDate }} and ends at {{ subject.toDate }}.
    {% if subject.description %}
       Optimize following description and include all information: {{ subject.description | raw }}
    {% else %}
       Generate a short description for that event.
    {% endif %}
  
# AI model to be used 
model: 'gpt-4o'

# Target data object field. 
target_field: 'description'

# Defines, if target field is localized. 
target_field_localized: true

# Defines the system prompt
system_prompt: 'All events should include a Star Wars reference.'

Detailed Configuration Options

• template: Template for initial prompt automatically sent to ChatGPT when action is started. It is a twig template with access to the subject on which action is started, e.g. data object. Thus, data from data object can be used to generate initial prompt and deliver additional information.
• model: AI model to be used
• target_field: Target data object field, where generated text should be assigned to.
• target_field_localized: Defines, if target field is localized. If set to true, then additional language selection is shown in Pimcore Copilot when assigning result to target field.
• system_prompt (optional): System prompt to be used for the AI model. This prompt is used in addition to the template prompt and acts as a general guideline for the AI model. It will be considered for alle requests sent to the model.

info

This action requires a OpenAI token, which needs to be provided as via OPEN_AI_AUTH_TOKEN environment variable:

OPEN_AI_AUTH_TOKEN=<YOUR_API_TOKEN>

Full Configuration Example

pimcore_copilot:
    interaction_actions:
        actions:
            event_description_generation:
                context_limitations: ['object_event']
                search_terms: []
                type: 'Pimcore\Bundle\CopilotBundle\Samples\InteractionAction\OpenAiTextPrompt'
                configuration:
                    template: |
                        We have an event with title {{ subject.title }} that starts at {{ subject.fromDate }} and ends at {{ subject.toDate }}.
                        {%% if subject.description %%}
                           Optimize following description and include all information: {{ subject.description | raw }}
                        {%% else %%}
                           Generate a short description for that event.
                        {%% endif %%}
                    model: 'gpt-4o'
                    target_field: 'description'
                    target_field_localized: true
                    system_prompt: 'All events should include a Star Wars reference.'
                icon: null