Prompts
As already mentioned, Prompts are at the heart of communication with artificial intelligence. You can compare it quite well to raising a child. It is not enough to tell a child what it has to do, but an important aspect is to mention what it should refrain from doing in this context.
Tip
The Microsoft website Introduction to Prompt Engineering and the concept articles accessible from there provide a good introduction.
Setup
| Field | Description |
|---|---|
| Code | Used to identify and uniquely name a prompt. |
| Description | Allows a more detailed description of the defined prompt. |
| Platform Code | Here you specify the configured Platform to which the request is sent. |
| Data Entity Code | Defines which data entity is to be used to generate the data structure. This also specifies the context (and the base table) for this prompt. If the data entity code is not specified, this prompt is available for all entities. |
| Language Code | The language code is used to translate option or enumeration fields. It is not (currently) used to define the language of the response. However, the request text itself can contain a phrase to force the response to be in a specific language. You can also use a placeholder for the language. |
| System Role Description | A text is stored here that defines the role of the AI and what freedoms or restrictions it has when responding. See also Prompt Samples and Results. |
| Prompt Type | This is used to pre-classify the content to be generated and primarily serves to reduce the number of available prompts for different cases. Currently, HTML and Text are available here. |
| Prompt | The prompt text itself defines the actual task, such as writing a text in HTML format. See also Prompt Samples and Results. |
| Placeholder Data Processing | Specifies whether and how placeholder data is generated. See Data Handling Options. |
| Structured Data Processing | Specifies whether and how structured data is generated. See Data Handling Options. |
Data Handling Options
- None: No data is generated.
- Original: The non-anonymized original data is transferred to the AI.
- Anonymize: The data is anonymized before transmission. This is done by converting the contents of code and text fields into a GUID.
Important
The use of anonymous placeholder data affects the prompt displayed in the AI dialog. This takes some getting used to, but ensures that the original data is not inadvertently sent despite anonymization being activated.
Additional Values
Attached to prompts you will also find Additional Values. These allow you to define additional values for e-mails, for example. Clicking on Value opens the editor in which you can use static text and also placeholders based on the configured Data Entity.
- Subject: Defines the subject line for an e-mail.
- Recipient: Specifies the recipient of an e-mail and should only contain a static e-mail address or an e-mail placeholder.
- Carbon Copy: Used to determine an e-mail address to which a copy (CC) is sent.
- Bling Carbon Copy: Here you specify the recipient of a blind carbon copy (BCC).
Prompt Parameters
The Prompt Parameters are intended for future extensions and are not currently used.