Problem

The composition of system-messages for translation/trainings prompts is not flexible enough

Solution

Currently the structure of the translation/trainings prompts sent to GPT/LLM is as follows:

{Internal system prefix} {User defined Instructions} {Internal translation Instruction}

 {Internal system prefix}: You are a machine-translation engine ...

{Internal translation Instruction}: Please translate the following texts from english "en-GB" to ...

This shall be made configurable:

a) The  {Internal system prefix} can be deactivated

b) the positions of {User defined Instructions} {Internal translation Instruction} can be switched to

{Internal translation Instruction} {User defined Instructions}