Context
I want to use an LLM model to generate specific proprietary tests that are very specific to an application in the Telecommunication industry.
What does it do?
The API calls fill parameters with specific values within an XML file - the XML file is then used to generate the test scripts that we are looking for.
Approach
Pre-train a seq-to-seq model like BART and use instruction fine-tuning to cater it to our specific use case. I am thinking of implementing an RAGs framework to build context for our niche use case.
Questions
(1) does this approach seem viable/valid for our goal?
(2) What other approach should I consider?