Is this few-shot fed in before the task starts? It is static. I want to dynamically update the prompt words based on the NER modified by the user as a few-shot.
Hi!
Before diving into this question it deserves to be highlighted that the OpenAI recipes will be deprecated in the future and the spacy-LLM variants are designed to replace them. More details on this can be found in the docs.
That said. The LLM recipes require a static configuration file, so the prompts will be static and won't get an update unless you update a local config file. This may seem as a limitation, but it can also be seen as a good practice because it keeps the prompt the same. A prompt that may change at any time depending on the order of the examples that it has seen can cause a lot of unexpected behavior which makes it hard to reason about.
Nothing is stopping you from adapting these recipes with custom code yourself though, so if you prefer to dynamically change the prompts then you can use your favorite Python tools to do this yourself via custom recipes.