For the past few weeks, I have been working on creating a RAG application.
I’m attempting to print the markdown-formatted model output from that RAG system into the front end of my RAG application.
However, I just attempted to create anything associated with that, and I neglected to include a prompt that indicates that responses should be made in markdown format.
However, I continued to receive in the form of a markdown.
I would like to know if the prompt is stored in the cache.
Hi and welcome to the Dev Community.
I don’t believe so, for normal gpt-4 API calls the model doesn’t remember anything between prompts unless you feed it the information.
It seems that these models just have a natural affinity towards responding in markdown format, likely a byproduct of the OpenAI’s fine-tuning.
If you are using a chat assistant, it is my understanding that chatGPT re-reads the ENTIRE conversation with every prompt. Which explains a noticeable response slow down in conversations that run several days. So I would assume the conversation is indeed loaded into cache when you open it back up to continue from yesterday.
Though I’m not 100% sure about this.