4 ms·
There are already ways of dynamically populating the context of the model with information retrieves as text embeddings from a vector store, based on the initia
by danielbln 4y ago
There are already ways of dynamically populating the context of the model with information retrieves as text embeddings from a vector store, based on the initial prompt. This way you can feed the model current and correct information which it will weave into its output