Why should a developer never attempt to access Vertex directly from a client-side app?	Because making requests requires secret credentials, which must be kept secure on a controlled server.
What are three required pieces of data when a server makes a request to Vertex?	The name of the model, a list of messages, and a max tokens value.
In the text generation process, what is a token?	A chunk of text, which can be a whole word, a part of a word, a space, or a symbol.
What is an embedding?	A long list of numbers that serves as a number-based definition of a given word.
What is the purpose of contextualization in the text generation process?	To adjust each embedding based on surrounding embeddings to refine its meaning to a single precise definition.
How does the model select the next word during the generation stage?	It uses a mix of probability and randomness to select words, rather than automatically picking the highest probability.
What are the two conditions that cause the model to stop generating text?	It reaches the max tokens parameter, or it generates a special end of sequence token.
What information is included in the API response after text generation?	The generated text, the usage count (input and output tokens), and the stop reason.