What is the initial step in the RAG pipeline when processing a source document?	Text chunking, which breaks the source document into separate pieces of text.
What is the purpose of generating text embeddings for chunks of text?	To convert the text into numerical vectors that capture the semantic meaning of the content.
What does the normalization step do to the generated embeddings?	It scales the magnitude (length) of each vector to 1.0.
What is a vector database optimized for?	Storing, comparing, and looking up long lists of numbers (embeddings).
How is a user query prepared before it is searched against the vector database?	It is run through the exact same embedding model used for the source chunks.
What mathematical calculation is used to determine which stored embeddings are most similar to the user's query?	Cosine similarity, which measures the cosine of the angle between the vectors.
What does a high cosine similarity score (close to 1) indicate?	That the user's query and the stored embedding are very similar in nature.
How is cosine distance calculated, and what does a value close to zero signify?	It is calculated as one minus the cosine similarity; a value close to zero means high similarity.
What is the final step in the RAG pipeline after a relevant text chunk is retrieved from the vector database?	The retrieved chunk and the user's query are combined into a prompt and sent to the LLM.