RAG - Retrieval Augmented Generation

RAG Applications (built in top of LLMs) FullStack Web RAG Application Select between multiple datasources and as using functions as tools to gather information

LLamaIndex

RAG -> solution to limited context windows (retrieve most relevant data form DB => augment query context => generate response)

Vector Search / Vector Embeddings

• emebedd your data into vector space (easy to locate relevant data)

1. embed (DATA => LLM => VECTOR SPACE)
2. retrieve context (PROMPT => LLM => VECTOR SPACE)
3. query (PROMPT + Context(2) )=> LLM => response)

Query Engine a) retriever (context out of the index, get nodes) b) post processing (process the nodes the retriever fetched, before giving to LLM) c) synthetizer: combine processed nodes, prompt template and query into a single prompt to the LLM