Tag Archives: RAG

RAG

What Does RAG Do

AI

RAG combines both retrieval and generation. rAG strengthens the LMMs by grounding their response in external, up-to-date, or domain-specific data. Such updates on these models enable the RAG to support applications such as question-answering systems, chatbots, and content generation that rely highly on accuracy, relevance, and context awareness.

Retrieval—the act of searching for relevant information

generation—using an LLM to produce a response

How RAG Works

Retrieval
Augmentation
Generation
  • When a user submits a query (e.g., “What are the symptoms of diabetes?”), the system searches a knowledge base (e.g., a vector database) for relevant information.
  • The knowledge base could include documents, FAQs, research papers, or other structured or unstructured data.
  • The retrieval process is often powered by vector embeddings and similarity search, which find the most semantically relevant information to the query.
  • The retrieved information (e.g., a medical guideline or research paper) is passed to the LLM as context.
  • This context helps the LLM understand the query better and generate a more accurate response.
  • The LLM uses the retrieved context and the user’s query to generate a natural language response.
  • The response is not only based on the LLM’s pre-trained knowledge but also on the specific, up-to-date information retrieved from the knowledge base.

Key Components of RAG

  1. Retriever:
    • A system, for example, a vector database that retrieves relevant information from a knowledge base.
    • It often uses vector embeddings and similarity search.
  2. Generator:
    • An LLM (like GPT-4, DeepSeek) generates natural language responses based on the retrieved context and the user’s query.
  3. Knowledge Base:
    • A collection of documents, FAQs, or other data that the retriever searches through.
    • It can be stored in a vector database for efficient retrieval.

Usecase – Healthcare Chatbot

Let’s discuss how a healthcare chatbot using RAG will work.

  1. User Query: “What are the symptoms of diabetes?”
  2. Retrieval: The system searches a vector database of medical guidelines and retrieves the most relevant document: “Symptoms of diabetes include frequent urination, excessive thirst, and unexplained weight loss.”
  3. Augmentation: The retrieved document is passed to the LLM as context.
  4. Generation: The LLM generates a response: “Common symptoms of diabetes include frequent urination, excessive thirst, and unexplained weight loss.”
RAG

Reference: https://cloud.google.com/use-cases/retrieval-augmented-generation?hl=en

Find more at https://twirltech.in/