🛫Azure AI Foundry Implementation - 1

What is RAG?

• It's called Retrieval Augmented Generation.

• R - What do we RETREIVE? --> Relevant Data based on user Input/prompt.

• A - What do we AUGEMENT? --> Our Interaction with LLM's in order to get more specific output.

• G - What do we GENERATE? --> A response or output grounded in our data.

What are the RAG benefits?

• Intelligent chatbot or webapp offers quick automated troubleshooting suggestions internal documentation uses company guides to provide accurate responses reduced workload less repetitive work for the network administrator and improved IT support delivers precise personalised answers and scalability groins with the company ensuring continues support.

How Does RAG Work?

• We need mode models to make this work an embedding model for the text vectorization and a GPT Model to generate accurate responses using your Data.

• We retrieve the data based on the user's input/prompt because when you interact with LLM it only gives output of what is trained but we want to use our data for that purpose we use an embedding model to query our files in a vector database.

• We use this embedding model to augment the LLM because we want to get more specific output.

• We generate output grounded in the retrieve information.

What is an embedding model for text vectorization?

• vectorization performed by embedding model is the process by which a machine learning model transforms a data element such as text or an image or an audio file into numerical representation in the form of a vector in a multi-dimensional space and these vectors preserve the semantic relationship between data allowing for efficient comparison and retrieval of similar information.

How does this process of RAG use our own Data?

1. Upload your raw data

2. Raw data is broken into chunks smaller pieces for easier processing

3. Each chunk turned into a numerical vector equal to embedding that captures its meaning.

4. Embeddings are stored in a vector database which enables fast retrieval of relevant information-based vector DB.

How does vector databases created?

Why is it important to use your own Data?

• Becos, you get access to personalized and private data greater control over the generated content and higher accuracy and less hallucinations no retraining required.

• Flexibility works with PDFs, databases, Api's and more and real time updates which means no need to retrain the model.