Invidious

How to Build a Basic RAG Application on AWS?

RAG is an AI pattern that combines a search step with text generation. It retrieves relevant information from a knowledge source (like a vector database) and then uses an LLM to generate accurate, context-aware responses.

Ingestion Stage

1. All raw documents (PDFs, text, etc) are first stored in Amazon S3.
2. When a file is added, AWS Lambda runs an ingestion function. This function cleans and splits the document into smaller chunks.
3. Each chunk is sent to Amazon Bedrock’s Titan embeddings model, which converts it into vector representations
4. These embeddings, along with metadata, are stored in a vector database such as OpenSearch serverless, DynamoDB

Querying Stage:

1. A user sends a question through the app frontend, which goes to API Gateway and then a Lambda query function.
2. The question is converted to an embedding using Amazon Bedrock Titan Embeddings.
3. This embedding is compared against the stored document embeddings in the vector database to find the most relevant chunks.
4. The relevant chunks and the user’s questions are sent to an LLM (like Claude or OpenAI on Bedrock) to generate an answer.
5. The generated response is sent back to the user through the same API.

Over to you: Which other AWS service will you use to build an RAG app on AWS?

--
We just launched the all-in-one tech interview prep platform, covering coding, system design, OOD, and machine learning.

Launch sale: 50% off. Check it out: bit.ly/bbg-yt

#systemdesign #coding #interviewtips
.