Hackforge Academy

Explanation

What is RAG (Retrieval-Augmented Generation)? RAG enhances LLM responses by retrieving relevant information from external data sources before generating an answer.

Code:

// User Question -> Retrieve Context -> 
LLM -> Answer

Explanation

Step 1: Load documents and create embeddings. Documents are split into chunks and converted into vector representations.

Code:

String text = Files.
readString(Paths.get("document.txt"));
Embedding embedding = 
embeddingModel.embed(text);

Explanation

Step 2: Store embeddings in a vector database such as Qdrant, ChromaDB, or Milvus for efficient similarity search.

Code:

vectorStore.add(List.of(new Document(text)));

Explanation

Step 3: Retrieve relevant document chunks based on the user's question and build a prompt with the retrieved context.

Code:

List<Document> docs = .
vectorStore.similaritySearch(question);
String prompt = "Context:\n" +
context + "\nQuestion:\n" + question;

Explanation

Step 4: Generate the final answer using an LLM. Common Java stack: Spring Boot + Spring AI + Ollama + Qdrant.

Code:

String answer = chatModel.call(prompt);