What is the Document class in Spring AI and how is it used in RAG?

The Document class is Spring AI's core data carrier for textual content flowing through a RAG pipeline. It wraps a piece of text together with a metadata map and an optional embedding vector, giving every chunk a consistent identity regardless of where it originated.

Key fields:

• id — auto-generated UUID uniquely identifying the chunk.
• content — the plain text of the chunk that gets embedded and later injected into prompts.
• metadata — a Map<String, Object> carrying provenance data (source filename, page number, URL, creation date). This metadata is preserved through the VectorStore and returned alongside similarity search results, so you can cite sources in answers.
• embedding — the float vector populated by the EmbeddingModel; null until the document is embedded.

// Creating a Document manually
Document doc = new Document(
    "Spring AI simplifies AI integration in Java applications.",
    Map.of("source", "spring-ai-docs.pdf", "page", 1)
);

// After similarity search you can access metadata
List<Document> results = vectorStore.similaritySearch(
    SearchRequest.query(question).withTopK(3));
results.forEach(d -> {
    System.out.println(d.getContent());
    System.out.println("Source: " + d.getMetadata().get("source"));
});

When you add documents to a VectorStore via vectorStore.add(docs), the store internally calls the EmbeddingModel to populate the embedding field before persisting. The caller does not need to embed documents separately in the typical flow.