LangChain4j Easy RAG

Zero-configuration RAG package for Java applications. Bundles document parsing (Apache Tika), embedding (BGE-small-en-v1.5), and text splitting with sensible defaults.

Installation

<dependency>
    <groupId>dev.langchain4j</groupId>
    <artifactId>langchain4j-easy-rag</artifactId>
    <version>1.11.0-beta19</version>
</dependency>

Quick Start

// Load document, create store, ingest with zero configuration
Document doc = FileSystemDocumentLoader.loadDocument(Paths.get("document.pdf"));
EmbeddingStore<TextSegment> store = new InMemoryEmbeddingStore<>();
EmbeddingStoreIngestor.ingest(doc, store);

// Create RAG-enabled assistant
Assistant assistant = AiServices.builder(Assistant.class)
    .chatModel(chatModel)
    .contentRetriever(EmbeddingStoreContentRetriever.from(store))
    .build();

Documentation

• Quick Start - Working examples to get started immediately
• Document Ingestion API - EmbeddingStoreIngestor API for document ingestion
• Content Retrieval API - EmbeddingStoreContentRetriever API for RAG queries
• Document Loading API - FileSystemDocumentLoader and document parsing
• Core Types - Document, TextSegment, Embedding, Metadata types
• Storage Types - EmbeddingStore and InMemoryEmbeddingStore APIs
• Chat Types - ChatModel and related types
• Architecture - How zero-configuration works via SPI
• Configuration - Default settings and customization options
• Examples - Complete working applications
• Troubleshooting - Common issues and solutions
• Reference - Dependencies, limitations, and links

What This Package Provides

• RecursiveDocumentSplitterFactory - 300 tokens/chunk, 30 token overlap
• Bundled Apache Tika - 200+ document formats (PDF, DOCX, etc.)
• Bundled BGE-small-en-v1.5 - In-process embedding model, 384 dimensions
• Automatic SPI discovery - Works with zero configuration