Module rag

Available on crate feature rag only.

Expand description

Retrieval-Augmented Generation (RAG) pipeline.

Modular RAG system with trait-based components:

Available with feature: rag

Modules§

chunking: Document chunking strategies.
config: Configuration for the RAG pipeline.
document: Data types for documents, chunks, and search results.
embedding: Embedding provider trait for generating vector embeddings from text.
error: Error types for the adk-rag crate.
inmemory: In-memory vector store using cosine similarity.
pipeline: RAG pipeline orchestrator.
reranker: Reranker trait for re-scoring search results.
tool: Agentic retrieval tool for ADK agents.
vectorstore: Vector store trait for storing and searching vector embeddings.

Chunk: A segment of a Document with its vector embedding.
Document: A source document containing text content and metadata.
FixedSizeChunker: Splits text into fixed-size chunks by character count with configurable overlap.
InMemoryVectorStore: An in-memory vector store using cosine similarity for search.
MarkdownChunker: Splits text by markdown headers, keeping each section as a chunk.
NoOpReranker: A no-op reranker that returns results unchanged.
RagConfig: Configuration parameters for the RAG pipeline.
RagConfigBuilder: Builder for constructing a validated RagConfig.
RagPipeline: The RAG pipeline orchestrator.
RagPipelineBuilder: Builder for constructing a RagPipeline.
RagTool: A retrieval tool that wraps a RagPipeline for agentic use.
RecursiveChunker: Splits text hierarchically: paragraphs → sentences → words.
SearchResult: A retrieved Chunk paired with a relevance score.

Chunker: A strategy for splitting documents into chunks.
EmbeddingProvider: A provider that generates vector embeddings from text input.
Reranker: A reranker that re-scores and reorders search results.
VectorStore: A storage backend for vector embeddings with similarity search.