Crate rag 

Re-exports

pub use aggregation::count_by;

pub use aggregation::group_by;

pub use aggregation::sum_by;

pub use chunker::FixedSizeChunker;

pub use chunker::ParagraphChunker;

pub use chunker::SentenceChunker;

pub use chunker::TextChunker;

pub use chunker::RecursiveChunker;

pub use chunker::SemanticChunker;

pub use chunker::StructuralChunker;

pub use contextual::ContextualRetrieval;

pub use dedup::content_jaccard;

pub use dedup::dedup_similarities;

pub use diversify::diversify;

pub use embeddings::EmbeddingModel;

pub use embeddings::EmbeddingRequest;

pub use embeddings::HttpEmbeddingModel;

pub use embeddings::OllamaEmbeddingModel;

pub use embeddings::OpenAIEmbeddingModel;

pub use errors::RagError;

pub use errors::Result;

pub use eval::EvalReport;

pub use eval::average_precision;

pub use eval::dcg_at_k;

pub use eval::evaluate;

pub use eval::graded_relevance;

pub use eval::idcg_at_k;

pub use eval::ndcg_at_k;

pub use eval::precision_at_k;

pub use eval::recall_at_k;

pub use eval::reciprocal_rank;

pub use eval::relevance_set;

pub use fuzzy::fuzzy_filter;

pub use fuzzy::is_fuzzy_match;

pub use fuzzy::levenshtein;

pub use graph::Community;

pub use graph::GraphEdge;

pub use graph::GraphNode;

pub use graph::GraphPath;

pub use graph::GraphPersisted;

pub use graph::GraphStore;

pub use graph_rag::EntityExtractor;

pub use graph_rag::EntityInfo;

pub use graph_rag::ExtractedEntity;

pub use graph_rag::GraphInfo;

pub use graph_rag::GraphRagEngine;

pub use graph_rag::GraphRagResult;

pub use graph_rag::GraphRagSnapshot;

pub use graph_rag::SeedEntityExtractor;

pub use graph_rag::SimpleEntityExtractor;

pub use hybrid::merge_hybrid;

pub use hybrid::rrf_fusion;

pub use index::DistanceMetric;

pub use index::FlatIndex;

pub use index::Index;

pub use index_hnsw::HnswIndex;

pub use index_ivf::IvfflatIndex;

pub use ingestion::CodebaseSource;

pub use ingestion::PdfSource;

pub use ingestion::WikiSource;

pub use ingestion::ExtractedDocument;

pub use ingestion::Source;

pub use keyword::Bm25Config;

pub use keyword::Bm25Index;

pub use keyword::FieldBm25Index;

pub use keyword::tokenize;

pub use mcp::RagMcpServer;

pub use quantize::QuantizationParams;

pub use quantize::QuantizedIndex;

pub use query_rewriting::QueryRewriter;

pub use rerank::PassthroughReranker;

pub use rerank::SimilarityReranker;

pub use rerank::rerank_similarities;

pub use rerank_external::CohereReranker;

pub use rerank_external::MixedBreadReranker;

pub use rerank_external::VoyageReranker;

pub use retriever::Retriever;

pub use sparse::SparseIndex;

pub use sparse::SparseVector;

pub use vector_store::Document;

pub use vector_store::InMemoryVectorStore;

pub use vector_store::JsonPersistentVectorStore;

pub use vector_store::MetadataFilter;

pub use vector_store::MinimalVectorDB;

pub use vector_store::Similarity;

pub use vector_store::VectorStore;

pub use vector_store::load_all_documents;

pub use wal::WalOp;

pub use wal::WriteAheadLog;

pub use wal::apply_ops;

Modules

aggregation

chunker

contextual

Contextual retrieval: rewrite chunks with surrounding document context before embedding.

dedup

Near-duplicate suppression using word-set Jaccard similarity on chunk text.

diversify

embeddings

errors

eval

Retrieval-quality evaluation metrics: Recall@k, Precision@k, MRR, MAP, NDCG.

fuzzy

Levenshtein edit distance for fuzzy string matching.

graph

graph_rag

hybrid

Combine vector similarity hits with BM25 keyword scores.

id

Minimal unique identifier generator with no external dependencies.

index

index_hnsw

index_ivf

Inverted-file (IVF) index: cluster vectors by nearest centroid, probe top clusters only.

ingestion

keyword

Okapi BM25-style lexical scoring over in-memory documents.

mcp

quantize

Scalar Int8 vector quantization for ~4x memory reduction vs f32.

query_rewriting

LLM-based query rewriting and multi-query generation.

rerank

Optional re-ranking hook for similarity lists (default: identity).

rerank_external

External reranker implementations via HTTP APIs.

retriever

simd

SIMD-accelerated distance / similarity kernels.

sparse

Sparse vector representation and search (SPLADE-style lexical-semantic signals).

vector_store

wal

Append-only write-ahead log for incremental VectorStore updates.