Trueno-RAG

Pure-Rust Retrieval-Augmented Generation (RAG) pipeline built on Trueno compute primitives.

Features

• Pure Rust - Zero Python/C++ dependencies
• SIMD-Accelerated - Built on Trueno compute primitives
• Flexible Chunking - 6 strategies: Recursive, Fixed, Sentence, Paragraph, Semantic, Structural
• Hybrid Retrieval - Dense (vector) + Sparse (BM25) search
• Multiple Fusion - RRF, Linear, DBSF, Convex, Union, Intersection
• Reranking - Lexical, cross-encoder, and composite rerankers
• Evaluation Metrics - Recall, Precision, MRR, NDCG, MAP
• Query Preprocessing - HyDE, multi-query expansion, synonyms

Quick Start

use trueno_rag::{
    pipeline::RagPipelineBuilder,
    chunk::RecursiveChunker,
    embed::MockEmbedder,
    rerank::LexicalReranker,
    fusion::FusionStrategy,
    Document,
};

// Build pipeline
let mut pipeline = RagPipelineBuilder::new()
    .chunker(RecursiveChunker::new(512, 50))
    .embedder(MockEmbedder::new(384))
    .reranker(LexicalReranker::new())
    .fusion(FusionStrategy::RRF { k: 60.0 })
    .build()?;

// Index documents
let doc = Document::new("Your content here...").with_title("Doc Title");
pipeline.index_document(&doc)?;

// Query with context
let (results, context) = pipeline.query_with_context("your query", 5)?;
println!("{}", context.format_with_citations());

Installation

[dependencies]
trueno-rag = "0.1"

Examples

# Basic RAG pipeline
cargo run --example basic_rag

# Compare chunking strategies
cargo run --example chunking_strategies

# Hybrid search with different fusion
cargo run --example hybrid_search

# Retrieval metrics evaluation
cargo run --example metrics_evaluation

Documentation

# Build and serve documentation book
make book-serve

# Or build only
make book

Chunking Strategies

Fusion Strategies

Development

# Run tests
make test

# Fast tests (release mode)
make test-fast

# Lint
make lint

# Format
make fmt

# Full CI check
make ci

# Coverage report
make coverage

Architecture

Document -> Chunker -> Embedder -> Index (Dense + Sparse)
                                        |
Query -> Preprocessor -> Retriever -> Fusion -> Reranker -> Context

License

MIT