RAG Pipeline

The traitclaw-rag crate provides a modular RAG pipeline with pluggable retrievers and chunkers.

Quick Start

use traitclaw_rag::{RagContextManager, KeywordRetriever, FixedSizeChunker};

// 1. Chunk your documents
let chunker = FixedSizeChunker::new(512);
let chunks = chunker.chunk(&document_text);

// 2. Create a retriever
let retriever = KeywordRetriever::new(chunks);

// 3. Use as a ContextManager
let rag = RagContextManager::new(retriever);

let agent = Agent::builder()
    .provider(provider)
    .context_manager(rag)
    .build()?;

Retrievers

Retriever	Method	Best For
KeywordRetriever	BM25	Fast keyword matching
EmbeddingRetriever	Cosine similarity	Semantic search
HybridRetriever	BM25 + Embeddings	Best of both

Chunkers

Chunker	Strategy
FixedSizeChunker	Fixed token/char count
SentenceChunker	Split on sentence boundaries
RecursiveChunker	Hierarchical splitting