Retriever Modules
We are actively adding more tailored retrieval guides. In the meanwhile, please take a look at the API References.
Index Retrievers
Section titled âIndex RetrieversâPlease see the retriever modes for more details on how to get a retriever from any given index.
If you want to import the corresponding retrievers directly, please check out our API reference.
Comprehensive Retriever Guides
Section titled âComprehensive Retriever GuidesâCheck out our comprehensive guides on various retriever modules, many of which cover advanced concepts (auto-retrieval, routing, ensembling, and more).
Advanced Retrieval and Search
Section titled âAdvanced Retrieval and SearchâThese guides contain advanced retrieval techniques. Some are common like keyword/hybrid search, reranking, and more. Some are specific to LLM + RAG workflows, like small-to-big and auto-merging retrieval.
- Define Custom Retriever
- BM25 Hybrid Retriever
- Simple Query Fusion
- Reciprocal Rerank Fusion
- Auto Merging Retriever
- Metadata Replacement
- Composable Retrievers
Auto-Retrieval
Section titled âAuto-RetrievalâThese retrieval techniques perform semi-structured queries, combining semantic search with structured filtering.
- Auto-Retrieval (with Pinecone)
- Auto-Retrieval (with Lantern)
- Auto-Retrieval (with Chroma)
- Auto-Retrieval (with BagelDB)
- Auto-Retrieval (with Vectara)
- Multi-Doc Auto-Retrieval
Knowledge Graph Retrievers
Section titled âKnowledge Graph RetrieversâComposed Retrievers
Section titled âComposed RetrieversâThese are retrieval techniques that are composed on top of other retrieval techniques - providing higher-level capabilities like hierarchical retrieval and query decomposition.
- Query Fusion
- Recursive Table Retrieval
- Recursive Node Retrieval
- Braintrust
- Router Retriever
- Ensemble Retriever
- Multi-Doc Auto-Retrieval
Managed Retrievers
Section titled âManaged RetrieversâOther Retrievers
Section titled âOther RetrieversâThese are guides that donât fit neatly into a category but should be highlighted regardless.