Skip to main contentCrate synaptic_retrieval
Source - BM25Retriever
- BM25 (Best Matching 25) retriever using Okapi BM25 scoring.
- ContextualCompressionRetriever
- A retriever that retrieves documents from a base retriever and then
compresses/filters them using a DocumentCompressor.
- Document
- A document with content and metadata, used throughout the retrieval pipeline.
- EmbeddingsFilter
- Filters documents based on cosine similarity between the query embedding
and document content embeddings.
- EnsembleRetriever
- A retriever that combines results from multiple retrievers using
Reciprocal Rank Fusion (RRF) with configurable weights.
- InMemoryRetriever
- A simple retriever that stores documents in memory and returns all of them for any query.
- MetadataFieldInfo
- Describes a metadata field for the LLM to understand available filters.
- MultiQueryRetriever
- A retriever that generates multiple query variants using a ChatModel,
runs each through a base retriever, and deduplicates results by document id.
- ParentDocumentRetriever
- Splits parent documents into children, stores both, and returns parent
documents when child chunks match a query.
- SelfQueryRetriever
- Uses a ChatModel to parse a user query into a structured query + metadata filters,
then applies those filters to results from a base retriever.
- DocumentCompressor
- Trait for compressing or filtering a set of documents based on a query.
- Retriever
- Trait for retrieving relevant documents given a query string.