rag-vector
A module that provides vector-based document storage and retrieval capabilities for Retrieval-Augmented Generation (RAG) systems.
Overview
The rag-vector module extends the rag-base module by implementing document storage with vector embeddings. It enables semantic search and similarity-based document retrieval by converting documents into vector representations. Key components include:
The
VectorStorageBackendinterface for low-level storage of documents with their pre-computed vector embeddingsThe
VectorStorageinterface that combinesWriteStorage,SearchStorage,DeletionStorage, andLookupStorageinto a single user-facing abstractionThe
DocumentEmbedderinterface for converting documents into vector representationsThe
TextDocumentEmbedderimplementation that works with text documentsThe
EmbeddingStorageclass that implementsVectorStorageby composing aDocumentEmbedderwith aVectorStorageBackend
This module bridges the gap between raw document storage and semantic search capabilities by leveraging vector embeddings to represent document content. It allows for efficient retrieval of documents based on semantic similarity to queries rather than just keyword matching.
For usage examples, see the ../../docs/docs/retrieval-augmented-generation.md guide.
A module that provides vector-based document storage and retrieval capabilities for Retrieval-Augmented Generation (RAG) systems.
Overview
The rag-vector module extends the rag-base module by implementing document storage with vector embeddings. It enables semantic search and similarity-based document retrieval by converting documents into vector representations. Key components include:
The
VectorStorageBackendinterface for low-level storage of documents with their pre-computed vector embeddingsThe
VectorStorageinterface that combinesWriteStorage,SearchStorage,DeletionStorage, andLookupStorageinto a single user-facing abstractionThe
DocumentEmbedderinterface for converting documents into vector representationsThe
TextDocumentEmbedderimplementation that works with text documentsThe
EmbeddingStorageclass that implementsVectorStorageby composing aDocumentEmbedderwith aVectorStorageBackend
This module bridges the gap between raw document storage and semantic search capabilities by leveraging vector embeddings to represent document content. It allows for efficient retrieval of documents based on semantic similarity to queries rather than just keyword matching.
For usage examples, see the ../../docs/docs/retrieval-augmented-generation.md guide.