Vector Store Retrievers

Store-specific retriever nodes that fetch documents by embedding similarity, with node types and connections.

Each supported vector store has a dedicated retriever node. All take a query embedding (list[float]) produced by a text embedder and return matching documents (list[Document]). Elasticsearch and OpenSearch variants accept query_embedding plus optional filters. For the store-agnostic alternatives, see Vector Store Search and Knowledge Base Search.

Store	Node label	Node type	Connection type	Inputs	Outputs	Notes
Weaviate	Weaviate Retriever	`dynamiq.nodes.retrievers.WeaviateDocumentRetriever`	`Weaviate`	`embedding`	`documents`	—
Pinecone	Pinecone Retriever	`dynamiq.nodes.retrievers.PineconeDocumentRetriever`	`Pinecone`	`embedding`	`documents`	—
Milvus	Milvus Retriever	`dynamiq.nodes.retrievers.MilvusDocumentRetriever`	`Milvus`	`embedding`	`documents`	—
pgvector	pgvector Retriever	`dynamiq.nodes.retrievers.PGVectorDocumentRetriever`	`PostgreSQL`	`embedding`	`documents`	—
Elasticsearch	Elasticsearch Retriever	`dynamiq.nodes.retrievers.ElasticsearchDocumentRetriever`	`Elasticsearch`	`query_embedding`, `filters` (optional)	`documents`	—
OpenSearch	OpenSearch Retriever	`dynamiq.nodes.retrievers.OpenSearchDocumentRetriever`	`AWSOpenSearch`	`query_embedding`, `filters` (optional)	`documents`	—
Chroma	Chroma Retriever	`dynamiq.nodes.retrievers.ChromaDocumentRetriever`	`Chroma`	`embedding`	`documents`	—
Qdrant	Qdrant Retriever	`dynamiq.nodes.retrievers.QdrantDocumentRetriever`	`Qdrant`	`embedding`	`documents`	—

For when to use these nodes instead of a managed Knowledge Base, see the Vector Store vs Knowledge Base guide; for end-to-end indexing and retrieval pipelines, see How nodes connect and the Knowledge Bases guides.