VectorStore To Document Loader
VectorStore To Document Node v V2
The VectorStore To Document loader allows you to reverse-query a Vector Store. Instead of using documents to build a vector store, you can query an existing vector store to retrieve relevant document chunks based on semantic similarity.
Inputs
- Vector Store: The target Vector Store node you wish to retrieve documents from.
- Top K: The number of top-matching search results to return (integer).
- Minimum Score (optional): The threshold similarity score required to include a document.
- Additional Metadata (optional): JSON object to append additional metadata.
Outputs
- Document: Array of document objects containing metadata and pageContent retrieved from the Vector Store.
- Text: Concatenated string of the retrieved semantic chunks.
Features
- Direct conversion of vectors back to raw Langchain document formats.
- Filtered retrieval using similarity thresholds.
- Compatible with all Agentflow V2 supported Vector Stores (Pinecone, Weaviate, Milvus, Qdrant, PGVector, etc.).
- Preserves original metadata attached to the vector embeddings.