Retriever
Represent dense similarity-based retrieval for a Basic RAG design.
Category: Query. Registry label: Retriever. Registry subtitle: Dense similarity search.
Purpose #
A retriever selects candidate chunks that appear relevant to a question. In a Basic RAG design, retrieval provides context that can be passed to an LLM.
Position in the graph #
Vector Store → Retriever → LLM
Design guidance #
Use Retriever for a standard dense-retrieval workflow. Ensure it receives the intended vector-store stage and connects toward answer generation. Use the node's current configuration dialog for exact controls, not a generic provider guide.
Current limitation #
The graph does not conduct live similarity searches in the frontend preview.