Skip to article
/ docs

Search documentation

Popular starting points. Search all 37 articles.

  • Get startedProduct availability and preview limitationsUnderstand what works in the frontend and which product capabilities are not connected yet.
  • Get startedBuild your first RAG pipelineBuild and inspect your first seven-node Basic RAG graph and learn what validation means.
  • Build pipelinesBasic RAG and Hybrid Search templatesCompare Ragify's two starting graphs and decide which pattern to explore.
  • Node referenceUnderstand the node libraryBrowse the ten registered node types by role in the workflow.

↑ ↓ to navigate · Enter to open · Esc to close

Back to Ragify ↗
On this page

Documentation

Configurable frontend node; re-scoring planned

Reranker

Represent a secondary ranking stage for retrieved candidates.

Category: Query. Registry label: Reranker. Registry subtitle: Cross-encoder re-scoring.

Purpose #

A reranker re-evaluates candidate results after initial retrieval and produces an ordering better aligned with a query. A cross-encoder-based reranker is one way to score query-and-passage pairs.

Position in the graph #

Hybrid Retriever → Reranker → LLM

Design guidance #

Place the Reranker after the Hybrid Retriever in the provided Hybrid Search template. Distinguish reranking from the initial retrieval stage when explaining a graph to collaborators.

Current limitation #

No candidates are scored by a cross-encoder or other runtime model in this frontend preview.

Hybrid Retriever, LLM.