NVIDIA Rerank
NVIDIA Rerank takes another retriever’s results and reorders them by how well they answer the question, with an NVIDIA model specialised in ordering results. That way the most useful ones reach the agent first.
A retriever receives a question and returns the document pieces that best answer it, so an agent or model can reply with your information.
When to use it
- The retriever finds the right documents, but not always at the top.
- You already use NVIDIA’s service for other models.
Configuration
| Field | What it does |
|---|---|
| Search Query | The question. |
| Base URL | The NVIDIA service address. Default: the official one. |
| Model | The model that orders results. Default and only option: nv-rerank-qa-mistral-4b:1. |
| API Key | Your NVIDIA key. |
| Retriever | The retriever whose results are reordered. |
Output
- Retriever: the search engine, to connect to components that ask for one, such as another in this family or a question-answering component.
- Search Results: the pieces found for Search Query, which you can pass to a prompt or an agent.
Example
A job portal wants the offers best matching each search to come first.
- Connect the Retriever output of its offers database to Retriever.
- Type your key in API Key and connect the candidate’s search to Search Query.
Related components
- Contextual Compression Retriever: refines results with a model.
- Cohere Rerank: reorders results with Cohere.