Scorer Retriever
Scorer Retriever searches a vector database and drops documents whose similarity to the question does not reach a minimum score. It keeps unrelated results away from the agent.
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
- You would rather the agent say it does not know than answer with loosely related documents.
- You want to control the minimum quality of the results.
Configuration
| Field | What it does |
|---|---|
| Search Query | The question. |
| Vector Store | The Vector Store output of the database to search. |
| Score Threshold | The minimum relevance score, from 0 (unrelated) to 1 (identical). Default: 0, dropping nothing. |
Advanced options
| Field | What it does |
|---|---|
| Number of Results | The maximum number of documents returned. Default: 10. |
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, each with its relevance in
score, which you can pass to a prompt or an agent.
Use as an agent tool
Connect Scorer Retriever to the Tools input of an Agent and the agent searches on its own: it writes Search Query with the question. It is called scored_search by default; in the Tool tab change the description to say what your documents contain.
Example
A health centre wants its assistant to answer only with protocols closely related to the question.
- Connect the Vector Store output of its protocols database to Vector Store.
- Run a few questions and look at the
scoreof each result: the right ones and the unrelated ones. - Set Score Threshold between the two.
Good practices
- Look at the scores before choosing the threshold: they depend on the embeddings model and on the database. In our test on 17 October 2026 (9 documents, 40 questions, a synthetic set), the questions with no answer in the documents scored as high as many that did have one, so a threshold could not separate them. To say “no answer”, use Minimum Relevance in Local Rerank.
Related components
- Chroma DB: a vector database these retrievers can start from.