Search(RAG) Node#
The Search (RAG) Node retrieves relevant information from trained knowledge or uploaded documents. It extracts contextually related content based on a given query.
Search Query#
Specify the query text used to explore the knowledge base.
The system searches for semantically similar content based on this query.
Number of Knowledge Chunks#
From all retrieved results, the system sorts chunks by relevance score and returns only the top N items you specify.
Increasing this value allows the model to reference a broader context, but may reduce precision.
Search Scope#
Defines the scope of the documents to be searched.
All documents trained within the agent (default)
Specific folders only — useful when filtering by domain, product, or department
Knowledge Chunk Recipe#
Defines how documents are segmented into chunks for retrieval.
Adjusting chunk size affects the balance between context length and precision:Smaller chunks → More focused but fragmented retrieval
Larger chunks → Broader context but higher risk of noise
Additional Parameters#
Allows manual configuration of advanced search parameters in JSON format.
Adjusting similarity thresholds
Controlling ranking weight
Filtering by metadata such as document type, author, or date
Modified at 2025-10-20 05:48:26
