RAG Datasource Node

Query your knowledge bases with semantic search.

Overview

The Datasource node allows you to query your uploaded documents using natural language. It supports multiple query modes including semantic search, structured queries, and hybrid approaches.

Configuration

Field	Description	Required
`Data Source`	Select one or more datasources to query	Yes
`Query`	The search query (supports variables)	Yes
`Output Variable`	Variable name to store the results	Yes

Query Modes

Semantic Search

Natural language queries using vector embeddings. Best for finding conceptually similar content.

Structured Query (Tables)

For table-type datasources, build queries with filters, sorting, and column selection.

Hybrid

Combines keyword search with semantic search for better precision.

Advanced Options (Semantic)

Field	Description	Default
`Search Method`	Semantic Search, Diverse Results (MMR), or Hybrid	Semantic
`Number of Results`	How many matching chunks to retrieve (1-20)	5
`Relevance Threshold`	Minimum match quality (0 = any, 1 = exact)	0.7

Search Methods

Method	Description
`Semantic Search`	Pure vector similarity search
`Diverse Results (MMR)`	Maximal Marginal Relevance - reduces redundancy
`Hybrid`	Combines keyword and semantic search

Structured Query Options (Tables)

For table datasources, you can build SQL-like queries:

Filters - Add conditions with AND/OR logic
Sort - Order results by column (ASC/DESC)
Columns - Select specific columns to return
Limit/Offset - Pagination controls

Filter Operators

Operator	Description
`eq`	Equals
`neq`	Not equals
`gt` / `gte`	Greater than / Greater or equal
`lt` / `lte`	Less than / Less or equal
`contains`	String contains
`starts_with`	String starts with
`ends_with`	String ends with
`is_empty`	Value is empty
`is_not_empty`	Value is not empty

Using Variables

{{user_question}}
{{search_term}}
{{prompt}}