Class RagExecutor

java.lang.Object

ai.nervemind.app.executor.RagExecutor

All Implemented Interfaces:

NodeExecutor

@Component
public class RagExecutor
extends Object
implements NodeExecutor

Executor for the "rag" node type - implements Retrieval-Augmented Generation.

Combines document retrieval with Large Language Models to answer questions based on specific context. This node orchestrates the full RAG pipeline: embedding generation for the query, vector similarity search against documents, and LLM generation with retrieved context.

Node Parameters

RAG node configuration parameters

Parameter	Type	Default	Description
provider	String	"openai"	LLM provider ("openai", "anthropic", "ollama")
chatModel	String	auto	Model for answer generation (e.g., "gpt-4o")
embeddingModel	String	auto	Model for embedding generation
apiKey	String	from settings	API key for the provider
baseUrl	String	provider default	Custom base URL for API
query	String	-	User's question/query (supports interpolation)
documents	List	[]	Documents to search (content + metadata)
topK	Integer	3	Number of documents to retrieve
includeContext	Boolean	true	Include retrieved text in output

Pipeline Steps

1. Embedding: Generates vector for the user query
2. Indexing: Generates embeddings for documents (if missing)
3. Retrieval: Finds top K documents using cosine similarity
4. Generation: Sends query + retrieved context to LLM

Output Data

Output keys added by this executor

Key	Type	Description
response	String	The LLM's generated answer
context	List	Retrieved documents (if includeContext=true)
sources	List	Metadata from retrieved documents
documentsRetrieved	Integer	Number of documents found

See Also:

• For standalone embedding generation
• For direct chat without retrieval

Constructor Summary

Constructors

Constructor	Description
RagExecutor()	Creates a new RAG executor with default HTTP client configuration.

Method Summary

Modifier and Type	Method	Description
Map<String,Object>	execute(ai.nervemind.common.domain.Node node, Map<String,Object> input, ExecutionService.ExecutionContext context)	Executes the business logic for this node type.
String	getNodeType()	Unique identifier for the node type this executor handles.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

RagExecutor

public RagExecutor()

Creates a new RAG executor with default HTTP client configuration.

Method Details

execute

public Map<String,Object> execute(ai.nervemind.common.domain.Node node,
 Map<String,Object> input,
 ExecutionService.ExecutionContext context)

Description copied from interface: NodeExecutor

Executes the business logic for this node type.

Specified by:

execute in interface NodeExecutor

Parameters:

node - The node definition containing parameters and configuration. Use Node.parameters() to access user settings.

input - Combined output from all upstream nodes that connected to this node. For simple flows, this contains the direct predecessor's output. For merge nodes, it contains combined data.

context - Verification context providing access to workflow-scoped services, logger, and execution metadata.

Returns:

A Map containing the results of this node's execution. Keys in this map become available variables for downstream nodes.
Note: Returning null is treated as an empty map.

getNodeType

public String getNodeType()

Description copied from interface: NodeExecutor

Unique identifier for the node type this executor handles. This must match the 'type' field in the JSON definition of the node.

Specified by:

getNodeType in interface NodeExecutor

Returns:

The unique type string (e.g., "httpRequest", "llmChat").