Skip to main content
The Agentic RAG node uses an intelligent agent-based approach to perform retrieval-augmented generation. It autonomously decides how to retrieve and process information based on the query context, providing adaptive and context-aware responses. Agentic RAG node overview

​
Overview

The Agentic RAG node provides intelligent retrieval by:
  1. Autonomous retrieval — Agent decides the best way to find relevant information
  2. Context-aware processing — Adapts retrieval strategy based on query context
  3. Source grounding — Returns references to source documents
  4. Zero configuration — Works out of the box with no setup required
Agentic RAG is the simplest RAG node to set up — it requires no configuration and handles document processing automatically.

​
When to Use Agentic RAG

ScenarioRecommendation
Simplest setup✅ Recommended — No configuration needed
Adaptive retrieval✅ Recommended — Agent optimizes retrieval strategy
Quick prototyping✅ Recommended — Get started immediately
Custom chunking control❌ Use Smart RAG or Chunking + Retrieval
Fine-tuned embedding models❌ Use Chunking + Retrieval

​
Using the Agentic RAG Node

​
Adding the Agentic RAG Node

  1. Open your flow in the Flow Builder
  2. Drag the Agentic RAG node from the sidebar onto the canvas
  3. Connect your Dataset node to Agentic RAG
  4. Connect a Question or Testset node for queries
  5. Double-click the Agentic RAG node to view settings

​
Input Connections

The Agentic RAG node accepts input from:
Source NodePurpose
DatasetRequired — Provides documents to process
QuestionProvides queries for retrieval (simulation/testing)
TestsetProvides multiple queries for comprehensive testing
To simulate Agentic RAG, you must connect a Question or Testset node as input. Without queries, the node cannot demonstrate retrieval results.

​
Output Connections

The Agentic RAG node can connect to:
Target NodeUse Case
RerankingFurther improve result quality with LLM-based scoring
LLMGenerate natural language responses
AnalysisEvaluate retrieval performance
ResponseOutput retrieved results directly

​
Configuration

The Agentic RAG node has no configurable parameters — it works out of the box with optimized default settings, making it the simplest RAG node to set up. Agentic RAG configuration

​
How Agentic RAG Works

​
Processing Flow

  1. Document Processing — Documents from Dataset are automatically processed
  2. Query Analysis — The agent analyzes the user query
  3. Autonomous Retrieval — Agent decides the best retrieval strategy
  4. Source Grounding — Results include source references
  5. Response — Retrieved chunks are returned with metadata

​
Key Differences from Other RAG Nodes

AspectAgentic RAGSmart RAGChunking + Retrieval
ChunkingAutomaticLocal (title-based)Local (configurable)
ConfigurationNoneLimitedFull control
Setup timeFastestFastLonger
Embedding modelAutomatictext-embedding-3-smallMultiple options

​
Pipeline Examples

​
Simple Agentic RAG Pipeline

Dataset → Agentic RAG → LLM → Response ← Question Best for: Quickest setup with intelligent retrieval.

​
Agentic RAG with Reranking

Dataset → Agentic RAG → Reranking → LLM → Response ← Question Best for: Enhanced precision through additional LLM-based scoring.

​
Agentic RAG Evaluation Pipeline

Dataset → Agentic RAG → Analysis ← Testset Best for: Evaluating retrieval quality across multiple questions.

​
Viewing Results

After running the pipeline (click Update Results):
  1. Results show documents grouped by question
  2. Each result displays:
    • Question — The query being answered
    • Content — Retrieved text
    • Score — Relevance score
    • File name — Source document

​
JSON View

Toggle JSON to see the raw result structure: 
{
  "page_content": "Retrieved chunk text...",
  "metadata": {
    "question": "What is machine learning?",
    "question_id": "abc-123",
    "expected_answer": "...",
    "score": 0.92,
    "file_name": "ml-guide.pdf"
  }
}

​
Agentic RAG vs. Smart RAG

FeatureAgentic RAGSmart RAG
ComplexitySimplestSimple
ConfigurationNoneTop K, metadata options
Document processingAutomaticTitle-based chunking
Embedding controlNoNo (fixed model)
Best forQuick start, zero configMore control over retrieval
Choose Agentic RAG when:
  • You want the simplest possible setup
  • You don’t need fine-grained control
  • You want agent-based adaptive retrieval
Choose Smart RAG when:
  • You need metadata/annotation options
  • You want Top K control
  • You need title-based chunking

​
Best Practices

  1. Ensure documents are parsed — Agentic RAG uses the active parsing version from your sources
  2. Use clear questions — Well-formed queries help the agent retrieve better results
  3. Test with Testsets — Evaluate performance across diverse questions before deployment
  4. Consider Reranking — Add a Reranking node if precision is critical
  5. Check source files — Verify retrieved content comes from expected documents

​
Troubleshooting

If Agentic RAG returns no results:
  • Verify a Question or Testset node is connected
  • Check that Dataset contains processed documents
  • Ensure documents have status “Processed”
If results aren’t relevant:
  • Improve document parsing quality
  • Rephrase questions to be more specific
  • Consider using Smart RAG for more control
  • Add Reranking node for better scoring
If Agentic RAG is slow:
  • Reduce number of documents in Dataset
  • Simplify query complexity
  • Consider using Smart RAG if you need faster local processing

​
Next Steps

Smart RAG

More configurable RAG with local processing

Dataset

Configure document selection for Agentic RAG

Reranking

Add LLM-based reranking for better results

Evaluation

Measure retrieval quality with metrics