Skip to main content

​
RAG Overview

Retrieval Augmented Generation (RAG) combines the power of large language models with your own data. Instead of relying solely on the model’s training data, RAG retrieves relevant information from your documents and uses it to generate accurate, grounded answers.

​
How It Works

  1. Index: Your documents are chunked and stored as embeddings
  2. Retrieve: When you ask a question, relevant chunks are found
  3. Generate: The LLM uses retrieved context to answer
  4. Cite: Sources are tracked for transparency

​
Architecture

PraisonAI’s RAG is built on a simple principle:
Knowledge indexes; RAG answers with citations.
  • Knowledge: Handles document ingestion, chunking, embedding, and retrieval
  • RAG: Thin orchestrator that combines Knowledge retrieval with LLM generation
This separation keeps the core SDK lean while providing powerful RAG capabilities.

​
Quick Example

from praisonaiagents import Agent

# Simplest approach - just works
agent = Agent(
    name="Research Assistant",
    knowledge=["research_paper.pdf", "data/"],
)

response = agent.start("What are the key findings?")
print(response)

​
When to Use RAG

Use CaseRAG Helps?
Q&A over documents✅ Yes
Summarizing reports✅ Yes
Code documentation lookup✅ Yes
General knowledge questions❌ No (use base LLM)
Real-time data❌ No (use tools/APIs)

​
Key Features

Citations

Every answer includes source references

Streaming

Real-time response streaming

Multi-Agent

Share knowledge across agents

CLI Support

Full CLI for indexing and querying

​
Next Steps

Quickstart

Get started in 5 minutes

RAG Module

Detailed API reference

CLI Commands

Command-line usage

Citations

Working with sources