Agentic RAG System: Two Approaches

This repository demonstrates two different approaches to building an agentic Retrieval-Augmented Generation (RAG) system focused on Pydantic AI documentation:

1. Pydantic AI Implementation: A Python-based implementation using Pydantic AI framework
2. n8n Implementation: A workflow-based implementation using n8n and Crawl4AI

Both implementations achieve the same goal: creating an AI assistant that can answer questions about Pydantic AI by retrieving information from its documentation.

Project Structure

• /paydanti-ai-version: Python-based implementation using Pydantic AI framework
• /n8n-version: Workflow-based implementation using n8n

Comparison

These implementations represent two fundamentally different approaches to building AI systems:

Feature	Pydantic AI Approach	n8n Approach
Development paradigm	Code-first	Visual workflow
Language	Python	JSON workflow definition
Learning curve	Python knowledge required	Visual interface, less coding
Customization	Highly customizable	Limited to available nodes
Deployment	Standard Python deployment	n8n server required
Web crawling	Using Crawl4AI library	Using Crawl4AI service
Database	Supabase vector database	Supabase vector database

Getting Started

Pydantic AI Version

See detailed instructions in the pydantic-ai-version README.

n8n Version

See detailed instructions in the n8n-version README.

Further Reading

For a more detailed comparison between these two approaches, check out my Medium article read me.

License

This project is licensed under the MIT License - see the LICENSE file for details.