🤖 Agentic AI Workshop: From RAG to Multi-Agent Systems

A comprehensive workshop repository demonstrating the evolution from basic AI concepts to sophisticated multi-agent systems. This hands-on learning path covers RAG (Retrieval-Augmented Generation), LLMs, and intelligent agents using modern frameworks and best practices.

🎯 Workshop Overview

This repository is structured as a progressive learning journey through four modules, each building upon the previous to create a complete understanding of modern AI agent architectures.

Learning Path

Module 1: Foundations (GIP - Get In Position)
    ↓
Module 2: RAG Systems (Vector Databases & Retrieval)
    ↓
Module 3: Single Agents (Autonomous Decision Making)
    ↓
Module 4: Multi-Agent Systems (Orchestration & Collaboration)

📚 Modules

Module 1: GIP - Get In Position

Location: src/mod-1-gip/

Foundation concepts and environment setup for AI development.

• Understanding LLMs and their capabilities
• Setting up development environments
• Basic prompt engineering
• Introduction to AI frameworks

Module 2: RAG - Retrieval-Augmented Generation

Location: src/mod-2-rag/

Building intelligent systems that combine retrieval with generation.

🔧 Components:

• rag-llama-index/ - Production RAG pipeline with LlamaIndex

  • Multi-collection vector database architecture (Astra DB)
  • Document preprocessing (PDF, JSON, CSV, DOCX)
  • Semantic chunking and embedding generation
  • Metadata extraction and quality assurance
  • Observability with Langfuse integration

• langflow-pipelines/ - Visual RAG workflows

  • PDF processing with Docling
  • Multi-modal file handling
  • ChromaDB vector storage
  • Online query systems

Module 3: Single Agent Systems

Location: src/mod-3-agent/

Creating autonomous agents with decision-making capabilities.

🔧 Components:

• dify-n8n-pipelines/ - Enterprise agent workflows

  • Spark specialist assistant ("Ask Lumi")
  • Knowledge base integration
  • Workflow automation

• langflow-pipelines/ - Visual agent builders

  • Agent with ChromaDB knowledge base
  • Tool integration patterns
  • Decision tree implementations

Module 4: Multi-Agent Orchestration

Location: src/mod-4-multi-agents/

Advanced multi-agent collaboration and orchestration patterns.

🔧 Components:

• crew-ai-agents/ - Restaurant Recommendation System

  • 3 specialized agents working in harmony:
    • 🍴 Restaurant Concierge (finds matching venues)
    • 🥗 Dietary Specialist (ensures food safety)
    • 💰 Promotions Manager (discovers best deals)
  • Sequential task execution
  • YAML-based configuration
  • Built-in CrewAI tools
  • LangFuse v3 observability integration
  • Production-ready with full tracing

• langflow-pipelines/ - Visual multi-agent workflows

  • Agent communication patterns
  • Parallel and sequential processing
  • Result aggregation strategies

🛠️ Technology Stack

Core Frameworks

• LlamaIndex - Advanced RAG and data ingestion
• CrewAI - Multi-agent orchestration
• Langflow - Visual workflow builder
• Dify - Enterprise AI application platform
• n8n - Workflow automation

Infrastructure

• Vector Databases: Astra DB, ChromaDB, Supabase
• LLMs: OpenAI GPT-4o-mini, Claude, local models
• Observability: Langfuse v3, OpenTelemetry
• Data Processing: Docling, pandas, embeddings

Languages & Tools

• Python 3.9+ - Primary development language
• YAML - Configuration management
• JSON - Pipeline definitions
• Rich - Terminal UI components

🚀 Quick Start

Prerequisites

python --version

python -m venv venv
source venv/bin/activate

Installation

git clone https://github.com/yourusername/ws-agentic-ai-smart-agents.git
cd ws-agentic-ai-smart-agents

cd src/mod-4-multi-agents/crew-ai-agents
pip install -r requirements.txt

cp .env.example .env

Running Examples

RAG Pipeline (Module 2)

cd src/mod-2-rag/rag-llama-index
python ingestion.py

Multi-Agent System (Module 4)

cd src/mod-4-multi-agents/crew-ai-agents
python main.py

🚀 Workshop Structure

Day 1: Foundations & RAG

• Morning: Module 1 - AI fundamentals, LLMs, prompt engineering
• Afternoon: Module 2 - Building RAG systems, vector databases

Day 2: Agents & Orchestration

• Morning: Module 3 - Single agent development, tools, decision-making
• Afternoon: Module 4 - Multi-agent systems, CrewAI, production deployment

🎓 Learning Objectives

By completing this workshop, you will:

1. ✅ Understand the fundamentals of LLMs and prompt engineering
2. ✅ Build production-ready RAG systems with vector databases
3. ✅ Create autonomous agents with decision-making capabilities
4. ✅ Orchestrate multi-agent systems for complex tasks
5. ✅ Implement observability and monitoring for AI systems
6. ✅ Deploy AI applications with enterprise-grade patterns

🛠️ Key Features Demonstrated

• Production Patterns: Error handling, retry logic, graceful degradation
• Observability: Full tracing with Langfuse v3, performance metrics
• Scalability: Multi-collection architectures, parallel processing
• Security: API key management, data validation, safe prompting
• User Experience: Rich terminal interfaces, progress tracking
• Testing: Quality assurance, validation pipelines

🤝 Use Cases

Restaurant Recommendation System (Module 4)

Complete multi-agent system that:

• Searches restaurant databases
• Validates dietary restrictions
• Finds promotional offers
• Generates personalized recommendations

Document Processing Pipeline (Module 2)

RAG system that:

• Ingests multiple file formats
• Creates semantic embeddings
• Stores in vector databases
• Enables intelligent retrieval

📚 Resources

Documentation

• CrewAI Docs
• LlamaIndex Docs
• Langflow Docs
• Langfuse Docs

Datasets

• Restaurant data (8 venues with full details)
• Promotional offers (65+ active coupons)
• Allergy guidelines and dietary information
• Multi-cuisine menu collections

🔧 Configuration

Each module includes:

• .env.example - Environment variable templates
• requirements.txt - Python dependencies
• Configuration files (YAML/JSON)
• README with specific instructions

🚨 Important Notes

• API Keys Required: OpenAI, Astra DB, Langfuse (depending on module)
• Python Version: 3.9+ required for latest features
• Storage: ~500MB for sample data and models
• Internet: Required for API calls and package installation

---

For questions or support, please open an issue in the repository.