W&B Community Workshop - Simple to Agentic RAG
[Workshop] From Simple to Agentic RAG: A Hands-on Workshop for Building Advanced Retrieval Systems
- by Bharat (param-bharat)
This hands-on workshop takes participants through the evolution of Retrieval Augmented Generation (RAG) systems, from basic implementations to sophisticated agentic architectures. Attendees will build progressively more complex RAG systems, learning key concepts and best practices at each stage. Starting with a simple RAG implementation, we'll explore advanced topics like query enrichment, hybrid retrieval strategies, and reranking, culminating in the development of an agentic RAG system. Through practical exercises and real-world examples, participants will gain hands-on experience with evaluation-driven development, understanding how to measure and improve RAG system performance. This workshop is ideal for developers and practitioners looking to build production-grade RAG applications.
Workshop Outline (1 hour 30 minutes):
1. Introduction and Setup
- RAG fundamentals and use cases
- Quick environment setup
- Overview of tools we'll use
2. Building a Basic RAG System
- Data ingestion basics
- Simple vector store setup
- Basic retrieval and generation
- Hands-on exercise: Building a simple documentation assistant
3. Advanced RAG Techniques
- Query enrichment overview
- Reranking implementation
- Hands-on exercise: Improving retrieval quality
- Quick evaluation metrics
4. Introduction to Agentic RAG
- Agentic RAG concepts
- Single-agent architecture
- Tool integration basics
- Hands-on exercise: Converting to simple agentic RAG
5. Wrap-up and Q&A
- Key takeaways
- Resources for further learning
- Discussion and questions
Prerequisites:
- Basic Python programming experience
- Familiarity with LLMs
- Laptop with Python 3.11+ installed
[Lightning Talk] Building Agents with Phidata
- by Manthan (manthanguptaa)
Agenda
Overview of Phidata.
Building an agent.
Demo in the playground.
Agentic Retrieval-Augmented Generation (RAG) example.
Part 1: Phidata Overview
Talk about the core features of Phidata
Part 2: Building an Agent
Showcase a simple example of building a finance agent with OpenAI and then how easy it is to change models by switching with Gemini and Ollama
Part 3: Playground Demo
Get this agent on the playground to showcase easy testing of the agent on the playground
Part 4: Simple RAG Example
Build a simple agentic rag example - more details https://x.com/manthanguptaa/status/1876856777108549989