Redis Context Engineering Workshop

This repository contains:

• Workshop notebooks in workshop/
• Six staged agent demos in demos/ that cover baseline RAG → context engineering → full agent → ReAct → memory.

---

Workshop Setup

Prerequisites

• Python: 3.11+
• Docker: for Redis + Agent Memory Server
• OpenAI API key: set OPENAI_API_KEY

Quick setup (recommended: uv)

Setup the Python environment:

uv sync

Create an env file:

cp .env.example .env

Edit env file as needed. Start the docker infra stack:

docker-compose up -d

Load the workshop data:

uv run load-hierarchical-courses \
  -i src/redis_context_course/data/hierarchical/hierarchical_courses.json \
  --force

Verification (optional)

uv run pytest tests/ -v

• Troubleshooting: see SETUP.md

---

Workshop Outline

This workshop guides you through the essential steps of building advanced agentic systems: starting with foundational context engineering concepts, progressing through RAG techniques, diving into practical data engineering, and culminating in the design of memory-enhanced AI agents.

Sections

Module	Time	Notebook	Key Highlights
1. Introduction	45 min	01_introduction_to_context_engineering.ipynb	Overview of context types, failures, and token budgeting strategies.
2. RAG Essentials	60 min	02_rag_essentials.ipynb	Semantic search, embeddings, and RAG patterns.
3. Data Engineering	75 min	03_data_engineering_theory.ipynb	Data pipelines, chunking methods, and preparing retrieval-ready data.
4. Memory Systems	90 min	04_memory_systems.ipynb	Working vs. long-term memory and memory-augmented RAG for agents.

Running notebooks

cd workshop

# Execute a specific notebook (optional)
jupyter execute 02_rag_essentials.ipynb --inplace

Module 4 note: the Redis Agent Memory Server must be running with OPENAI_API_KEY set (the provided docker-compose.yml loads it from your .env).

---

Agent Demos

Six CLI demos that progressively add capabilities. Use --help for all options, --quiet for minimal output, --show-reasoning for ReAct traces (stages 4–6).

# 1. Baseline RAG — naive retrieval, no optimization
uv run 1-baseline-rag "What machine learning courses are available?"

# 2. Context-engineered — cleaned/transformed context, progressive disclosure
uv run 2-context-engineered "What machine learning courses are available?"

# 3. LangGraph agent — structured workflow with intent routing
uv run 3-langgraph-agent "What courses teach machine learning?"

# 4. Hybrid + ReAct — adds NER-based hybrid search and visible reasoning
uv run 4-hybrid-react --show-reasoning "What are the prerequisites for CS002?"

# 5. Working memory — multi-turn conversations within a session
uv run 5-working-memory --student-id alice --session-id s1 "What is CS004?"

# 6. Full memory — working + long-term memory with preference tracking
uv run 6-full-memory --student-id alice --show-reasoning "What courses do you recommend?"