Context Engineering with Redis and LangChain

Modern AI systems are shifting from prompt engineering to context engineering, where developers program how information flows into an LLM rather than relying on a single prompt. Instead of treating the model as a chatbot, this approach builds structured pipelines that assemble instructions, retrieved knowledge, memory, and tool outputs into the context window — turning stateless models into reliable, production-ready agents.
 

This workshop explores how Redis and LangChain enable scalable context architectures by combining vector search, semantic caching, and agent memory into a unified system. Attendees will learn how to design efficient context pipelines that reduce token costs, improve latency, and maintain persistent state across sessions — key capabilities for building real-world AI applications beyond demos. 

 

1. Foundations of Context Engineering

• Moving beyond prompt engineering into programmable context pipelines

• Understanding the context window as working memory for agents

• Strategies for selecting, compressing, and isolating context signals

2. Memory Architectures for AI Agents

• Short-term vs long-term memory design

• Retrieval-augmented generation (RAG) patterns

• Using Redis vector search and agent memory systems for persistent context 

3. High-Performance Context Pipelines

• Semantic caching to reduce token usage and latency

• Context pruning and summarization strategies

• Designing for throughput, reliability, and real-time agent interactions

4. Orchestrating Context with LangChain

• Building multi-step agent workflows

• Managing tool outputs, conversation history, and state

• Structuring context flows for scalable production deployments