Real-time RAG: How to augment LLMs with Redis and Amazon Bedrock
summary
Built over 100 RAG systems this year, the approach pairs vector search in Redis Enterprise with models via Amazon Bedrock and a client called Redis VL to deliver real‑time updates and semantic caching. Claims: LLM QPS often sits around two, fine‑tuning is slow and costly, while inserting a JSON or hash into Redis takes ~2–4 ms. Semantic cache uses similarity thresholds (e.g., 98–99%) to avoid repeated LLM calls, cutting latency and invocation costs significantly.
AI-written summary · about
chapters (8)
- 0:00Introduction to rag challenges
- 1:15Data strategy and LLM costs
- 4:26Redis enterprise capabilities
- 7:50Principles of RAG
- 10:24Benefits of RAG systems
- 12:35Semantic caching techniques
- 16:25Practical research examples
- 18:10Bedrock and closing remarks
related talks
14mOptimizing RAG with Semantic Caching & LLM Memory
59mThe Evolution of RAG
14mSmarter RAG, Smaller Bill: Optimize for Performance and Price
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
28mNavigating Challenges and Enhancing Performance of LLM based Applications
22mPractical Strategies for Implementing Domain-Specific LLMs
48mCrafting your own RAG system: Leveraging 30+ LLMs for enhanced performance
31mBeyond LLMs and Maturing RAG use cases