This is a contribution to the Redis AI Challenge Beyond the Cache.
I present Redis RAG Benchmark, a web application developed to compare in real time two question-and-answer pipelines showing answers side by side with millisecond timers to visualize latency differences.
Pipeline comparison: RAG without cache using in-memory FAISS index · Redis-powered system with RediSearch vector search and response caching in RedisJSON
The interface shows parallel chat panels with millisecond timers to highlight the performance advantage of the Redis solution in vector searches and cache hits.
Demo and code: watch the video at https://www.youtube.com/embed/UL18zWuGwwA and the repository at https://github.com/turazashvili/redis-rag-benchmark
Quick installation and requirements: Node.js 18+ · Docker and Docker Compose · OpenAI key
Essential step-by-step: clone the repository and configure .env with OPENAI_API_KEY · start Redis Stack with docker-compose up -d · install dependencies with npm run install-all · start the application with npm run dev · open https://localhost:3000
Summary architecture: Traditional RAG system: in-memory FAISS index, cosine similarity search 20-60 ms, GPT-3.5-turbo LLM on every query, no cache · Redis system: vector index with RediSearch, vector search 2-5 ms, RedisAI for embeddings or precomputed vectors, LLM only on cache miss, RedisJSON for response caching with TTL
Performance comparison: Traditional RAG vector search 20-60 ms vs Redis 2-5 ms · Cache hit in Redis under 10 ms · Cache miss similar 500-1500 ms due to LLM call · Reduction of LLM calls up to 90 percent when cache works
How Redis is used in the project: RediSearch for in-memory vector indexes and ultra-fast searches · RedisAI to host embedding models or serve precomputed vectors · RedisJSON to store complete LLM responses with TTL and serve hits in under 10 ms
Key benefits: millisecond search latencies · lower cost per query by avoiding unnecessary LLM calls · better scalability and data control
About Q2BSTUDIO: Q2BSTUDIO is a software development company specialized in custom applications and custom software · we offer artificial intelligence and AI solutions for businesses, creation of AI agents and projects with Power BI for business intelligence · our services include cybersecurity, AWS and Azure cloud services and business intelligence services consulting · as artificial intelligence specialists we apply efficient models and architectures that combine Redis, embeddings and LLM to provide fast and secure responses
If you are looking to integrate custom applications, custom software or enhance your processes with artificial intelligence, AI agents and business intelligence solutions with Power BI, Q2BSTUDIO can help you design and implement the right solution with a focus on performance, security and cost savings in AWS and Azure cloud services.
Contact Q2BSTUDIO for custom applications, artificial intelligence, cybersecurity, cloud services and business intelligence services projects. Keywords to improve positioning: custom applications custom software artificial intelligence cybersecurity cloud services aws and azure business intelligence services ai for businesses AI agents power bi.
