Event Streaming and its Potential for the GenAI World
summary
Introduces LangStream as an open‑source, low‑code event‑streaming library that connects Kafka/Pulsar sources to LLMs (OpenAI, Vertex AI, Hugging Face) and vector stores (Astra/Cassandra, Pinecone, Milvus) to implement RAG (retrieval-augmented generation). Shows embedding dimensions (OpenAI 1536), LLM token limits (4k / 16k), and uses ANN/HNSW-style nearest‑neighbor search and a JVector implementation for fast vector queries. Demo runs locally via Docker and YAML pipelines, streams real‑time data into embeddings, and returns working chat responses though latency varies.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 1:56Background on AI and GenAI
- 8:14Generative AI and LLMs explained
- 14:37Vector databases and embeddings
- 23:42RAG pattern for LLMs
- 32:33LangStream introduction
- 35:00LangStream architecture and demo
- 43:55Resources and closing
related talks
46mEvent Streaming for Multi-Agentic Generative AI Workflows
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases
47mReal-Time AI with Open Source
48mEnter the Brave New World of GenAI with Vector Search
33mEnter the Brave New World of GenAI with Vector Search
41mAdding Generative AI to Real Time Streaming Pipelines
57mEnter The Brave New World of GenAI with Vector Search
26mUnleashing the Potential of Cloud Native Open Source Vector Databases