Building easily deployable RAG applications with Llamastack
summary
A working GitHub RAG assistant is built: repos are cloned, documents chunked, embedded and stored in Postgres + pgvector, and semantic search returns source chunks that the model uses to answer repo questions. The stack uses the Llamastack rag tool and Llamastack’s agent framework to orchestrate retrieval and tool calls, with inference from a VLM server or cloud provider. Deployment uses Helm charts for OpenShift from the AI architecture charts repo and runs at scale.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:43What is RAG
- 2:06RAG system components
- 3:30Introducing Llama Stack
- 4:51GitHub RAG assistant demo
- 6:59Simplifying RAG workflows
- 7:19Deploying RAG with Helm/K8s
- 8:46Resources and closing
- 9:07Q&A on vector DBs
- 10:13RAG model challenges Q&A
related talks
16mOpenRAG: An open-source stack for RAG
20mProduction RAG apps made easier with Astra
36mProduction Ready RAG
01h 12mBuild RAG using Llama 3.1
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
42mTo RAG or not to RAG
19mThe RAG Stack We Landed On After 37 Fails
31mBuild an Agentic RAG system using Langchain, Ollama and Milvus