Build a production-ready RAG agent
summary
Builds a production RAG system: a Streamlit RAG agent wired to Gemini 3.5 Flash via the Google Agent Development Kit, seeded with a small JSON menu (eight items) and migrated to Cloud Firestore with text-embedding-005 vectors and a 768‑dim index using cosine nearest‑neighbor search. Deploys from source to Cloud Run via Cloud Build, creates a least‑privilege service account, and demonstrates live recommendations that call a get_menu tool before answering — the app deploys and returns grounded responses.
AI-written summary · about
chapters (30)
- 0:00Building RAG agent
- 1:00Workshop overview
- 3:35Coffee shop journey
- 4:46AI agent pillars
- 5:54Technical objectives
- 7:44Workshop logistics
- 10:38Cloud Shell setup
- 13:35Virtual TA access
- 15:02Menu JSON source
- 20:06Grounding agent in data
- 23:19File RAG vs vector search
- 26:13Dependencies and agent
- 27:18Agent orchestrator with ADK
- 31:40Streamlit UI
- 33:52Deploy to Cloud Run
- 34:13Least privilege IAM
- 37:17Binding roles to Cloud Run
- 38:24Cloud Run flags
- 40:07Security for production
- 42:52Recap deployed app
- 44:22Test recommendation agent
- 45:53Conversation state handling
- 55:33Migrate to Firestore
- 1:04:28Enable Firestore database
- 1:05:35Seed Firestore data
- 1:12:01Embedding model need
- 1:13:32Firestore vector indexes
- 1:19:13Grant permissions setup
- 1:20:01Firestore migration recap
- 1:21:54Q&A advanced agents
related talks
27mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
53mBuild RAG from Scratch
50mPractical Agentic RAG with SpringAI in Modern Enterprise Applications
01h 16mForget RAG Pipelines—Build Production Ready Agents in 15 Mins
11mBuilding easily deployable RAG applications with Llamastack
56mRAG: from dumb implementation to serious results
41mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
25mSet Up an Open Source RAG in 30 Minutes