Building Smarter Mobile Apps with On-device AI Agents & GraphRAG
summary
Shows a concrete developer path: start with an LLM-only agent (Claude 3.7 Sonnet on Amazon Bedrock) that confidently misanswers (claims Transcendence’s Wally Pfister as a multi-film director), add vector RAG and a search tool to get better but still wrong numeric/metadata checks (suggests Tenet but misses the runtime <150 constraint), then switch to a GraphRAG using Neo4j where the agent calls get graph schema and execute Cypher via the Strands agent SDK in React Native and returns five results that satisfy all five constraints.
AI-written summary · about
chapters (7)
- 0:00Introduction to agentic apps
- 1:53What makes an agentic app
- 8:59Building with Strands SDK
- 14:01Limitations of RAG alone
- 19:53Why knowledge graphs matter
- 23:43Introducing GraphRAG
- 29:46Key takeaways and conclusion
related talks
53mAgents, Tools, and Mcp, Oh My! Next Level AI Concepts for Developers
45mAgents, Tools, and MCP, oh my! Next-level AI concepts for developers
12mLeveraging Neo4j Graph Data and AI Across the App Development and Deployment Lifecycle
30mGenAI Beyond Chat with RAG, Knowledge Graphs, and Python
26mJust Ask: Building a Natural Language GraphQL Agent from the Graph You Already…
12mRevolutionizing Platform Engineering with AI Agents and GraphRAG
29mBuilding Modular AI Agents with LangGraph, MCP, and
55mStop AI Agent Hallucinations: 5 Techniques + Production Patterns