Day 1 13 Jason Mayes
summary
Demonstration builds an on-device agent that runs the Gemma 2B model in the browser via the Mediap web LLM library, exposes existing site functions as callable tools, and orchestrates follow‑ups and structured JSON to autonomously book flights and even spin up a DJ using Spotify. Prototype downloads the model once, runs locally on a five‑generation old Nvidia 1070 GPU, uses a caching proxy library for large binaries, yields useful results with some hallucination risk, and shows hybrid cloud patterns for sensitive or up‑to‑date data. agentic behaviors
AI-written summary · about
chapters (10)
- 0:00Introduction to web AI agents
- 5:48Demo of agentic web prototype
- 10:21Caching proxy library overview
- 17:03Web AI future and hybrid approaches
- 20:53Q&A on privacy and model security
- 23:49Q&A on API tool access and data flow
- 25:34Q&A on accessibility and language support
- 27:19Q&A on JSON output consistency
- 29:27Q&A on fine-tuning vs RAG
- 30:54Q&A on testing and validation
related talks
29mClient side Web AI Agents Building smarter user experiences for a future age
30mLook ma, no hands! Multimodal AI-Agents in the browser
30mUnleashing AI in Your Web Apps
31mBringing and Running AI Agents in the Browser
22mBoring but Effective: WebAI That Ships
14mGemma on mobile and web. Best and worst practices
01h 00mHow to work with generative AI in JavaScript
20mFrom ML to LLM: On-Device AI in the Browser