Building a Weather Agent on Android with Koog
summary
Adds the Koog Kotlin dependency and an OpenAI LLM executor, builds an agent with an executor, system prompt, and model; implements two concrete tools—search locations and fetch 5‑day forecast—groups them into a tool registry, and exposes them via annotated functions so the agent calls the weather API. An event handler logs before/after LLM calls. The single‑run agent fetches keys, retrieves forecasts, and returns results (including packing advice); it works but cannot handle clarifying questions within one run without a different strategy.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 1:08What is an AI agent
- 6:30Introducing Cook framework
- 7:35Building weather agent
- 8:18Adding Cook dependency
- 9:02First AI agent instance
- 13:44Tools in Cook framework
- 15:10Aqua weather API
- 16:14Weather tools creation
- 18:03Grouping tools
- 18:24Registering tools
- 19:06Testing agent queries
- 21:55Debugging with event handlers
- 25:09Agent workflow
- 26:11Customizing agent behavior
- 26:32Agent strategy in Cook
- 28:23Predefined strategies
- 29:47Custom agent strategy
- 32:37Advanced Cook features
- 32:57Summary
related talks
15mBuilding AI Agents in Kotlin with Koog
51mReliable AI Agents Using Domain Modeling with Koog in Java
46mPowering Compose Multiplatform Apps with Koog & MCP
41mBuilding LLM Agents – Challenges and Learnings
33mCopilot in Android: AI for boring stuff, people - for important
51mSupercharging Spring AI: Scalable AI Agents with Koog + Spring AI + Spring Boot
01h 06mBuilding Smart Agents with LlamaIndex: A Comprehensive Overview
41mBuilding with AI in Kotlin