Week 1: Agent Fundamentals & Tooling
Learn the foundational concepts of AI agents by implementing them from scratch.
Goal: By the end of this week, you'll understand how agents work internally and be able to build basic tool-using agents.
Learning Objectives
- ✅ Understand the ReAct (Reasoning + Acting) pattern
- ✅ Implement Thought → Action → Observation loops
- ✅ Define tool schemas with Pydantic
- ✅ Use OpenAI Function Calling API
- ✅ Integrate real web search with Tavily
Sessions
Week Overview
Key Concepts
What is an Agent?
An agent is an LLM that can:
- Reason about what to do next
- Act by calling external tools
- Observe the results and continue
This is fundamentally different from a simple chatbot that only generates text.
The ReAct Pattern
Thought: I need to search for information about X
Action: search["X"]
Observation: [search results]
Thought: Now I have the information, I can answer
Final Answer: Based on my research, X is...Prerequisites
Before starting this week:
- Complete the Getting Started guide
- Ensure your
sotaazenvironment is active - Have your API keys configured
Time Estimate
| Session | Duration |
|---|---|
| Session 1: ReAct | ~2 hours |
| Session 2: Tool Calling | ~2 hours |
| Weekend Project | ~3-4 hours |
Ready? Let's start with ReAct Pattern →