Agentic AI Crash Course
A small Python crash-course repository with runnable examples that progress from a plain OpenAI chat loop to tool-using AgentSpan agents with memory, web search, guardrails, structured output and human-approval steps.
This crash course is a set of runnable teaching examples for building AI agents: a minimal terminal chatbot using the OpenAI Responses API directly, an AgentSpan-based chatbot with conversation memory, a required greeting tool and Tavily-powered internet search, and a customer-support agent demonstrating input guardrails against prompt injection, a mock order-lookup tool, a refund tool that requires human approval, and structured output through a Pydantic model.
Built with Python 3.11 or newer, dependency-managed with uv, using the openai package for direct model access, the agentspan package for agents, tools, memory, guardrails, runtime and approvals, and tavily-agent-toolkit for search; requires an OpenAI API key and a Tavily API key. The scripts are interactive teaching examples rather than production services.