AI agents represent a shift from simple chatbots by operating independently based on triggers and using data, tools, and memory.
OpenAI provides powerful built-in tools like Code Interpreter and Web Search, supported by an evolving set of APIs for agent development.
The Agents SDK simplifies development by providing a clear structure for defining an agent, its instructions, and running it synchronously or asynchronously.
The Agents SDK allows you to easily extend an agent's capabilities by decorating a standard Python function and assigning it as a tool.
Agents can be equipped with pre-built functionalities like the web search tool to access and process up-to-date information from the internet.
Maintain context across multiple turns in a conversation by collecting the message history and passing it as a chained list to the agent.
The handoff pattern uses a central triage agent to analyze a request and delegate it to the most appropriate specialized agent in a team.
Create a collaborative team where a primary agent orchestrates complex tasks by using other specialized agents as callable tools.
Guardrails act as programmable input or output filters that check agent messages against predefined rules to ensure safe and appropriate responses.
Building effective AI agents is less about mastering complex APIs and more about applying the right architectural concepts and patterns.
24:22Philipp Schmid
39:11Thomas Dohmke
24:12Brad Axen
23:33Oliver Will
26:35Ricardo
30:01Mario Fusco
23:39Dennis Zielke & Rene Pajta
27:07Eric Jadi & Rene Pajta
Jobs that call for the skills explored in this talk.
Deutsches Medizinrechenzentrum GmbH
4Com GmbH & Co. KG