ReAct Agents & LangChain create_agent

Practise and Check

Deliverable

ReAct research agent: give a `create_agent` agent your RAG retriever and a web-search tool, add a middleware that blocks off-topic questions, and compare its answers with your T6 system.

Assessment

Answer the five Interview questions without notes, then run the lab.

Interview probe

Larkspur's support agent cost three times as much yesterday as the day before. Traces show some runs making a dozen `search_handbook` calls with reworded queries before giving up. What do you check, and what do you change?

Common mistakes

• Raising the call cap when the index stored questions without answers
• Making refund_amount required, so the model invents a value
• Sending a new user message to a thread paused for approval

Related Modules

• T7.1The Agent Loop & Tool Calling
• T7.3Agentic Workflow Patterns