PRODUCTION SCENARIO
A Generative AI Engineer has a LangGraph agent that already works in a notebook. The team wants to try it in AI Playground, run Agent Evaluation against it, and deploy it with monitoring.
What should the engineer do to the agent code?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Wrap the existing agent in MLflow's ResponsesAgent interface
Databricks recommends MLflow ResponsesAgent to build agents: wrapping an agent written in any framework gives out-of-the-box compatibility with AI Playground, Agent Evaluation and Agent Monitoring, plus automatic tracing and an inferred signature. Logging it as a bare LangChain model deploys it but skips that compatibility, and a hand-written ChatModel subclass is the older interface ResponsesAgent supersedes. A Unity Catalog function reshapes an agent into a batch SQL tool.