Question 12
Domain 3: Application DevelopmentOn serverless compute, what must you do to get automatic LangChain traces in MLflow?
Correct answer: D
Explanation
MLflow autologging is the mechanism for automatic trace capture, and the exam guide explicitly lists “Use MLflow and Agent Framework for developing agentic systems” and “Evaluate agent performance using MLflow scoring and tracing.” Calling `mlflow.langchain.autolog()` enables LangChain integration so traces are logged automatically in MLflow on serverless compute.
Why each option is right or wrong
A. Enable Delta Change Data Feed
Delta Change Data Feed tracks row-level table changes, not LangChain execution traces.
B. Create a Direct Vector Access index
Vector Search indexes support retrieval over embeddings, not tracing or logging chain runs.
C. Wrap the chain in a SQL function
SQL functions can wrap logic, but they do not enable MLflow LangChain autologging.
D. Call `mlflow.langchain.autolog()`
MLflow’s LangChain integration is enabled through the autologging API, which is the mechanism that automatically captures traces without manual span logging. In the Databricks exam guide, this aligns with the objective to use MLflow and Agent Framework for agentic systems and to evaluate agent performance using MLflow scoring and tracing; on serverless compute, you must invoke `mlflow.langchain.autolog()` in the active Python session before running the chain so the traces are recorded automatically in MLflow.