mlflow-evaluation

What problem does it solve?

This Skill provides an end-to-end GenAI evaluation framework for ML agents using MLflow GenAI evaluation workflows, enabling structured assessment across multiple dimensions.

Core Features & Use Cases

• End-to-end evaluation orchestration using mlflow.genai.evaluate with datasets, scorers, and trace analysis
• Support for safety, correctness, relevance, and grounding checks across traces, datasets, scorers, and reference guidance
• Standardized evaluation runs and cross-version comparisons for CI/CD, experimentation, and benchmarking

Quick Start

1. Install MLflow with Databricks extras: pip install "mlflow[databricks]>=3.x" (adjust to your environment)
2. Prepare evaluation data and a local predict_fn wrapper for your agent
3. Run an evaluation: mlflow.genai.evaluate(data=eval_data, predict_fn=predict_fn, scorers=[...])