Quickstart
Follow these quickstart guides to start using MLflow for your GenAI app.
1. Setup your environment
Follow the setup your environment quickstart to install MLflow and connect your development environment to an MLflow Experiment.
2. Instrument your application using MLflow Tracing
MLflow Tracing helps you debug and iterate on your GenAI applications by capturing the entire execution flow, including prompts, retrievals, and tool calls. This allows you to monitor and improve the quality, cost, and latency of your application.
Based on your development environment, follow the appropiate quickstart below:
2. Evaluate your application's quality
MLflow empowers you to iteratively improve your GenAI application's quality by embedding evaluation directly into your development workflow. Systematically test changes to prompts, models, or application logic using mlflow.genai.evaluate() with LLM-based and custom scorers.
3. Collect human feedback
Incorporate human insight by capturing domain expert and end-user feedback to understand desired application behavior and align your custom LLM-judge metrics with expert judgment.
Next steps
Continue your journey with these recommended actions and tutorials.
- Debug & observe your app - Learn to use traces for debugging and understanding your app's behavior
- Evaluate & improve your app - Dive deeper into evaluation techniques and quality improvement
Reference guides
Explore detailed documentation for concepts and features mentioned in this guide.
- Tracing concepts - Understand the fundamentals of MLflow Tracing
- Evaluation concepts - Learn about scorers, judges, and evaluation methodology
- MLflow data model - Explore experiments, traces, and runs in depth