mlflow

Why we included this project

When you're running LLM apps and agents in production, you need to see what they're actually doing, and MLflow gives you that in one open-source tool. It records full traces of agent and LLM calls, runs evaluations with built-in metrics and LLM judges, and keeps an eye on quality and cost over time so problems surface before users hit them. It also still handles the classic ML workflow: experiment tracking, a model registry, and deployment to batch or real-time serving on Kubernetes, SageMaker, and similar targets. That breadth makes it a practical default for small teams that want tracing, evaluation, monitoring, and model management in one place instead of stitching together several point solutions, and it plugs into OpenTelemetry and most major agent frameworks and model providers without forcing a rewrite.