v3.4.0rc0
📦 mlflowView on GitHub →
✨ 8 features🔧 1 symbols
Summary
MLflow 3.4.0rc0 introduces major new capabilities including OpenTelemetry metrics, MCP server integration, a custom judges API, correlations backend, evaluation datasets, Databricks backend support, Claude autologging, and Strands agent tracing.
✨ New Features
- OpenTelemetry Metrics Export: MLflow now exports span-level statistics as OpenTelemetry metrics for enhanced observability.
- MCP Server Integration: Added Model Context Protocol (MCP) server enabling AI assistants and LLMs to interact programmatically with MLflow.
- Custom Judges API: Introduced `make_judge` API for creating custom evaluation judges for LLM output assessment.
- Correlations Backend: Implemented backend infrastructure for storing and computing correlations between experiment metrics using NPMI.
- Evaluation Datasets: Added support for storing and versioning evaluation datasets within experiments.
- Databricks Backend for MLflow Server: MLflow server can now use Databricks as a backend.
- Claude Autologging: Automatic tracing support for Claude AI interactions.
- Strands Agent Tracing: Added comprehensive tracing support for Strands agents with automatic instrumentation.