v3.4.0

✨ New Features

• OpenTelemetry Metrics Export: MLflow now exports span-level statistics as OpenTelemetry metrics, providing enhanced observability and monitoring capabilities for traced applications.
• MCP Server Integration: Introducing the Model Context Protocol (MCP) server for MLflow, enabling AI assistants and LLMs to interact with MLflow programmatically.
• Custom Judges API: New `make_judge` API enables creation of custom evaluation judges for assessing LLM outputs with domain-specific criteria.
• Correlations Backend: Implemented backend infrastructure for storing and computing correlations between experiment metrics using NPMI (Normalized Pointwise Mutual Information).
• Evaluation Datasets: MLflow now supports storing and versioning evaluation datasets directly within experiments for reproducible model assessment.
• Databricks Backend for MLflow Server: MLflow server can now use Databricks as a backend, enabling seamless integration with Databricks workspaces.
• Claude Autologging: Automatic tracing support for Claude AI interactions, capturing conversations and model responses.
• Strands Agent Tracing: Added comprehensive tracing support for Strands agents, including automatic instrumentation for agent workflows and interactions.
• Experiment Types in UI: MLflow now introduces experiment types, helping reduce clutter between classic ML/DL and GenAI features; auto-detects the type and allows adjustment via a selector next to the experiment name.
• Add ability to pass tags via dataframe in mlflow.genai.evaluate.
• Add custom judge model support for Safety and RetrievalRelevance builtin scorers.
• Add AI commands as MCP prompts for LLM interaction.
• Add MLFLOW_ENABLE_OTLP_EXPORTER environment variable.
• Support OTel and MLflow dual export.
• Make set_destination use ContextVar for thread safety.
• Add MLflow commands CLI for exposing prompt commands to LLMs.
• Add 'mlflow runs link-traces' command.
• Add 'mlflow runs create' command for programmatic run creation.
• Add MLflow traces CLI command with comprehensive search and management capabilities.
• Add --env-file flag to all MLflow CLI commands.
• Backend for storing scorers in MLflow experiments.
• Allow cross-workspace copying of model versions between WMR and UC.
• Add automatic Git-based model versioning for GenAI applications.
• Improve WheeledModel._download_wheels safety.
• Support resume run for Optuna hyperparameter optimization.
• Add MLFLOW_DEPLOYMENT_CLIENT_HTTP_REQUEST_TIMEOUT environment variable.
• Add ability to hide/unhide all finished runs in Chart view.
• Add MLflow OSS telemetry for invoke_custom_judge_model.