.NET: [Feature]: MLflow integration

[ramtinz]

Description

Problem

I’d like to use MLflow together with Agent Framework for:
• experiment tracking (inputs/outputs, latency, tool usage)
• saving artifacts (logs, traces, evaluation results)
• optionally integrating with MLflow Model Registry and/or MLflow Serving

Currently I’m not sure what the best integration pattern is, or whether there are extension points/hooks intended for this.

Proposal

Provide guidance (and ideally an example) for integrating Agent Framework with MLflow.

Desired capabilities
• Log each agent run as an MLflow run (metadata + artifacts)
• Capture tool calls / execution traces (if available)
• Optional: load a model from MLflow registry or call an MLflow serving endpoint

Acceptance criteria
• Documentation section: “Observability / Tracking with MLflow”
• Minimal code example (Python and/or .NET) showing:
• starting an MLflow run
• logging key metrics (latency, tokens/cost if available)
• logging prompts/responses as artifacts

Thanks! 🙏

Code Sample

Language/SDK

Python