[feat] MLflow Integration

[tdadela]

Summary

This PR adds MLflow integration to the ReAgentAI project, enabling experiment tracking and visualization. The integration includes:

1. A new MLflowTracker class for tracking experiments, metrics, and artifacts
2. Docker configuration for running MLflow alongside ReAgentAI
3. Updated documentation and setup instructions

Changes

Features

• Added MLflow integration for experiment tracking
• Created Dockerfile.mlflow for running MLflow server in Docker
• Added docker-compose.yml for running both ReAgentAI and MLflow together
• Updated pyproject.toml to include MLflow dependency

[Copilot]

Pull Request Overview

Adds MLflow experiment tracking to ReAgentAI, updates the UI and main app to log metrics and parameters, and provides Docker and documentation support for MLflow.

• Introduces MLflowTracker for run management and logging
• Updates main.py and UI handlers to start runs and log to MLflow when enabled
• Adds Dockerfile, docker-compose setup, and documentation for running MLflow alongside the app

Reviewed Changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/test_reagentai.py	Adds tests for MLflowTracker initialization, lifecycle, and error paths
src/reagentai/common/mlflow_tracking.py	Implements the MLflowTracker class with start/end run and logging methods
src/reagentai/ui/app.py	Extends UI callbacks to accept an optional mlflow_tracker and log data
src/reagentai/main.py	Initializes and uses MLflowTracker, logging system info and ending runs
pyproject.toml	Adds mlflow to dependencies and pytest configuration
docker-compose.yml	Defines services for running ReAgentAI and MLflow together
README.md	Updates setup and run instructions to include MLflow usage
Dockerfile.mlflow	Provides a container image for the MLflow server

Comments suppressed due to low confidence (4)

README.md:58

• [nitpick] The command uv run mlflow server appears incorrect; it should likely be mlflow server (or use the correct CLI tool) without the uv run prefix.

uv run mlflow server

src/reagentai/ui/app.py:96

• The variable user_query is not defined in handle_bot_response. You should either include user_query in the function signature or extract it from inputs before using it.

f"query_{len(chat_history)}": user_query[:100]  # Truncate long queries

src/reagentai/main.py:23

• logger is not defined in this module; replace logger.info with logging.info or initialize a module-level logger via logger = logging.getLogger(__name__).

logger.info(f"MLflow tracking {'enabled' if run_id else 'disabled'}")

src/reagentai/main.py:59

• logger is not defined here either; use logging.info or define logger = logging.getLogger(__name__) at the top of the file.

logger.info("MLflow tracking session ended")

[janumiko]

Maybe add this? Because it doesn't work without that for me (the reagent module is not found when running tests).

pythonpath = [
"."
]

[janumiko]

nitpick: missing new line at the end

[janumiko]

I wouldn't pass the mlflow instance to the Gradio frontend, I don't think we should keep logic like that there.

Maybe we should pass the tracker to the Agent class itself.

[janumiko]

Maybe add a function like setup_mlflow for this?