mlflow-skinny

What it is and what it does

mlflow-skinny is a stripped-down MLflow client designed for environments where you want to log experiments, traces, and metrics to a remote MLflow server without installing heavy dependencies like SQL databases or data science libraries. It includes core tracking APIs, OpenTelemetry integration for distributed tracing, and support for logging LLM and agent interactions. The package is built on FastAPI, Pydantic, requests, and standard HTTP libraries, making it suitable for lightweight deployments, containerized applications, and edge environments.

You point it at a remote MLflow server via environment variable or API call, then use its tracking and logging functions to record experiment parameters, metrics, model artifacts, and execution traces. It is particularly useful for teams building AI agents or LLM applications who want observability without the overhead of a full MLflow installation, or for CI/CD pipelines and serverless functions that need to report results back to a central MLflow instance.

Use it for

• Log LLM application traces and metrics to a remote MLflow server from a lightweight client without installing data science libraries
• Integrate OpenTelemetry tracing into AI agent frameworks and capture execution spans in MLflow for debugging and monitoring
• Track experiment parameters and results in containerized or serverless environments where minimal dependencies are critical
• Enable distributed tracing across microservices by using mlflow-skinny as a common observability layer
• Record model predictions and performance metrics from production inference endpoints without bundling a full MLflow installation

Worth the install?

AI-flagged interpretation of the facts on this page. Verify before relying on it.

Install

mlflow-skinny on PyPI

Before you install

Low install friction with a pure-Python wheel and 20 runtime dependencies that are widely available. Active maintenance with a release 11 days old and 27503 repository stars indicate ongoing support.

Requires a remote MLflow server running at the tracking URI; set MLFLOW_TRACKING_URI environment variable or call mlflow.set_tracking_uri() before logging.

License in practice

Apache License 2.0 (permissive) allows commercial and private use with minimal restrictions; you must include a copy of the license and note any modifications you make.

Quickstart

pip install mlflow-skinny

import mlflow
mlflow.set_tracking_uri("http://localhost:5000")
mlflow.openai.autolog()

Package facts

Evidence: mlflow_skinny-3.15.1-py3-none-any.whl

Tags

More Python Modules packages

See also mlflow · mlflow-tracing · azureml-mlflow · mlserver-mlflow · arize · dagster-mlflow · openlit · arize-phoenix · promptflow-tracing · sagemaker-mlflow