querying-mlflow-metrics

Name: querying-mlflow-metrics
Availability: InStock
Author: MLflow

Fetch and aggregate MLflow metrics with time-series and custom filters.

69|21|Updated Jan 14, 2026

Search Tags:#metrics #data retrieval #mlflow #model performance #time-series analysis

Ready to run

npx skills add https://github.com/mlflow/skills --skill querying-mlflow-metrics-mlflow

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: querying-mlflow-metrics
Source: https://github.com/mlflow/skills/tree/main/querying-mlflow-metrics
Command: npx skills add https://github.com/mlflow/skills --skill querying-mlflow-metrics-mlflow

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires requests, pandas, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill streamlines the process of querying and analyzing MLflow metrics, providing quick insights into model performance and usage patterns.

Core Features & Use Cases

Fetch Aggregated Metrics: Access and aggregate trace metrics such as token usage, latency, and trace counts.
Time-Series Analysis: View trends and changes over time for various metrics.
Custom Filters: Apply filters to narrow down results by experiment, trace, or other dimensions.
Use Case: For a data scientist looking to monitor the performance of a model over time, this Skill can provide a quick overview of key metrics like latency and token usage.

Quick Start

Fetch the average latency and 95th percentile of latency for all traces in the last 24 hours.

python scripts/fetch_metrics.py -s http://localhost:5000 -x 1 -m latency -a AVG,P95 -d trace_name -t 3600 --start-time="-24h" --end-time=now

Frequently Asked Questions about querying-mlflow-metrics

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I retrieve MLflow metrics for time-series analysis?▼

You can retrieve MLflow metrics for time-series analysis by running a script that connects to your MLflow tracking server, fetches trace data, and aggregates metrics like token usage and latency over specified time intervals.

What MLflow metrics can I fetch and analyze using Python?▼

You can fetch and analyze MLflow trace metrics including token usage, latency, and trace counts. The Skill aggregates these metrics to provide insights into model performance and usage patterns over time.

How do I calculate average and 95th percentile latency from MLflow traces?▼

Calculate average and 95th percentile latency from MLflow traces by running the fetch_metrics script with parameters for your server URL, metric name, aggregation functions like AVG and P95, and the desired time window.

Do I need pandas and requests to query MLflow metrics?▼

Yes, you need both the requests and pandas Python libraries. Requests handles connecting to the MLflow tracking server, while pandas structures the fetched metric data for aggregation and time-series analysis.

Can I filter MLflow metrics by experiment or trace dimensions?▼

Yes, you can filter MLflow metrics by experiment, trace, or other dimensions. The Skill supports custom filtering to narrow down results, allowing you to isolate specific data points for targeted performance analysis.