skills.rest
aaronachermann
aaronachermannCOMMUNITY
Explore all skills by this author →
GitHub

querying-mlflow-metrics

Fetch aggregated MLflow trace metrics with time-range bucketing and JSON or table output.

Updated Apr 18, 2026
Search Tags:#metrics#aggregation#observability#latency#traces#mlflow#token-usage
One-click install
npx skills add https://github.com/aaronachermann/PolentaEncoders --skill querying-mlflow-metrics-aaronachermann
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: querying-mlflow-metrics
Source: https://github.com/aaronachermann/PolentaEncoders/tree/main/.github/skills/querying-mlflow-metrics
Command: npx skills add https://github.com/aaronachermann/PolentaEncoders --skill querying-mlflow-metrics-aaronachermann

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Aggregates and summarizes MLflow trace metrics so teams can quickly understand token usage, latency, trace counts, and quality evaluations without manually querying raw run data.

Core Features & Use Cases

  • Flexible Aggregations: Compute COUNT, SUM, AVG, MIN, MAX, P50, P95, and P99 across specified metrics.
  • Time and Dimension Bucketing: Produce hourly/daily buckets and group-by dimensions like trace_name or trace_status to reveal trends and hotspots.
  • Quality & Assessment Reporting: Summarize assessment scores and counts by evaluator to monitor LLM output quality.
  • Use Case: An ML engineer analyzing LLM inference costs and latency trends across experiments to optimize model routing and billing.

Quick Start

Use the querying-mlflow-metrics skill to fetch SUM and AVG of total_tokens for experiment 1 from http://localhost:5000 over the last 24 hours.

Frequently Asked Questions about querying-mlflow-metrics

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

FAQPage Schema
How do I aggregate MLflow trace metrics like token usage and latency percentiles?▼

You can aggregate MLflow trace metrics by fetching COUNT, SUM, AVG, MIN, MAX, and latency percentiles (P50, P95, P99) across specified metrics using experiment IDs and a reachable MLflow tracking URI.

What's the best way to group MLflow trace metrics by time windows and dimensions?▼

Grouping MLflow trace metrics is done by applying time-range bucketing for hourly or daily intervals and using dimensions like trace_name or trace_status to reveal trends and hotspots.

Does this approach require a specific MLflow tracking server setup to query metrics?▼

Querying MLflow metrics requires a reachable MLflow tracking URI, specific experiment IDs, and selected metric names, ensuring your MLflow tracking server is accessible to fetch the trace data.

Can I summarize LLM assessment scores and evaluator counts using MLflow traces?▼

You can summarize LLM assessment scores and counts by evaluator from MLflow traces to monitor output quality, applying aggregations to assessment metrics across experiments and time windows.

What output formats are supported when fetching aggregated MLflow trace metrics?▼

Fetching aggregated MLflow trace metrics supports JSON and table output formats, allowing teams to inspect token usage, latency percentiles, trace counts, and assessment scores in their preferred structure.

Related Skills

Explore More Data & Analytics Skills→
teegin-gteegin-g

querying-mlflow-metrics

Fetch aggregated MLflow tracking metrics from servers in table or JSON formats.

Community
Intermediate
BlackkadderBlackkadder

querying-mlflow-metrics

Fetch aggregated trace metrics from MLflow tracking servers.

Community
Intermediate
ScottHMcKeanScottHMcKean

querying-mlflow-metrics

Fetch aggregated trace metrics from MLflow tracking servers.

Community
Intermediate
JingyiWngJingyiWng

querying-mlflow-metrics

Fetch aggregated MLflow trace metrics for token counts, latency, and assessment scores.

Community
Intermediate