retrieving-mlflow-traces

Name: retrieving-mlflow-traces
Availability: InStock
Author: aaronachermann

Retrieves MLflow trace records by ID or using flexible search filters via CLI and Python API.

Updated Apr 18, 2026

Search Tags:#performance #debugging #python #cli #search #tracing #mlflow

Ready to run

npx skills add https://github.com/aaronachermann/PolentaEncoders --skill retrieving-mlflow-traces-aaronachermann

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: retrieving-mlflow-traces
Source: https://github.com/aaronachermann/PolentaEncoders/tree/main/.github/skills/retrieving-mlflow-traces
Command: npx skills add https://github.com/aaronachermann/PolentaEncoders --skill retrieving-mlflow-traces-aaronachermann

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Helps engineers and analysts find and retrieve MLflow trace records for debugging, performance analysis, and investigation when only partial identifiers or behavioral criteria are available.

Core Features & Use Cases

Search traces by run ID, experiment ID, session, user, status, tags, metadata, time range, execution duration, or free-text match.
Retrieve a single trace by trace ID to inspect spans, attributes, and human assessments for root cause analysis.
Support pagination, output formatting (JSON/table), span detail inclusion, and ordering to investigate slow traces, failed runs, or session-level sequences.

Quick Start

Search experiment 1 for traces with status ERROR and retrieve full details for any matching trace IDs using the MLflow CLI or Python API.

Frequently Asked Questions about retrieving-mlflow-traces

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

FAQPage Schema

How do I search MLflow traces to debug failed runs?▼

You can search MLflow traces using filter expressions for status, execution_time_ms, tags, or metadata via the mlflow CLI or Python search_traces API to identify failed runs and inspect span-level details for root cause analysis.

Can I retrieve a single MLflow trace by trace ID?▼

Yes, you can retrieve a single MLflow trace by its trace ID to inspect spans, attributes, and human assessments. This allows detailed root cause analysis for specific slow traces or failed runs without searching the entire experiment history.

What filters can I use to find slow MLflow traces?▼

Filter slow MLflow traces by applying execution_time_ms conditions in your search expression. You can also combine filters for timestamp_ms, tags, metadata, span attributes, and assessment fields to narrow down performance bottlenecks across sessions and users.

How do I paginate MLflow trace search results?▼

Paginate MLflow trace search results by applying pagination parameters alongside filter expressions in either the mlflow CLI or Python search_traces API. This allows you to manage large result sets when investigating traces across multiple sessions, users, or experiments.

Does the MLflow Python API support filtering traces by span attributes?▼

Yes, the MLflow Python search_traces API supports filtering traces by span attributes, along with execution_time_ms, timestamp_ms, tags, metadata, and assessment fields. This enables precise querying for debugging and performance analysis.

What is the best way to format MLflow trace search output for debugging?▼

Format MLflow trace search output using JSON or table formatting options available in the mlflow CLI and Python API. Including span details in the output allows you to inspect span-level attributes and assessments effectively for root cause analysis.