retrieving-mlflow-traces

Retrieve MLflow trace records and span details by trace ID or filtered search.

Updated May 20, 2026

Search Tags:#performance #debugging #tracing #mlflow #metadata filters #cli search #span inspection

Ready to run

npx skills add https://github.com/ice-droid-99/BI-Dashboard-Generation-Agent-in-Databricks- --skill retrieving-mlflow-traces-ice-droid-99

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: retrieving-mlflow-traces
Source: https://github.com/ice-droid-99/BI-Dashboard-Generation-Agent-in-Databricks-/tree/main/.gemini/skills/retrieving-mlflow-traces
Command: npx skills add https://github.com/ice-droid-99/BI-Dashboard-Generation-Agent-in-Databricks- --skill retrieving-mlflow-traces-ice-droid-99

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Retrieves and inspects MLflow tracing data when you need to identify what happened in a model/session execution, especially for failures.

Core Features & Use Cases

Single-trace retrieval: Fetch full details by trace ID to inspect status, timing, tags/metadata, and spans.
Trace searching with precise filtering: Find traces by status, tags, metadata fields, session/user identifiers, execution time, and time ranges.
Debugging and performance triage: Locate slow or failed traces and narrow down root-cause investigation using span tree details and assessments.

Use case examples:

Find all ERROR traces for a given session to pinpoint the failing step.
Search for traces slower than a threshold to identify performance regressions.
Retrieve a specific trace by ID from UI logs and review spans and metadata for root-cause analysis.

Quick Start

Tell the AI to search MLflow traces for experiment 1 where trace.status equals ERROR and return the matching trace IDs and key metadata.

Frequently Asked Questions about retrieving-mlflow-traces

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

FAQPage Schema

How do I retrieve MLflow traces for a specific experiment to debug failed executions?▼

You can retrieve MLflow traces by using the MLflow CLI trace search and get commands with filter-string queries. Filter by status, tags, or execution time to locate failed executions and inspect span details for root-cause analysis.

Can I search MLflow traces using metadata filters and time range constraints?▼

Yes, you can search MLflow traces using metadata filters and time range constraints. Construct filter-string queries consistent with MLflow tracing fields to search by status, tags, session, user, and execution time to identify performance regressions.

What is the best way to inspect span details for a single MLflow trace?▼

The best way to inspect span details is to fetch a single MLflow trace by its trace ID using the CLI get command. This retrieves full trace details, including span tree information, status, timing, and metadata for root-cause analysis.

Does the MLflow CLI trace search support pagination and span-inclusive output?▼

Yes, the MLflow CLI trace search supports pagination and optional span-inclusive output. You can manage large sets of trace records by paginating results and choosing whether to include span details in the returned output.

How do I find slow MLflow traces to identify performance regressions?▼

To find slow MLflow traces and identify performance regressions, search for traces using execution-time constraints. Apply filter-string queries with execution-time thresholds to locate traces slower than your specified limit and narrow down root causes.