An MLflow record of a whole execution, arranged as a tree of spans and capturing inputs, outputs, steps along the way, latency and errors. Scorers use it in evaluation and monitoring.
Read more: Databricks docs
In the Ultra Transcenders books
Each book explains Trace in context, with comparison tables and the common traps.
Terms in this definition
- MLflow
An open-source framework for experiment tracking and model management. Azure Machine Learning adopts it as its tracking API, so each workspace acts as a compatible MLflow tracking server.
Related terms
- Foundry Control Plane
Lets organisations govern, monitor and trace their agents from one central place in Microsoft Foundry. Agents built outside Foundry can be included after being registered via an AI gateway into a project that is wired to Application Insights.
- Lineage view
Diagram of a workspace showing how its items feed one another, from sources through to reports, making it easy to trace data origins and dependants. Impact analysis begins here.
- make_judge
Build your own LLM-based judge in MLflow with this function. The instructions you give it have to mention
inputs,outputs,expectationsortrace, as template variables. - operation_Id
An Application Insights field naming the root operation, shared by every telemetry item belonging to one distributed trace. It corresponds to trace-id in W3C Trace Context.
- PF_DISABLE_TRACING
Environment variable governing tracing in prompt flow, which is disabled unless changed. Setting it to false reveals the Trace tab, showing duration and token cost for each node.
- Production monitoring
Scores live agent traffic while in Beta: you choose what share of traces to sample, MLflow 3 scorers examine them, and every result lands on its trace as feedback.
- purple screen
What an ESX host displays after an unrecoverable VMkernel error. It lists details such as the message, stack trace, CPU registers, current world and uptime; support will want the message and the stack.
- RAISERROR
A legacy way to signal an error in T-SQL, with a chosen severity and printf-style placeholders; severities from 19 upwards also need WITH LOG and either sysadmin or ALTER TRACE. Microsoft advises THROW for anything new.