sync_to_mlflow

Sync a training run to a local or remote MLflow tracking server.

Overview

sync_to_mlflow reads the metrics, parameters, and artifacts from a Fovux training run and creates a corresponding MLflow run entry. Requires an accessible MLflow tracking server.

Input Schema

Parameter	Type	Required	Default	Description
run_id	string	Yes	—	Fovux run ID to sync.
tracking_uri	string	No	"http://localhost:5000"	MLflow tracking server URI.
experiment_name	string	No	"fovux"	MLflow experiment name.

Output Schema

Field	Type	Description
run_id	string	Fovux run ID.
mlflow_run_id	string	Created MLflow run ID.
tracking_uri	string	MLflow server URI used.
experiment_name	string	MLflow experiment name.
synced_metrics	integer	Number of metric entries synced.

Examples

CLI

curl -X POST http://127.0.0.1:7823/tools/sync_to_mlflow \
  -H "Authorization: Bearer $(cat ~/.fovux/auth.token)" \
  -H "Content-Type: application/json" \
  -d '{"run_id": "abc123", "tracking_uri": "http://localhost:5000"}'

Python

from fovux.tools.sync_to_mlflow import sync_to_mlflow
result = sync_to_mlflow("abc123", tracking_uri="http://localhost:5000")

Notes & Limits

• Requires the mlflow Python package to be installed.
• Network connectivity to the MLflow server is required.
• This tool is the only Fovux tool that makes external network requests.

Failure Modes

• ImportError if mlflow is not installed.
• Connection errors if the MLflow server is unreachable.
• FovuxRunNotFoundError if the run ID is not in the registry.