mlflow_integration¶

MLflow integration helpers for AdalFlow tracing.

set_trace_processors(processors)[source]¶: Set the global trace processors.

enable_mlflow_local(tracking_uri: str = None, experiment_name: str = 'AdalFlow-Agent-Experiment', project_name: str = 'AdalFlow-Agent-Project', port: int = 8080) → bool[source]¶

Enable MLflow local tracing without auto-starting server.

This function sets up MLflow tracing but does NOT automatically start an MLflow server. You need to have an MLflow server already running.

To start an MLflow server: `bash mlflow server --host 127.0.0.1 --port 8080 `

Parameters:

tracking_uri – MLflow tracking server URI. If None, defaults to http://localhost:{port}
experiment_name – Name of the MLflow experiment to create/use
project_name – Project name for the MLflow tracing processor
port – Port for the default tracking URI if tracking_uri is None

Returns:

True if MLflow was successfully enabled, False otherwise

Return type:

bool

Example

>>> from adalflow.tracing import enable_mlflow_local
>>> # Use with existing MLflow server on default port
>>> enable_mlflow_local()
>>> # Or with custom tracking URI:
>>> enable_mlflow_local(tracking_uri="http://localhost:5000")

enable_mlflow_local_with_server(tracking_uri: str = None, experiment_name: str = 'AdalFlow-Agent-Experiment', project_name: str = 'AdalFlow-Agent-Project', use_adalflow_dir: bool = True, port: int = 8080, kill_existing_server: bool = False) → bool[source]¶

Enable MLflow local tracing with auto-starting server.

This function sets up MLflow tracing and automatically starts an MLflow server if needed.

Parameters:

tracking_uri – MLflow tracking server URI. If None, will auto-start server
experiment_name – Name of the MLflow experiment to create/use
project_name – Project name for the MLflow tracing processor
use_adalflow_dir – If True and server fails to start, use ~/.adalflow/mlruns as fallback
port – Port to run MLflow server on
kill_existing_server – If True, kill any existing MLflow server on the port before starting

Returns:

True if MLflow was successfully enabled, False otherwise

Return type:

bool

Example

>>> from adalflow.tracing import enable_mlflow_local_with_server
>>> # Auto-start server on port 8080
>>> enable_mlflow_local_with_server()
>>> # Or kill existing and start fresh:
>>> enable_mlflow_local_with_server(kill_existing_server=True)
>>> # Or use custom port:
>>> enable_mlflow_local_with_server(port=5000)

get_mlflow_server_command(host: str = '0.0.0.0', port: int = 8080) → str[source]¶

Get the MLflow server command with AdalFlow backend store.

Parameters:

host – Host to bind the server to (default: “0.0.0.0”)
port – Port to run the server on (default: 8080)

Returns:

The MLflow server command to run

Return type:

str

Example

>>> from adalflow.tracing import get_mlflow_server_command
>>> print(get_mlflow_server_command())
mlflow server --backend-store-uri file:///Users/yourname/.adalflow/mlruns --host 0.0.0.0 --port 8080

start_mlflow_server(host: str = '0.0.0.0', port: int = 8080, wait: bool = True, kill_existing: bool = False) → bool[source]¶

Start MLflow server with AdalFlow backend store.

Parameters:

host – Host to bind the server to (default: “0.0.0.0”)
port – Port to run the server on (default: 8080)
wait – If True, wait for server to be ready before returning
kill_existing – If True, kill any existing process on the port before starting

Returns:

True if server started successfully, False otherwise

Return type:

bool

Example

>>> from adalflow.tracing import start_mlflow_server
>>> start_mlflow_server()  # Starts server on http://0.0.0.0:8080
>>> # Or with custom settings:
>>> start_mlflow_server(port=5000)