DataRobot MLOps

11.1.28 verified Fri May 01 auth: no python

Python SDK for DataRobot MLOps to read and report model monitoring statistics (accuracy, drift, data quality) to DataRobot's MLOps platform. Current version 11.1.28, requires Python >=3.9. Released as part of DataRobot's MLOps cloud or on-premise deployment with frequent minor releases.

pip install datarobot-mlops

Common errors

error AttributeError: module 'datarobot' has no attribute 'mlops' ↓

cause Importing as 'datarobot.mlops' instead of 'datarobot_mlops'.

fix

Use correct import: 'from datarobot_mlops import MLOpsReporting'

error TypeError: MLOpsReporting.__init__() got an unexpected keyword argument 'application_key' ↓

cause datarobot-mlops 11.x removed application_key; use api_token.

fix

Replace application_key='...' with api_token='...'.

error datarobot_mlops.exceptions.MLOpsException: (403) Forbidden ↓

cause Invalid or expired API token, or insufficient permissions on the deployment.

fix

Verify API token and permissions; check server_url and token environment variables.

Warnings

breaking In version 11.x, MLOpsReporting no longer accepts 'application_key' argument; use 'api_token' instead. ↓

fix Replace 'application_key' with 'api_token' in constructor.

deprecated The method 'report_binary_classification_stats' is deprecated in 10.x+ and removed in 11.x. ↓

fix Use 'report_classification_stats' with 'class_names' parameter instead.

gotcha The SDK uses blocking HTTP calls. In production, run reporting in a separate thread or async task to avoid blocking model inference. ↓

fix Wrap report calls with threading or use an async HTTP client if available (not natively supported).

gotcha Deployment ID must be a string; integer or UUID object will raise AttributeError. ↓

fix Always pass deployment_id as str(deployment_id).

Imports

MLOpsReporting
wrong
```
from datarobot.mlops import MLOpsReporting
```
correct
```
from datarobot_mlops import MLOpsReporting
```
Old pattern used dot-based submodule; correct import uses underscore

mlops

wrong

import datarobot.mlops as mlops

correct

import datarobot_mlops as mlops

Common mistake due to PyPI name

Quickstart

Basic usage to report model predictions and execution time to a DataRobot MLOps deployment.

import os
from datarobot_mlops import MLOpsReporting

# Initialize with API credentials from environment
mlops_report = MLOpsReporting(
    server_url=os.environ.get('DATAROBOT_API_ENDPOINT', ''),
    api_token=os.environ.get('DATAROBOT_API_TOKEN', '')
)

# Report deployment stats
mlops_report.report_deployment_stats(
    deployment_id='your-deployment-id',
    predictions=100,
    execution_time=0.5,
    model_id='your-model-id'
)
print("Stats reported successfully.")