DataRobot Python Client

3.13.0 · active · verified Thu Apr 16

The DataRobot Python Client Library (v3.13.0) provides a programmatic interface for interacting with the DataRobot AI Platform, enabling users to manage projects, create models, perform predictions, and automate MLOps workflows. It follows a regular release cadence, typically monthly or bi-monthly, aligning with platform updates.

Common errors

```
datarobot.errors.ClientInitializationError: API token or endpoint not provided
```
cause The DataRobot client could not find the necessary API token or endpoint to establish a connection. This often happens if environment variables (`DATAROBOT_API_TOKEN`, `DATAROBOT_ENDPOINT`) are not set, or if they are not passed explicitly to `dr.Client()`.

fix Set the `DATAROBOT_API_TOKEN` and `DATAROBOT_ENDPOINT` environment variables. For example: `export DATAROBOT_API_TOKEN='your_token'` and `export DATAROBOT_ENDPOINT='https://app.datarobot.com/api/v2'`. Alternatively, initialize the client with `dr.Client(token='your_token', endpoint='your_endpoint')`.
```
requests.exceptions.SSLError: HTTPSConnectionPool(...) Max retries exceeded with url: (...) Caused by SSLError(CertificateError(...))
```
cause SSL certificate verification failed, typically due to a corporate proxy or firewall interfering with the TLS handshake. Python's `requests` library does not trust the custom certificates presented by the proxy.

fix Configure the `REQUESTS_CA_BUNDLE` environment variable to point to your corporate CA certificate bundle. For example: `export REQUESTS_CA_BUNDLE='/etc/ssl/certs/ca-certificates.crt'`. You may also need to set proxy environment variables (`HTTP_PROXY`, `HTTPS_PROXY`).
```
AttributeError: module 'datarobot' has no attribute 'Client'
```
cause This error usually indicates an outdated version of the `datarobot` library, a conflict with a local file named `datarobot.py`, or an incorrect import pattern.

fix Ensure you have the latest version of the `datarobot` library by running `pip install --upgrade datarobot`. Check your project directory for any files named `datarobot.py` that might shadow the installed package. Confirm you are using `import datarobot as dr` and then `dr.Client()`.
```
datarobot.errors.NotFound: No such project found with ID '...'
```
cause The specified project ID does not exist, is incorrect, or the authenticated user does not have permission to access it.

fix Verify the project ID is correct. Ensure the API token belongs to a user with appropriate permissions to view or interact with the project. You can list all projects to which you have access using `dr.Project.list()` to find correct IDs.

Warnings

breaking The `datarobot.Client` initialization changed significantly in version 3.0. Prior versions often required explicitly passing `endpoint` and `token` arguments to the `Client` constructor. Version 3.0+ defaults to auto-discovery from environment variables (DATAROBOT_ENDPOINT, DATAROBOT_API_TOKEN) or a configuration file.
Fix: For versions 3.0 and newer, set `DATAROBOT_ENDPOINT` and `DATAROBOT_API_TOKEN` environment variables. If you need to explicitly pass them, use `dr.Client(endpoint='...', token='...')`.
breaking Methods for interacting with the prediction API (e.g., `predict_raw`) may have changed signatures or availability across major versions, particularly regarding how data is passed and results are retrieved.
Fix: Consult the DataRobot Python Client documentation for your specific version, especially for `Project.get_recommended_model().predict()` or `Deployment.predict()`, to ensure correct method signatures and return types.
gotcha Connecting from corporate networks can often lead to `SSLError` due to proxy configurations or custom CA certificates. The Python `requests` library (used by DataRobot client) might not trust these certificates by default.
Fix: Set the `REQUESTS_CA_BUNDLE` environment variable to the path of your corporate CA certificate bundle or `CURL_CA_BUNDLE` for similar effect. Alternatively, you might need to configure proxy settings via `HTTP_PROXY` and `HTTPS_PROXY` environment variables. As a last resort (not recommended for production), `dr.Client(verify_ssl=False)` can temporarily bypass SSL verification.
gotcha When working with large datasets, uploading or downloading them directly through the client can be slow or hit memory limits. DataRobot provides specific methods for large data handling.
Fix: For large dataset uploads, consider using `dr.Dataset.create_from_file(filepath, max_wait=3600)` and ensure your client timeout is sufficient. For downloading predictions, use methods that allow for streaming or chunking, or retrieve predictions from a prediction server/deployment directly.

Install

pip install datarobot Install latest version

Imports

Client
wrong:
```
from datarobot import Client
```
correct:
```
import datarobot as dr
dr.Client()
```
While 'from datarobot import Client' works, the recommended and common practice is 'import datarobot as dr' and then using 'dr.Client()' to maintain consistency with other DataRobot objects like 'dr.Project' or 'dr.Model'.
Project
```
import datarobot as dr
dr.Project
```
Deployment
```
import datarobot as dr
dr.Deployment
```

Quickstart

This quickstart demonstrates how to initialize the DataRobot client, which by default reads configuration from environment variables (DATAROBOT_ENDPOINT, DATAROBOT_API_TOKEN). It then lists the first five projects accessible to the authenticated user. Ensure your environment variables are set for successful execution.

import datarobot as dr
import os

# Configure client using environment variables
# DATAROBOT_ENDPOINT e.g., 'https://app.datarobot.com/api/v2'
# DATAROBOT_API_TOKEN
# DATAROBOT_USERNAME (optional, for specific auth methods)
# DATAROBOT_PASSWORD (optional, for specific auth methods)

# For testing, ensure these are set in your environment or provide defaults.
endpoint = os.environ.get('DATAROBOT_ENDPOINT', 'YOUR_DATAROBOT_ENDPOINT')
token = os.environ.get('DATAROBOT_API_TOKEN', 'YOUR_DATAROBOT_API_TOKEN')

if 'YOUR_DATAROBOT_ENDPOINT' in endpoint or 'YOUR_DATAROBOT_API_TOKEN' in token:
    print("Please set DATAROBOT_ENDPOINT and DATAROBOT_API_TOKEN environment variables or replace placeholders.")
else:
    try:
        # Client auto-discovers config from environment variables by default.
        # Explicit config can be passed via dr.Client(token='...', endpoint='...').
        dr.Client()
        print("Successfully connected to DataRobot.")

        # List first 5 projects
        projects = dr.Project.list(max_projects=5)
        if projects:
            print("\nFirst 5 projects:")
            for p in projects:
                print(f" - {p.name} (ID: {p.id})")
        else:
            print("No projects found in your account.")

    except dr.errors.ClientInitializationError as e:
        print(f"Error initializing DataRobot client: {e}")
        print("Ensure DATAROBOT_ENDPOINT and DATAROBOT_API_TOKEN environment variables are correctly set.")
    except Exception as e:
        print(f"An unexpected error occurred: {e}")

view raw JSON →