GrowthBook Python SDK

Warnings

• breaking Version 2.0.0 removed async tracking callback executions, enforcing synchronous execution for reliability. If you previously relied on async callbacks, you will need to adjust your implementation to handle tracking synchronously or implement your own async wrapper.
  Fix: Migrate `on_experiment_viewed` or `set_event_logger` callbacks to be synchronous, or wrap them with your own asynchronous handling mechanism if needed.
• gotcha If 'Secure Attributes' are enabled in your GrowthBook SDK Connection, you must manually hash any `secureString` or `secureString[]` attributes using SHA-256 with your organization's secure attribute salt before passing them to the SDK.
  Fix: Hash sensitive attributes using `hashlib.sha256(attribute_value + secure_salt).hexdigest()` before including them in the `attributes` dictionary.
• gotcha To analyze experiment results in the GrowthBook platform, you must provide an `on_experiment_viewed` callback (for `GrowthBook` class) or configure an event logger via `set_event_logger` (for `GrowthBook` or `GrowthBookClient` from v2.2.0+). Without this callback, feature flags will work, but experiment assignments will not be tracked.
  Fix: Implement and pass an `on_experiment_viewed` callable to the `GrowthBook` constructor or use `gb.set_event_logger` to record experiment view events to your analytics system.
• gotcha For improved performance and better resource utilization, especially in async web applications (e.g., FastAPI, Django Async), the `GrowthBookClient` class is recommended over the synchronous `GrowthBook` class. It's designed to reuse a single client instance.
  Fix: Use `GrowthBookClient` and its `async with` context manager for async applications, and `await gb.load_features()` for fetching data.

Install

• pip install growthbook Install with pip

Imports

• GrowthBook

  from growthbook import GrowthBook

• GrowthBookClient

  from growthbook import GrowthBookClient

  Recommended for async applications for improved performance and resource utilization.

Quickstart

This quickstart demonstrates how to initialize and use the `GrowthBookClient` for asynchronous applications, fetch feature flags, check feature status, retrieve feature values, and run A/B experiments. It uses environment variables for `client_key` and `api_host` for secure and flexible configuration. The `GrowthBookClient` is recommended for performance and resource utilization in async environments.

import asyncio
import os
from growthbook import GrowthBookClient, Options, UserContext, FeatureRefreshStrategy

async def main():
    # Get client key from environment variable for security
    client_key = os.environ.get("GROWTHBOOK_CLIENT_KEY", "sdk-YOUR_CLIENT_KEY")
    api_host = os.environ.get("GROWTHBOOK_API_HOST", "https://cdn.growthbook.io")

    # User attributes for targeting and experimentation
    attributes = {
        "id": "user-123",
        "country": "US",
        "loggedIn": True
    }

    # Configure GrowthBook client options
    options = Options(
        api_host=api_host,
        client_key=client_key,
        # Optional: Enable real-time feature updates via Server-Sent Events (SSE)
        refresh_strategy=FeatureRefreshStrategy.SERVER_SENT_EVENTS
    )

    # Create a UserContext instance
    user_context = UserContext(attributes=attributes)

    # Create an async GrowthBookClient instance (recommended for async apps)
    async with GrowthBookClient(options=options, user_context=user_context) as gb:
        # Load features from the GrowthBook API
        await gb.load_features()

        # Simple on/off feature gating
        if gb.is_on("my-awesome-feature"): # Replace with your feature key
            print("My awesome feature is ON!")
        else:
            print("My awesome feature is OFF.")

        # Get the value of a feature with a fallback
        button_color = gb.get_feature_value("button-color-feature", "blue") # Replace with your feature key
        print(f"Button color feature value: {button_color}")

        # Example for an experiment
        # Make sure 'pricing-experiment' and variations are configured in GrowthBook
        pricing_result = gb.run(
            {
                "key": "pricing-experiment",
                "variations": ["control", "variant-a", "variant-b"]
            }
        )

        if pricing_result.in_experiment:
            print(f"User in pricing experiment, assigned variation: {pricing_result.value}")
        else:
            print("User not in pricing experiment.")

    print("GrowthBook client closed.")

if __name__ == "__main__":
    # Set dummy keys for local testing, replace with actual keys or environment variables
    os.environ["GROWTHBOOK_CLIENT_KEY"] = "sdk-example12345"
    os.environ["GROWTHBOOK_API_HOST"] = "https://cdn.growthbook.io"
    asyncio.run(main())