ABMeter Python SDK

ABMeter is a feature-flag and A/B-testing platform. You define parameters, experiments, and feature flags in the ABMeter Lab; this package reads the value assigned to each user in your Python app and reports events back.

Supported Python versions

abmeter supports Python 3.11 and newer (tested against 3.11, 3.12, 3.13).

Installation

pip install abmeter

Getting an API key

Sign up at abmeter.ai, then in the Lab open API Keys and create one. Pass it to configure (e.g. from an environment variable).

Usage

Configure the client once at startup:

import abmeter

abmeter.configure(api_key="your-api-key")

The example below assumes a parameter welcome_text (in a space, with a variant, controlled by a running experiment or feature flag) and an event type user_purchases_plan already exist — create them through the ABMeter API or, more easily, your AI assistant over MCP. Until they do, resolve_parameter just returns the parameter's default and track_event rejects unknown event types.

# In request handling: read the value assigned to this user. `user_id` is the
# only required field; an optional `email=` is available for predicate-based
# (email-pattern) audience targeting.
user = abmeter.User(user_id=current_user.id)
text = abmeter.resolve_parameter(user, "welcome_text")

# In business logic: record an action that feeds a metric.
abmeter.track_event(
    "user_purchases_plan",
    current_user.id,
    {"plan": purchased_plan.name, "price": purchased_plan.price},
)

The two calls are one loop: the experiment varies welcome_text, and the event feeds a metric whose effect the experiment measures across variants. Note that resolve_parameter is not a pure read — it fetches the assignment config, resolves the variant client-side (no per-call round-trip), and records an exposure in the background; use get_exposure for the same resolution without submitting (useful for debugging). Full guides live at abmeter.ai.

What this SDK is

A thin client: it fetches the assignment config, resolves parameters locally, and submits exposures/events. It deliberately does not contain the platform-side concerns of the Ruby gem (config serialization, allocation range-building, type-validation registry). Those live in the ABMeter API.

Development

The package uses uv. It is a pure client library, so the unit suite needs no database or any other running service.

uv sync --extra dev   # install dependencies
uv run pytest         # tests
uv run ruff check .   # lint
uv run mypy           # type check

License

Available as open source under the terms of the MIT License.