miniAB

miniAB is a lightweight A/B testing and experimentation framework designed to help developers run small-scale experiments in web and mobile applications. It emphasizes minimal setup, predictable performance, and a straightforward workflow for defining experiments and interpreting results. The project targets teams that want to validate ideas quickly without deploying large instrumentation or complex analytics pipelines.

Key features include client-side and server-side experiment execution, deterministic traffic allocation based on user identifiers, support for feature flags, versioned experiments, built-in analytics adapters, and a pluggable data store. It uses JSON or YAML for experiment definitions and provides a simple API for integration with existing logging and metrics systems. The framework is designed to be unobtrusive, with a small runtime footprint and easy onboarding for engineers who are new to experimentation.

Architecture of miniAB is composed of an experiment designer, a runtime engine, and an adapter layer. The designer stores experiment definitions; the engine assigns users to variants and records outcomes; the adapters push telemetry to external analytics services. It can be deployed as a library within applications or as a standalone service, prioritizing portability and a clear separation between definition, execution, and data reporting.

Usage commonly involves creating experiments, instrumenting events, and viewing results through built-in dashboards or compatible external BI tools. miniAB supports use cases such as feature rollouts, UI experiments, and content variants, particularly where teams require quick iteration and low operational overhead.

History and reception: miniAB originated as an open-source project to address the need for lightweight experimentation tooling. It has attracted contributions from volunteers and is described in community discussions as easy to adopt but with fewer enterprise features than larger platforms. Documentation emphasizes straightforward setup, privacy-conscious sampling, and local testing options.