Versioning & Changelog

The MMM Framework is at version 0.1.0 — a pre-1.0 release. The mathematical foundations and core API are intentionally stable, but the surface area is still evolving. This page tracks where we are and what to expect.

Pre-1.0: read before depending

Public API may have backwards-incompatible changes between minor versions until 1.0. Pin a specific version (mmm-framework==0.1.0) for production use, and review the release notes before upgrading. Once 1.0 ships, the framework will follow semantic versioning strictly.

API Stability by Module

Different parts of the framework have matured at different rates. Use this as a guide for what to depend on:

Stable

Core modeling

  • BayesianMMM
  • MMMResults / PredictionResults
  • Builders (ModelConfigBuilder, etc.)
  • config (Pydantic dataclasses)
  • Adstock, saturation, seasonality, trend transforms
  • MFFLoader / data loading
Beta

Active surface

  • reporting (charts, extractors, generator)
  • analysis (counterfactual, marginal)
  • serialization (MMMSerializer)
  • FastAPI endpoints (api/)
  • React frontend (frontend/)
Experimental

Subject to change

  • mmm_extensions (NestedMMM, MultivariateMMM, CombinedMMM)
  • dag_model_builder
  • Excel config workflow
  • Streamlit UI (legacy)

Versioning Policy

Until 1.0:

After 1.0, the project will follow SemVer strictly: patches are bug fixes only, minors add backward-compatible features, and majors are reserved for breaking changes.

Release Notes

For ongoing release tracking, see the GitHub Releases page. The history below summarises the current version.

0.1.0 initial release Current

First public release of the MMM Framework. Includes:

  • Bayesian MMM core (BayesianMMM) extending PyMC-Marketing with hierarchical specs, configurable adstock/saturation, and reproducible model serialization
  • Fluent builders for model, channel, and variable configuration
  • MFF (Master Flat File) data loader with validation
  • Counterfactual and marginal analysis utilities
  • HTML report generator with Plotly charts and design-token theming
  • Extension models (NestedMMM, MultivariateMMM, CombinedMMM) — experimental
  • FastAPI service + ARQ worker for async job execution
  • React (TypeScript) frontend; legacy Streamlit UI retained
  • Sphinx-generated API reference + this static documentation site

Contributing

Bug reports, feature requests, and pull requests are welcome on GitHub. Please open an issue before submitting a substantial PR so we can discuss scope.

For questions about the methodology, the FAQ is a good starting point; for the underlying math, the Bayesian Workflow and Causal Inference guides cover most of the foundations.