Back to Tool
D

Decision Forge

Tool Nov 3, 2026

An alpha Python pipeline that compiles organizational policy into reviewed rules, reproducible scenarios and decision-model training data.

Visit primary link 3 sources attached
Decision Forge project diagram shows organization sources, compilation and business review, scenario distributions, policy replay and train, development and test exports.
Juneo Lab's diagram of the Decision Forge data pipeline. It illustrates the project's architecture and review stages, not measured downstream agent performance. Source: Juneo Lab / Decision Forge contributors
agent-ecosystemdecision-modelsverification

What happened lately

Decision Forge timeline

Decision Forge is an Apache-2.0 Python command-line application for preparing organizational decision training data. Its four decision surfaces cover allocating work, sharing information, selecting actions and evaluating outcomes. Reviewed policy rules supply deterministic labels; model-assisted extraction and synthesis are separate, explicitly configured paths.

Version 0.3.2 adds persistent family split locks across dataset versions and broadens conflict review beyond seed states. The project remains alpha: reproducible dataset preparation does not establish improved downstream model judgment. Python 3.11 or newer and uv are documented for the source-install quickstart, with Linux and macOS in its CI matrix. The offline build path needs no model credentials; document compilation sends requests to a selected model endpoint using a provider key.