Introduction¶
WorldForge is a harness framework for building world-model-based workflows for physical AI. It is the application builder's counterpart to model-training stacks like Stable World Model: it helps roboticists and physical-AI builders compose, evaluate, and benchmark workflows built on top of world models, rather than train them.
The whole framework is organized around one backbone loop: planning and scoring action candidates with an action-conditioned predictive world model, in latent space.
- a
policyprovider proposes candidate action chunks from an observation - a
predictprovider rolls candidates forward as action-conditioned dynamics - a
scoreprovider ranks candidates as a cost oracle LatentMPCControllerowns the caller-side CEM/receding-horizon optimizer that ties them together- evaluation and benchmarking compare configurations so a builder can select the best one
The provider boundary is strict and fail-closed, which keeps semantics honest: a JEPA cost model is not treated as a video generator, and a VLA robot policy is not treated as a predictive dynamics model. The world model is a dynamics/cost oracle, not a controller — the controller stays a pure optimizer.
WorldForge is for Python developers and roboticists composing provider-backed planning loops, evaluation harnesses, and benchmarks. It is not a hosted control plane, a world generator, or a training framework, and it does not own checkpoints, robot runtimes, production telemetry, or durable multi-writer persistence.
Documentation Map¶
For the complete reader-path map, use Documentation Map.
| Need | Start here |
|---|---|
| Install, create a world, and run the mock provider | Quick Start |
| Find an exact CLI command or optional runtime smoke | CLI Reference |
| Understand capability names and provider boundaries | World Model Taxonomy |
| See module responsibilities and planning pipelines | Architecture |
| Record events, worlds, plans, and benchmarks in Rerun | Rerun Integration |
| Validate a checkout, diagnose providers, run smokes, or prepare a release | User And Operator Playbooks |
| Add or promote a provider adapter | Provider Authoring Guide |
| Check package, testing, typing, and ML-runtime quality rules | Engineering Quality |
If a page changes public behavior, update the linked operational page and the changelog in the same branch. WorldForge docs are treated as part of the package contract, not as post-release notes.