Platform concepts
In this guide, the basic concepts of the Timefold Platform are explained.
High-level architecture
The diagram below shows how the different components of Timefold fit together to solve your planning problem.
-
Models: Each of the Timefold models solve a specific type of planning problem. You can use one or more of the expert-built models.
-
Model SDK: The Model SDK connects the models to the underlying supporting features.
-
Enterprise Solver: This is the Enterprise version of the Timefold Solver tasked to find the most optimal solution possible.
-
Timefold Cloud Platform: This is the infrastructure that coordinates solving planning problems and adds a powerful UI.
For more details on the architecture of the Timefold Platform itself, please refer to Installation instructions for more details.
Tenants
A platform installation can have one or more tenants. A single tenant can hold one or more models and can have multiple members. Tenants allow you to organize your resources so that the members of a tenant only see the models that are available to that tenant.
Members
A member is a user of the Timefold Platform. Members log in to the platform with one of the options (e.g. Google or Microsoft). A single user can be a member of one or more tenants.
If a user is a member of two or more tenants, they can switch between tenants via the tenant drop-down in the platform’s header.
See Member management and roles for more details.
Models
On the Dashboard of the Timefold Platform you see which models (and versions of models) are added to your tenant. You can manage available models via the Settings pages.
Each model has Documentation pages explaining what type of problems it can be used for and which features it supports. The documentation also includes API specifications as well as user guides for the model.
In the Runs tab of a model you can see how often planning problems were provided to the model and the solutions that were computed. Each run (with its KPIs and metrics) is shown in the overview.
See Model catalog and documentation for more details.
Model runs
A run of a model happens each time a user or process submits a planning problem to the Timefold Platform. This can be done in the UI (by uploading a planning problem or running a demo data set) or by posting to the API. All runs display in Timefold Platform’s UI.
For each model run, Timefold Platform shows:
-
Details about the run, its status, its constraint scores and when it started and was completed.
-
An overview of the KPIs associated with the computed planning solution.
-
A score graph showing the evolution of the scores during solving.
-
A detailed score analysis of the planning solution, with scores for each constraint as well as justification examples.
-
A visualization of the planning solution that can be used, for example, for spot checks.
-
Pages showing the input, output, and logs of the model run.
Each model run has a unique ID, and can also be given a custom name and have associated tags. Tags can be used to separate production runs from test runs, or to separate partitions of the planning problem.