Concepts
How DataMaker workspaces, reusable assets, runs, and generated files fit together.
A team holds membership and permissions. A project groups the assets you work with. Select the correct team and project before creating resources or running a workflow; an ID from another project does not grant access to it.
Reusable definitions
| Object | Purpose |
|---|---|
| Template | Field definitions and generator options used to produce data. |
| Custom data type | A reusable field configuration, including weighted value libraries. |
| Connection | A database target or source. |
| Endpoint | A saved REST or OData request, with its method, URL and options. |
| Integration | A business system and its authentication, shared by related endpoints. |
| Scenario | Python code, requirements, environment variables and workspace files for a repeatable workflow. |
| Plan | A reviewable specification: entities, tasks, flow validation or a source-to-target mapping. Different plan kinds have different execution behavior. |
| Masking policy | Explicit per-field rules for changing sensitive source values. |
| Skill | Instructions the agent can use to perform a recurring task. |
| Hub pack | Portable templates, skills, plans, data types, value libraries and scenarios. |
Data and results
A set stores a small reusable collection of rows inline. A dataset stores chunked data suitable for larger processing. A data job extracts, transforms, masks or loads a dataset and records progress. A key map records relationships such as source IDs and the target IDs created by a load.
My data lists your generated chat files. It is separate from the project's sets and datasets. Scenario files are inputs and outputs attached to scenario workspaces.
Choose a workflow
- For new synthetic rows, start with a template.
- For a one-off request, use chat.
- For reviewed migration rules, use a mapping plan and data jobs.
- For custom orchestration or test automation, use a scenario.
A generated response is not automatically saved as a reusable set or loaded into another system. Review and save or deliver it explicitly. A sensitive-field label is also not a masking policy: see Sensitive fields.