| Project |
A container for jobs, notebooks and models. Has a name and an owner |
| Job |
Python code that reads, transforms and writes data. Runs on a runtime, on demand or on a schedule |
| Sub-job |
A job from the same project that runs together with a main job |
| Task |
A scheduled run of a job in the Scheduler |
| Run |
One execution of a job. Its status is Running, Success, Failed or aborted |
| Notebook |
An interactive document of code and Markdown cells that runs on a kernel |
| Session |
A running notebook. You start and stop it |
| Kernel |
The Python process in a session that runs your cells |
| Checkpoint |
A restore point of a notebook that you can go back to |
| Dataset |
A saved SQL or code query whose result is shown in pages and used by charts and dashboards |
| Standard runtime |
The dataset option that runs one SQL query directly on a source, with no compute to size |
| Runtime (RTE) |
The environment that runs your code, such as a Spark or DuckDB runtime with its libraries |
| Compute |
The memory and processor size given to a runtime |
| Minimum Workers |
The number of Spark workers to keep. Spark runtimes only |
| GPU Size |
The number of GPUs for a runtime |
| Custom RTE |
A runtime an administrator builds from a base runtime and a list of extra packages |
| Connection |
A saved link to an outside system. Code uses it by name and never holds credentials |
| System variable |
A placeholder such as ${JOBNM} that Groundzero fills in before your code runs |
| Published (exported) item |
A copy of a job, notebook or model shared with everyone on the site |
| MLflow |
The experiment tracking and model registry behind the Models tab |
| MCP |
Model Context Protocol, the standard through which assistants work with Groundzero |
