Which interface should I use to run the pipeline?¶

Use run_together. It is the recommended interface: one command runs the whole pipeline (SGE conversion → FICTURE → cell analyses → images → packaging → optional publishing) for any supported platform, for one sample or many, as a single resumable Makefile.

The individual commands remain available when you need control that run_together does not expose.

Interface	What it is	Best for
`run_together` (recommended)	One command for the whole pipeline, with per-platform defaults, sample sheets, and a JSON config for full control	Almost every run: single samples, cohorts with a joint model, batches of independent samples
Individual commands	Call `sge_convert`, `run_ficture2`, `run_cartload2`, `import_`, `upload_` yourself	Custom pipelines, a parameter `run_together` does not expose, debugging one stage, tight HPC integration
`run_xenium` / `run_visiumhd` (legacy)	Older per-platform orchestrators	Existing scripts only; see Legacy Orchestrators

Quick decision rule¶

Running a supported platform, one sample or many? → run_together. Start with Get Started, then the single-sample or multi-sample guide.
A platform without a built-in profile, or a custom format? → still run_together, with --platform generic (see the Generic platform page); drop to the individual commands only if that is not enough.
Need to customize or debug one stage? → the individual commands, introduced in Step-by-step.

They interoperate

run_together calls the same commands and produces the same catalog.yaml + PMTiles. You can prototype with run_together, then rerun a single stage by hand.