Architecture

SciWIn is five repositories that only make sense together. This page draws the two relationships that come up most often: how the pieces fit together, and how a tool gets a container without you writing a Dockerfile.

Component hierarchy

SciWIn-Client (s4n) and SciWIn-Studio are two front ends over the same shared core. Both operate on the exact same on-disk project, a workflow.toml, your CWL files, and a git repository, so a project started in one opens cleanly in the other.

flowchart TD
  Studio["SciWIn-Studio<br/>Tauri + Svelte GUI"]
  CLI["s4n<br/>command-line interface"]
  Core["sciwin<br/>shared core: authoring · execution · provenance"]
  CWL["commonwl<br/>CWL parsing + execution engine"]
  RO["rocrate<br/>RO-Crate read/write"]
  REANA["reana<br/>REANA client"]

  Studio --> Core
  CLI --> Core
  Core --> CWL
  Core --> RO
  Core --> REANA

• sciwin is the shared core: authoring CWL CommandLineTool/Workflow files, running them, and building provenance graphs. Neither front end duplicates this logic.
• commonwl parses CWL documents and provides the execution engine (local, Docker, and TES backends) behind s4n execute.
• rocrate reads, writes, and validates RO-Crates, including the Provenance Run Crates s4n execute run --rocrate produces — see Provenance & Publishing for what those actually contain.
• reana is the client used when execution targets a remote REANA server instead of running locally. A compatibility layer for tools that only carry a Dockerfile exists in the sciwin crate — see Provenance & Publishing — but isn’t wired into execution yet.

Automatic container resolution

For a Python or R tool, s4n create --auto-container can skip writing a Dockerfile entirely. It reads the project’s own dependency file, resolves those dependencies against a FAIRagro-run registry, and attaches whichever pre-built image satisfies them, picking the smallest match.

flowchart LR
  Deps["requirements.txt /<br/>pyproject.toml / DESCRIPTION"]
  Create["s4n create<br/>--auto-container"]
  Registry["sciwin-container-registry<br/>package → image digest index"]
  Containers["sciwin-containers<br/>built images"]

  Deps --> Create --> Registry --> Containers

Two repositories do the real work here: sciwin-container-registry is the index mapping package requirements to image digests, and sciwin-containers is where those images are actually built. The resolved image is written straight into the generated CWL tool’s DockerRequirement, no Dockerfile authoring required.