Architecture

This page explains the main code paths in terms of responsibilities rather than individual source files.

High-level flow

The active project flow looks like this:

nextflow working directory
  -> run discovery
  -> run selection
  -> analysis/log staging
  -> desktop-style metadata synthesis
  -> manifest creation
  -> .2me tarball
  -> import

Main areas of the codebase

CLI entry points

src/create_2me/create_from_cli_run.rs and src/importer/import_from_2me.rs define the active command-line flows.

Nextflow capture

src/nextflow/nextflow_toolkit.rs indexes historical runs with nextflow log and orchestrates CLI-run packaging.

Analysis staging

src/nextflow/nextflow_analysis.rs resolves output directories, finds matching logs, distills nextflow.stdout, and synthesizes progress.json.

Metadata extraction

src/nextflow_log_parser.rs parses the reduced Nextflow transcript into workflow identity fields such as project, repository, revision, and version.

Desktop analysis model

src/epi2me_desktop_analysis.rs defines the EPI2ME-style analysis record that is serialized into the archive payload.

Workflow payload model

src/epi2me_workflow.rs inventories installed workflow files for packaging and import.

Manifest and archive semantics

src/xmanifest.rs defines the portable archive structure, provenance, and manifest verification logic.

Design intent

The recurring design theme is translation.

epi4you is not trying to replace Nextflow or EPI2ME Desktop. Instead it translates between:

• raw CLI-oriented run artifacts,

• EPI2ME-style metadata expectations, and

• a portable archive form suitable for transfer.

This translation is why the codebase contains both low-level filesystem work and higher-level domain models such as manifests and desktop analyses.

Relationship to the broader repository

The repository contains additional code for workflows, containers, and database operations that reflects the wider original ambition of the project.

Even where those paths are not the current main CLI entry points, they still matter architecturally because they explain why the manifest supports multiple payload types and why the project thinks in terms of “bioinformatics assets” rather than only “analysis result folders”.