FAST-HEP separates the scientific description of an analysis from the implementations and computing infrastructure used to execute it.
An analysis begins with a workflow description. Flow validates and normalises that description, resolves registered capabilities and dependencies, constructs a logical graph, and produces an execution plan before runtime execution begins.
The following pages describe the main concepts behind this process.
How analyses are described in workflow.yaml, including datasets, sources,
analysis operations, outputs, and rendering.
How Flow turns a workflow description into a logical graph and execution plan, and how that plan separates scientific workflow semantics from runtime execution.
How workflow operations expose their parameters, inputs, outputs, dependencies, and runtime behaviour through stable contracts.
How capabilities are discovered and selected, and how profiles compose registries for different analysis and execution contexts.
How the same execution plan can be mapped onto different backends, computing resources, and provisioning strategies.
How complete analyses can be organised into reproducible repositories containing workflows, configuration, custom code, data references, and supporting material.