The medium by which the developer can communicate capability with the PI
This section defines core concepts used throughout VoxKit so developers and researchers share a common vocabulary.
A low-level processing component that implements a specific algorithm or toolkit (for example: forced alignment, voice activity detection, feature extraction, or model inference). Engines load models/assets, expose a simple run API, and produce machine-readable artifacts (JSON, arrays, files). Engines are intended to be implemented by developers and documented so pipelines can reference them declaratively.
A higher-level component that analyzes datasets, generates reports, or computes summary statistics. Analyzers are implemented by developers and documented for pipeline use.
The orchestration layer that composes Engines and Analyzers into an ordered step with a GUI. The Stacker reads pipeline definitions and displays them for tailored context. Stacker configs are intentionally declarative so non-developers can modify workflows via JSON.
A named sequence of steps that implements a study-specific processing flow. Pipelines are defined in `config/pipeline_definitions.json` and outline the guidance and order of opertions.
A developer-provided routine that runs once (or on first launch) to prepare the applications assets: verify or download model files, fetch large datasets, and perform one-time initialization. (In the future) Startup behavior can be configured in `config/startup.json`.
Human-editable JSON/YAML files in the `config` folder that drive behavior without changing code. Researcher teams can customize definitions like (pipeline selection, step parameters, study metadata).
Artifacts are outputs produced by steps (alignment files, feature arrays, plots). Provenance is metadata that records which pipeline, config version, component versions, and timestamps produced those artifacts; recording provenance is essential for reproducibility and auditability.