Skip to content

Reference

This section is the technical map for AnomaLog.

Use it for two things:

  • navigating the public API
  • orienting yourself in the codebase before drilling into the generated reference pages

Reference pages

  • Top-level API DatasetSpec and SplitLabel

  • Presets Built-in dataset specifications

  • Parsers Structured parsers, template parsers, sinks, and TemplatedDataset

  • Sources Dataset source implementations

  • Labels Label readers and lookups

  • Sequences Sequence builders and sequence objects

  • Representations Model-facing sequence representations and lazy representation views

  • Experiments Config-driven experiment configs, model runners, and result helpers

Codebase layout

  • anomalog/__init__.py exposes the small top-level public API
  • anomalog/dataset.py contains the fluent DatasetSpec builder
  • anomalog/presets.py defines built-in dataset specifications
  • anomalog/parsers/ contains structured and template parser abstractions
  • anomalog/sources/ contains dataset materialisation logic
  • anomalog/labels.py contains label readers and lookups
  • anomalog/sequences.py contains sequence grouping and split behavior
  • anomalog/representations/ contains model-facing sequence representations
  • anomalog/_runtime/ contains internal orchestration code
  • experiments/ contains the repository-local experiment configs, runners, and result helpers

Experiment layer

The experiments/ directory is useful for reproducible model experimentation after preprocessing, but it is separate from the core public API of the library. See Experiments for the module reference and docs/experiments.md for the workflow overview.

Note

The reference pages document symbols. Start with Getting Started for the onboarding path and Pipeline Concepts for the stage-by-stage mental model.