Overview and Ownership

This chapter defines what PLEB is responsible for and how it integrates with PQC.

PLEB owns

  • data ingest/discovery/copy/verification,

  • branch-aware dataset update control,

  • fix policies for tim/par edits,

  • stage ordering and run layout (serial/parallel/workflow grouping),

  • optional post-fit tools (whitenoise and public release comparison),

  • output file placement and reproducibility metadata.

PQC owns

  • detector-level feature and outlier decisions,

  • row-level QC labels,

  • detector diagnostics and model-specific statistics.

Operational model

Think of PLEB as:

  1. a stage coordinator,

  2. a policy applier,

  3. an output/reproducibility manager.

The practical boundary is:

  • PLEB decides what to run, where to read/write, and how to apply outputs.

  • PQC decides which points/events are flagged by detector logic.

UX boundary

The UX module (pleb init/run/doctor) is a configuration front-end:

  • UX builds project-style TOML files,

  • UX translates them to the flat runtime config,

  • core execution remains in the same ingest/fix/pipeline/workflow code paths.

Cross-reference