Skip to main content

Modules Overview

The raw output bundle makes the module layer explicit. Each module is not just a score source. It is an evidentiary producer with its own status, result payload, metadata, and provenance hooks.

What the example bundle shows​

The example raw/ directory contains module-native artifacts such as:

  • admet_predictions_raw.json
  • clinical_precedent_prior_raw.json
  • pbpk_sensitivity_raw.json
  • toxicity_profile_raw.json
  • sar_explorer_raw.json
  • target_biology_risk_raw.json

Most follow a family shape built around:

  • module_id
  • generated_at
  • status
  • reason
  • result
  • metadata
  • provenance and evidence-link fields

How module status should be read​

A module can be:

  • executed successfully and evidentiary
  • planned but skipped because of missing input
  • out of scope by category policy
  • partial or failed

That status should always be read together with the module plan and evidence-state machine.

Required template for every module page​

  • Purpose
  • When it runs
  • Required inputs
  • Outputs
  • Interpretation guidance
  • Limitations
  • Caveats