Changes in version 0.6.0                        

  - Standardize the named v5 package logo across README, documentation
    pages and available application headers, with matching light and
    dark artwork.

  - Reframe the public model as abstract logistics-chain simulation with
    changing item specifications.
  - Rename condition-model functions, observation fields, viability
    endpoints and workbench labels consistently.
  - Preserve the simulation equations and random-draw schedules; this is
    a terminology and interface revision, not a newly calibrated model.
  - Earlier serialized objects and scripts require explicit migration to
    the renamed interface.

                        Changes in version 0.5.0                        

  - Introduce the public MIT-licensed academic edition with abstract
    network stages and stochastic disruption terminology. Public
    identifiers use stage and disruption; earlier analysis environments
    require explicit migration.

                        Changes in version 0.4.0                        

  - Add an opt-in degradation-trauma item model with actual servicing
    events, finite procedural capacity and overall viability at 30 days
    from damage.
  - Add aggregate calibration and independent stochastic admission-batch
    replay.
  - Preserve legacy execution by default and support separately pinned
    item comparisons with frozen planning assumptions and observed
    information.

                        Changes in version 0.3.0                        

  - Add matched initial-observation and updated-observation policy
    controls.
  - Separate reactive current-risk planning from the P3 response
    forecast while retaining the same decision architecture.
  - Preserve uncontrolled simulation behavior and document the
    servicing-free endpoint and separate-experiment requirement.

                        Changes in version 0.2.0                        

  - Added explicit replication-seed and bootstrap primitives for
    external study schedulers while preserving caller RNG state and
    original numerical methods.
  - Added a local Shiny scenario workbench with independent and paired
    designs, diagnostics, validated graph editing, provenance and
    replayable exports.
  - PDF exports recover when Cairo is advertised but its runtime cannot
    load; the portable fallback preserves existing graphics devices.
  - Expanded the README, reference help and three executable tutorials.
  - Added a reciprocal analysis/package synchronization contract and
    local gate.
  - Expanded package checks to Linux, macOS, Windows and three R release
    tracks; documentation and coverage artifacts remain within private
    CI.

                        Changes in version 0.1.0                        

  - Initial private package extraction of the five-leg SHIELD
    simulation.
  - Explicit validated scenario inputs and two neutral teaching
    examples.
  - Reproducible batch simulation with preserved caller RNG state.
  - Independent-arm, multi-arm and paired-intervention comparison
    functions.
  - Documented endpoint and information-timing assumptions.
  - Stable log-ratio forecasts for extreme response parameters outside
    the original study range, with unchanged arithmetic within that
    range.