Skip to content

container-conversion.md

Latest commit

 

History

History
66 lines (54 loc) · 3.57 KB

container-conversion.md

File metadata and controls

66 lines (54 loc) · 3.57 KB
Raw

Container conversion

The container importer translates the initial-world KCD1 Stash family to native KCD2 storage. Temporary unowned access is intentional until homes and NPC ownership are converted. Inactive quest layers remain inactive.

Source clothing support

Container loot uses the original item GUIDs. Missing wearables are imported with their KCD1 meshes, materials, colour parameters and available morph variants. Equipment roles are selected from source garment archetypes, then assigned native KCD2 equipment slots and layers. The source material-type number is not treated as a native equipment-slot enum.

Supported paths include male and female clothing, armour, helmets, gloves, boots, jewellery and horse equipment. Native definitions already present in the installed item registry are reused. This is not a blanket replacement of native items with KCD1 models.

The current dependency run resolves all 589 item types referenced by the audited stash inventory presets. It produces 312 additional wearable components. Twenty source entries reference morph names absent from their mesh: those use the original base mesh, and the missing request is recorded in the dependency report. They do not silently select another variant.

Every generated PresetItem includes both its registered native Name and original item GUID. GUID-only entries passed schema validation but produced blank loot screens in the development game. The correction covers all 174 generated inventory tables (2,208 item entries), and the user confirmed the starting-home fruit sacks work after the correction. The converter rejects items without a resolved native name. This is a sampled gameplay check, not whole-map acceptance or a save/load test.

The skin converter preserves render geometry and maps source weights to the native skeleton. An explicitly enabled correction handles uniformly offset internal bind-position buffers, with a 25 cm limit; nonuniform discrepancies still fail. This does not establish that every garment fits every native body or layers correctly with every outfit. Runtime fitting remains a separate acceptance check.

Build and evidence

  1. tools/container_audit.py reads active source placements, inventories, locks, phases, ownership and service references.
  2. tools/container_items.py resolves item, wearable, model, material and localisation dependencies before container registration.
  3. tools/build_world_containers.py stages native stashes, inventory presets, key dependencies, original models and phase models in bounded asset shards. It preserves the existing shared personal-storage registrations.

The staging builder does not install its output. Its localisation archive is an addition to the installed archive and must be merged, not copied over it. The pre-conversion level is required as the baseline; repeated registration of the same source container fails.

Reports distinguish generated data from runtime verification. Native loot generation is not seed-identical to KCD1. The current translation retains direct rows and weighted preset choices, with explicit quantity ranges. Restock intervals are translated to integral native days; sub-day intervals round up and are identified in the report. Source random-selection semantics and restoration after save/load require further acceptance testing.

Source merchant generation, home/NPC-generated contents, confiscation and campaign-driven transfers are additional contracts. A converted physical stash is not evidence that those services are active. They must not be reported as completed solely from entity or item-definition counts.