Shared SliceCache ("S-cache") for per-slice decoded-geometry reuse.
The source under the caching store that serves a packed store's small nodes
from one request each (luxar optimize --pack); hand it the store's index
with adoptChunkPacks. null without the caching store (?noCache), where
every read is plain.
A fresh sidecar read, sharing the already-open archive reader for zip stores.
Resolved cache telemetry state for the UI monitor. Reflects the actual policy decision the cache stack made:
disabled-no-cache: URL ?noCache flag.disabled-config : appConfig.cache.enabled === false.enabled : at least one tier (L0 or L1/L2) is active.
Pushed to the monitor via setCacheTelemetryState() in
wireMonitorAfterLoad.The resolved per-tier budgets (and their source: heap / explicit / device-class / fixed) — surfaced so the Settings popover can show the user what their budget actually resolved to.
The load-time root-document fetch validation and the store open share
(cache/root-document-prefetch.ts), or null for a URL it does not apply
to (zipped / presigned / non-http). The loader reads the response's ETag from
it for the identity watchdog.
Stop answering from rootDocument and drop the registry's reference — call once the root has been opened. Later root reads (an invalidation) then reach the server as before. Idempotent.
True when the root metadata the store opened with came from the network this
load (no cache tier served it), so its consolidated index is current. False
when L2 served it: L2 revalidates by content_hash alone, which a sidecar
edit such as luxar env attach leaves unchanged.
Result of cache setup: the three layers and a ready-to-open store.