Downloads
| Architecture | MacOS | Linux | Windows |
|---|---|---|---|
| x86 | Mac Intel | Linux x86 | Win x86 |
| ARM | Mac ARM | Linux ARM | Win ARM |
This release introduces on-demand, per-user file-index sharding across the server and app layers, significantly reducing startup work and scoping file discovery to only the relevant user's data.
Highlights
- Per-user file-index shards loaded on demand – File and module APIs now demand-load the caller's L2 file-index shard via
context.ensureUserFileIndex?.(username)instead of scanning the full app index at startup. - Async HTTP handlers – Many HTTP handlers have been converted to async to support the new demand-loading flow.
- Shard-scoped pattern discovery –
stateSystemis now passed intofile_pathsso pattern discovery operates within the correct shard boundary. - New
file_index_store– Added a dedicated store and related file-watch adjustments to manage sharded index lifecycle. - Documentation updates – Server and app docs now describe sharded
file_index, on-demand L2 loading, auth-only hints, and watchdog/reconcile behavior. - Reduced startup cost – The system no longer scans all users on boot; discovery and listing are scoped to relevant shards only.
Commits
- Ensure per-user file-index shards on demand: demand-load caller's L2 shard in file/module APIs, convert handlers to async, pass
stateSystemintofile_pathsfor shard-scoped discovery, addfile_index_store, update file-watch logic, and revise documentation and tests to reflect new semantics.