Keep filter-only route changes on the current session page, prevent superseded analysis results from overwriting the latest time range, and hide unsupported capture controls in Web WASM.
Keep automatic format detection consistent with accepted JSONL syntax and reject unresolved owner metadata. Regenerate the standalone skill validators from the canonical implementation.
Resolve legacy and active data directories through real paths and reject parent-child overlaps before any recursive deletion. Cover settings migration, parent paths, and path aliases in the desktop path regression test.
Emit explicit members before the first streamed message batch so merge previews do not retain metadata-free member stubs. Add parser ordering and temporary database regression coverage.
Clarify that skipped files can leave unique data on either side, so cleanup records intentionally remain unchanged and deletion performs the final active-path safety check.
A retried migration can skip stale files already copied to the target. Keep the source directory outside manual cleanup whenever any files were skipped so newer source data cannot be deleted.
Show the pending cleanup notice in CLI Web after initialization while excluding the browser-only runtime.\n\nStub config persistence in the cleanup test so it cannot overwrite the user's real data directory settings.
Keep the renderer fail-closed until lock state is known, use native modal dialogs without unmounting business state, and coordinate overlapping dialogs while locked. Isolate desktop E2E data from real user settings and keep startup migration inspection read-only.
Persist failed unlock attempts in the lock flag so restarting the desktop app cannot reset the limit. Block further attempts for 30 seconds after five failures and surface the remaining cooldown on the lock screen.
Load standalone CLI assistant configs through the shared manager so fresh data directories receive locale defaults. Back up custom assistants that occupy newly reserved default IDs before installing canonical templates.
Let status-coded 4xx errors reach the shared Fastify handler so missing sessions remain distinguishable from transaction failures. Add regression coverage for the missing-session response.
Normalize member responses from both database and Desktop providers so the public REST endpoint keeps its legacy name field while retaining the newer detailed fields.
Document the single-OS-user product boundary beside the default private temp root so future reviews do not treat unsupported multi-UID sharing as a regression.
Add an exact Vite alias before the workspace catch-all so Electron main-process imports resolve the exported temp-workspace source instead of a nonexistent package-root path.
- Move AI config/CLI guides under /ai with redirect stubs at old /usage paths
- Rewrite why-chatlab: lead with value, drop table for a two-item list
- Rename nav items ("为什么选择 ChatLab", "对接 ChatLab") for clarity
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Treat more than two real senders as stronger evidence than a conflicting private marker while preserving genuine private chats with system placeholders.
Use timestamp and message ID as the stable viewer cursor so historical messages imported later remain in chronological order. Resolve timeline anchors by message time and cover out-of-order IDs with SQLite regressions.
Share auto-import planning between preview and write paths, expose a machine-readable CLI result envelope, and preserve explicit parser options during analysis.
Keep the sidebar header below the fixed 32px title bar after removing the outer margin. This prevents header controls from being covered on non-macOS platforms.
Route Desktop JSON dry-runs through a shared read-only push analyzer so validation, deduplication, and member inference match actual writes. Keep JSONL on the file analyzer path and cover new and incremental predictions without database mutations.
Use the long-running worker timeout and refresh Desktop caches after successful push imports.
Align auto-created member metadata and counts across initial and incremental imports, exclude SYSTEM from member updates, and keep existing-session writes validated and atomic.
digua
