Skip to content

VectorDash 0.3.45

Latest

Choose a tag to compare

@github-actions github-actions released this 09 Nov 03:49
52dca6b

VectorDash 0.3.45

New controls

  • GPU idle threshold selector: choose which detected GPU the idle threshold applies to, or select Use model GPU. CUDA selections resolve to their Windows adapter identifier through the NVIDIA driver. Schedule status names the monitored GPU. A busy iGPU no longer blocks indexing on a selected NVIDIA GPU.

Fixes since 0.3.44

  • Network drives: resolve mapped-drive monitoring, traversal and content reads through the UNC share while retaining configured drive-letter paths and source policies. Add a five-minute network-only catch-up for missed SMB notifications, respecting idle/pause rules and retaining records when a share is unavailable.
  • Quick-search dismissal: handle Escape before response widgets consume it and detect outside clicks using actual window bounds. Keep the popup open during AI generation and active/paused speech; restore dismissal once both are inactive.
  • Idle clipboard activity: stop repeatedly querying empty OCR/image-embedding queues every two seconds. New captures/imports and relevant setting changes trigger a fresh check, with a one-minute safety recheck. Add indexes for background candidates and retention, and skip repeated schema inspection after migration.
  • Clipboard pause: background image embedding now respects capture pause, matching OCR.

Validation

Unit coverage includes per-adapter GPU filtering, saved selector settings, history migration, indexed candidate lookup without payload-table reads, empty-queue timing and image-capture wakeup. A live CUDA check mapped CUDA0 to the NVIDIA GeForce RTX 5080 and sampled its utilization. Previous beta GUI checks verified popup protection during AI/speech and dismissal after completion/stop; a native watcher test passed on a temporary folder on T:.

On a disposable copy of the reported clipboard database, the indexed query reduced SQLite page-cache misses from 7,985 to 7 and median lookup time from 15.485 ms to 0.0335 ms. These are query benchmarks, not physical disk or whole-app CPU measurements.

All fixes from 0.3.45-beta.1 and beta.2 are included. Multi-adapter compute APIs that cannot be resolved automatically require an explicit GPU selection rather than using another adapter's load.