github ampersandru/VectorDash v0.2.23
VectorDash 0.2.23 - Download rate-limit recovery

one month ago

VectorDash 0.2.23 fixes rate-limited setup downloads reported in issue #2.

Download VectorDash-Setup-0.2.23-x64.exe below. Exit VectorDash, install the update, and reopen it. No index rebuild is needed.

  • Setup now identifies GitHub runtime catalog versus Hugging Face model failures, instead of showing a bare HTTP 403. Access-denied errors are distinguished from rate limits.
  • Automatic bounded retries handle temporary server/network errors and short rate-limit waits. Long waits show a suggested retry time; cancellation interrupts waiting.
  • Public publisher catalogs are cached for ten minutes, with a last-known catalog fallback up to 24 hours during service failures. This avoids repeated catalog calls on every setup attempt.
  • Models & hardware → Download and set up now offers optional GitHub and Hugging Face tokens. Save before retrying. Tokens are masked, protected using Windows DPAPI, scoped to their matching HTTPS service, stripped on cross-origin redirects, and excluded from parser workers.
  • Existing partial-download resume and publisher checksum requirements are preserved. Successful runtime installation is retained if a later model download fails, so setup can be retried.
  • OCR language catalog requests also use the cache/retry/token handling.

Validation: 173 automated tests, including rate-limit recovery, cancellation, resume/checksum verification, token protection and redirect isolation; live public catalog retrieval including Vulkan/Q8_0/encoder metadata; mocked setup/retry checks; packaged UI/dependency/indexing/encryption/MCP checks.

The installer is unsigned. Windows may show a SmartScreen or unknown-publisher warning. Developed with AI assistance; AI search can miss or misrank results.

Don't miss a new VectorDash release

NewReleases is sending notifications on new releases.