Skip to content

vMLX v1.0.7

Choose a tag to compare

View all tags
@jjang-ai jjang-ai released this 18 Mar 02:01
· 4127 commits to main since this release
v1.0.7
071183d

vMLX v1.0.7

Image System Overhaul

  • All mflux models supported: Klein 4B/9B, Kontext, Fill, Qwen Image, FIBO — each with correct class dispatch (no regex)
  • img2img variations: Generate models support iterative variations via source image + strength
  • Clear separation: Gen models generate + vary, Edit models do instruction-based editing
  • Image fetch timeout fixed: Node.js http.request replaces Electron fetch (was timing out at 5 min)
  • Iterate UI: Click "Iterate" on any generated image to create variations with modified prompts
  • Settings persist: Steps, size, guidance, strength, negative prompt saved across restarts

Critical Fixes

  • Tool calling restored: Reverted asyncio.wait_for → asyncio.wait (was cancelling generators, killing tool calls)
  • JANG config hardening: Corrupt/invalid config files handled gracefully
  • HF README display: HTML stripped, whitespace cleaned, emojis preserved

New

  • Code tab placeholder (Coming Soon)
  • Klein 4B/9B restored with correct Flux2Klein class
  • Kontext, Fill, QwenImage, FIBO added to model picker

Install

pip install --upgrade vmlx

Desktop: MLX Studio v1.2.5

Assets 2
Loading