v0.1.1

Patch release for dsh-koboldcpp-hands (KoboldCpp for DeepSeek Harness).

Changes since v0.1.0:

• README: new Uninstall section (remove the profile row, per-stopBehavior server handling, no-residue notes) including one-command removal (
  pm uninstall dsh-koboldcpp-hands / Remove-Item -Recurse -Force ).
• README: new FAQ: my main model is text-only - how do images get in? - pasted images land as temporary file paths in text-only conversations; the model passes them via image_paths, no harness upload involved (same approach as the community's flagship vision bridge).
• Vision tool description now tells the online model to use image_paths when the main model cannot receive images.
• Removed third-party attribution mentions from credits.

Same compatibility: DeepSeek Harness 0.1.0-rc series, Node >= 20, MIT.

v0.1.0

First release of dsh-koboldcpp-hands - KoboldCpp for DeepSeek Harness.

A third-party tool plugin that lets the harness online model hand repetitive text and vision (OCR) labor to a local KoboldCpp (llama.cpp) server.

Tools: koboldcpp_run (text) + koboldcpp_vision (images/OCR, analyze/ocr/compare modes).
Lifecycle: on-demand auto-launch with your .kcpps, idle/exit stop, external servers never killed.
Compatibility: DeepSeek Harness 0.1.0-rc series, Node >= 20. MIT licensed.