Skip to content

Hermes Local v0.18.0 — Portable configuration

Choose a tag to compare

View all tags
@xdCloudy xdCloudy released this 28 Jul 11:56
· 387 commits to main since this release
v0.18.0
b31d4e7

Portable configuration release

Hermes Local no longer assumes one drive, GPU, CUDA architecture, model, profile, or port pair.

Highlights

  • Register and switch arbitrary local GGUF models from Hermes Launcher.
  • Create, edit, select, and delete inference profiles.
  • Configure CPU/CUDA/auto acceleration, loopback host and ports, build workers, CUDA architecture, Python line, and startup model verification.
  • Resolve project paths from the clone and tune thread/VRAM defaults from the current machine.
  • Keep personal choices in an ignored, atomically written config/launcher/user-settings.json.
  • Generate Hermes provider configuration dynamically while preserving unrelated user configuration.
  • Validate portable defaults, model manifests, profiles, and user settings with versioned JSON Schemas.

Install

Clone the repository, run Setup-Hermes-Local.ps1, then use either Windows launcher asset below. The release binaries contain the launcher, not model weights or the managed runtimes.

The binaries are not Authenticode-signed, so Windows may show a reputation warning. Verify SHA256SUMS.txt before running them.

Verification

Validated on native Windows with the live local stack, an alternate model/profile/port smoke configuration, packaged and portable Electron interaction tests, schema validation, PowerShell parsing, TypeScript checks, focused Electron tests, and a clean Windows Defender distribution scan.

Assets 7
Loading