Skip to content

Releases: linuxserver/pelorus

Release list

0.1.1

0.1.1 Latest
Latest

Choose a tag to compare

View all tags
@thelamer thelamer released this 19 Jul 17:49
0.1.1
d2f67da
This commit was signed with the committer’s verified signature.
thelamer Ryan Kuba
SSH Key Fingerprint: MyTvFbYGAe023hGJyvPdzDlaxzgmX3pe55FoU5X+aY8
Verified
Learn about vigilant mode.

0.1.1 - Initial Release

Pelorus is an AI navigator for Selkies-powered Linux desktops. It gives an LLM agent control over mouse, keyboard, screenshots, and window management via the Pixelflux computer-use backend.

What's included

  • FastAPI server with REST API and WebSocket support
  • LLM providers: Ollama, OpenAI-compatible, and Gemini
  • Desktop control via AT-SPI accessibility tree (Fully supported on KDE running under Wayland)
  • KWin D-Bus integration for precise window geometry on KDE
  • Vision mode fallback for apps without accessibility support and models that support vision (games, Electron apps, etc.)
  • Web interface frontend with chat UI, session tabs, streaming output, and server management
  • Configurable via TOML config file or environment variables

Getting started

pelorus
# listens on 0.0.0.0:5100 (or $PELORUS_PORT)

Open http://localhost:5100 for the chat UI or use the REST API.

Known limitations

  • Vision mode is less reliable for precise coordinate targeting than text-based AT-SPI control
  • KWin D-Bus integration is KDE-specific (works without it on other DEs with reduced window geometry accuracy)

Links

Assets 2
Loading

0.1.0

0.1.0

Choose a tag to compare

View all tags
@thelamer thelamer released this 19 Jul 16:57
0.1.0
3d5d61e
This commit was signed with the committer’s verified signature.
thelamer Ryan Kuba
SSH Key Fingerprint: MyTvFbYGAe023hGJyvPdzDlaxzgmX3pe55FoU5X+aY8
Verified
Learn about vigilant mode.

0.1.0 - Initial Release

Pelorus is an AI navigator for Selkies-powered Linux desktops. It gives an LLM agent control over mouse, keyboard, screenshots, and window management via the Pixelflux computer-use backend.

What's included

  • FastAPI server with REST API and WebSocket support
  • LLM providers: Ollama, OpenAI-compatible, and Gemini
  • Desktop control via AT-SPI accessibility tree (Fully supported on KDE running under Wayland)
  • KWin D-Bus integration for precise window geometry on KDE
  • Vision mode fallback for apps without accessibility support and models that support vision (games, Electron apps, etc.)
  • Web interface frontend with chat UI, session tabs, streaming output, and server management
  • Configurable via TOML config file or environment variables

Getting started

pelorus
# listens on 0.0.0.0:5100 (or $PELORUS_PORT)

Open http://localhost:5100 for the chat UI or use the REST API.

Known limitations

  • Vision mode is less reliable for precise coordinate targeting than text-based AT-SPI control
  • KWin D-Bus integration is KDE-specific (works without it on other DEs with reduced window geometry accuracy)

Links

Loading