Msty
Talk to us

Changelog

Msty Nexus Changelog

Release notes for Msty Nexus, newest first.

Msty Nexus Changelog


Advanced presets and seamless app updates🔗

2026-08-06
Msty Nexus 0.4.1

Msty Nexus 0.4.1 expands preset controls for reasoning and caching, improves model handling, and makes Nexus easier to keep updated and available in the background.

Features

  • Added preset controls for thinking effort, reasoning display, Anthropic prompt caching, and advanced provider options, with compatibility warnings for unsupported settings.
  • Added in-app and system update notifications with downloads for the correct installer on each supported operating system.
  • Added close-to-tray behavior so Nexus remains available after its window is closed.

Improvements

  • Improved downloaded model handling so aliases such as a base model name and its :latest tag no longer appear as duplicate entries.
  • Improved Ollama embedding model loading reliability.
  • Hid Nexus MLX options on operating systems that do not support MLX.
  • Improved Smart Routes for mixed text and image workloads with endpoint-compatible model selection, dedicated image fallbacks, and clearer validation.

Fixes

  • Fixed running LM Studio and Msty Studio Local AI providers appearing as Setup instead of Live on the Overview page.

Gauge usage workspace, expanded gateway routes, and model health checks🔗

2026-07-27
Msty Nexus 0.3.0

Features

  • Added Gauge as an opt-in AI usage and runway workspace.
  • Added image, speech, and rerank gateway routes.
  • Added model library health checks.
  • Added multipart transcription and image editing.
  • Added loaded model memory state visibility.
  • Added an auto-start option in settings.

Improvements

  • Added Always on Top to the app menu and window controls.
  • Hid models from Smart Route model pickers when they are hidden on the Models page.

Fixes

  • Fixed Msty Studio detection as local AI so it can be added correctly.
  • Fixed the tray icon not appearing when Nexus runs in the background.
  • Fixed custom OpenAI-compatible provider names not appearing in usage reports.
  • Fixed Nexus Dock UI issues.
  • Fixed rescanning so onboarding providers are not the only providers shown.
  • Fixed Llama.cpp model scanning so it no longer installs the default model automatically.
  • Fixed added Ollama providers and models not appearing in model lists.
  • Fixed app restart behavior for DEB and AppImage builds.
  • Fixed push notification toast messages showing duplicated words.

Local AI discovery, Smart Routes, and accelerated presets🔗

2026-07-09
Msty Nexus 0.2.0

Features

  • Discover and connect to local AI runtimes already running on your machine.
  • Create Smart Routes that send requests to the best configured model.
  • Speed up local model presets with optional acceleration.

Improvements

  • Improved first-time setup and provider onboarding.
  • Made connection endpoints clearer and easier to copy.
  • Clarified which models run locally and which use online providers.
  • Rename client tokens for easier identification.
  • View usage broken down by client token.
  • Improved Msty Nexus icon visibility across desktop themes.
  • Improved model and provider setup guidance.

Fixes

  • Fixed provider menus being cut off near the bottom of the list.
  • Fixed the restart-required status remaining after restarting the runtime.
  • Fixed model search results remaining after clearing a search.
  • Fixed runtime and provider status not updating between the main window and menu bar.
  • Fixed the Advanced Options model search field becoming unusable.
  • Fixed Linux AppImage and DEB launch and compatibility issues.
  • Fixed system tray icon rendering issues.
  • Fixed saving an empty allowed-origins list.
  • Fixed hidden models remaining visible in connected clients.
  • Fixed duplicate Msty Nexus windows appearing from the Dock and menu bar.
  • Fixed Claude models failing through the OpenAI-compatible endpoint.
  • Fixed duplicate model-pull actions while another pull is in progress.
  • Fixed removed providers and their models remaining visible.
  • Fixed duplicate preset names causing inconsistent availability in connected clients.