v2.4.8: Set a Thinking Level for Any Model, and a UI That Honors Your Brand
The headline in v2.4.8 is a small word with a big effect: thinking. You can now dial the reasoning effort of any thinking-capable custom LLM, exactly the way you already could with Claude. Around it sits a batch of quality work: a smoother hiring wizard, a UI that finally honors your brand colors everywhere, license and SMTP changes that land without a restart, and desktop installers that ship to every instance out of the box.
Here is the full release.
Pick a thinking level for any custom LLM
Until now, the reasoning effort selector was a Claude privilege. In v2.4.8 it works for your custom models too, right from the model picker in chat.
The part worth appreciating is how little you have to do. When you add or edit a model, Geta.Team probes it to find out whether it actually supports reasoning and stores the correct request format for it. That works across OpenRouter models, OpenRouter presets, and self-hosted or OpenAI-compatible endpoints alike, with no manual setup on your end.
What you see is one simple scale for every model: Off, Low, Medium, High, Xhigh. Behind that scale, the right wire format is used per model, and on OpenRouter the effort is mapped to the correct token budget even for models that only think in tokens. For the rare endpoint we cannot probe automatically, there is a Reasoning effort toggle with a format choice and an Auto-detect button when you edit the model.
The level is set per employee and survives restarts. Switch to a model that does not think, and the selector simply disappears. No clutter, no guesswork.
A guided hiring wizard that tells you what it needs
Two fixes here, both about clarity. The Next button on the first step used to do nothing when required fields were empty, which left you clicking into the void wondering what was wrong. It is now disabled until those fields are filled, so the state of the form is obvious at a glance.
And validation errors and required-field markers now use your theme's primary color instead of red. The old red was genuinely hard to read on dark backgrounds, so this is a small accessibility win as much as a cosmetic one.
No more flashing suggestion cards
Refreshing the page mid-conversation used to briefly flash the starter suggestion cards, then hide them again once history loaded. That little flicker is gone. The cards now appear only once the conversation is confirmed empty, so a refresh stays calm.
Your brand color, everywhere it should be
Input focus highlights, like the one on the file explorer search box, used to stay stubbornly blue no matter what custom theme color you had set. White-label colors now drive the focus ring too, so every focus highlight matches your brand. If you run Geta.Team white-labeled, the small inconsistencies that gave it away are getting swept up one by one.
License and SMTP changes that apply on the spot
Activating an Enterprise license or saving your DKIM and SMTP settings now takes effect immediately, with no restart of the stack. Email sending and receiving unblock the moment you save: the SMTP service reloads its configuration, and employee mailboxes are provisioned automatically right then. Previously this kind of change meant a restart and a wait. Now it is instant, which matters a lot when you are mid-setup and just want your employee's inbox to start working.
Desktop installers, baked in and seeded everywhere
The desktop installers are now built into the image and seeded to each instance on startup, so a fresh instance serves them out of the box. Before this, the macOS .dmg and its auto-update feed only existed where someone had placed them by hand, which meant other instances greeted you with "No installer available yet." That gap is closed. The Windows installer seed is refreshed to the current build, and the two platforms are now seeded independently, so a Mac-only or Windows-only update can propagate on its own.
One to watch next
A note on what is queued: the desktop app now persists your signed-in session across launches, and even across a reinstall, using the OS-encrypted store in a location that survives an uninstall. No more logging in every single time you open the app. That one ships the moment a new desktop installer is built and seeded, so consider it a preview of the next release.
The short version
Set a thinking level on any custom model, get a hiring wizard that explains itself, see your brand color on every highlight, change your license or SMTP without a restart, and find the desktop installers waiting on a fresh instance. If you are running Geta.Team, it is already live for you.
Want to test the most advanced AI employees? Try it here: https://Geta.Team
