Skales v12.9.15 - Hotfix

Here's a look at what's new in Skales v12.9.15. We've focused on making the experience more reliable and predictable, particularly around how Skales interacts with your local models and your data.

Mailbox Handling

We've adjusted how Skales reads your emails. Previously, asking for a single message could inadvertently load your entire mailbox, sometimes leading to unexpected crashes. Now, Skales fetches only the specific message you need, or the ones relevant to a reply, making the process much more efficient and stable.

Updates and Shutdowns

If you've ever updated Skales on Linux and found old versions lingering, that should be resolved. We've added clearer shutdown procedures for updates, ensuring that previous instances and their associated local model servers are properly closed down, preventing unnecessary CPU usage.

Model Context and Capabilities

Skales now more accurately reflects the context window of your local AI models. It will correctly identify and use the model's actual capabilities, preventing errors and ensuring conversations fit within the intended limits. We've also made sure that smaller models are no longer asked to perform tasks they aren't designed for, improving their stability.

Google Account and Model Identification

The Google account card now provides clearer feedback if there's an issue with token refreshes or status reads, explaining exactly what happened. Additionally, when you look at your AI models, you'll now see which provider each one comes from, making it easier to distinguish between models with similar names.

Documentation and Connectivity

We've documented the incoming webhook, so you can now find information on how to use it. Furthermore, Skales is now more resilient to temporary connection drops. Instead of restarting the entire application, it will handle these events as normal occurrences, ensuring your work isn't interrupted by minor network blips.

Download Skales