Andromeda v0.8.0

The Wakeword is finally here. You no longer need to touch your keyboard, or if you are gaming, you don't have to alt-tab out of the game to speak with Andromeda.

I experimented for a long time and finally found a clean solution.
Additionally, Andromeda can now see images.
(Only models that support Vision)

So have fun.

Patchnotes

New Features

• Wakeword Recognition: Added hands-free control via voice activation. Preconfigured as well as custom wake words are supported.
• Vision Support: Integrated image and graphics analysis for compatible multimodal models.
• Audio Feedback & Sounds: Added new sound effects for wake word recognition and sending messages, including dedicated sound settings.

Improvements & Optimizations

• Intelligent Model Selection: Optimized automatic model selection to ensure smooth performance, especially during resource-intensive applications (e.g., gaming).
• Hardware Protection: If hardware resources are insufficient, the thinking process is intercepted beforehand and the user is informed directly via a chat message.
• Interactive Onboarding: Extended the onboarding tour with a help card for setting up and using wake words.
• Chat Performance: Noticeably improved processing speed and rendering in the chat area.
• Security & Stability Checks: Implemented extended safeguards and validations for Vision and wake words.

Bug Fixes

• Loading Screen: Fixed an erroneous additional window in the taskbar during the loading process.

Model Library Update & Outlook

Additionally, here is a small update regarding the model availability:
My plan is to significantly expand the list of available models later and keep them up to date.

Since I am currently focusing on core features, no new models will be added to Andromeda for now. I have already prepared the underlying system, but because I write a description for every model, each one needs to be translated into all 22 languages and assigned to categories. With well over 300 models, that is quite a task.
By version 1.0 at the latest, most existing models will be included, and I will keep them updated.

I have already built an internal page specifically to display all models and the latest releases according to my needs.

Since I am working solo, I have to set priorities.

So I hope you can wait a little longer. The library will definitely be very extensive by 1.0.

So then I wish you have fun with the Wakeword and Vision :)