Sophia v1.1 Released -- Contextual Awareness

Posted Fri May 29th 10:55

Sophia v1.1 is now available, and can be installed via the Settings -> Addons and Settings -> Devices and Services menu of your HA device.

After listening to beta testers, it was clear an architectural mistake was made. During initial development in an attempt to conform with HA norms, it was decided Sophia itself would be scaled back and instead of simply understanding your message, would require the friendly name in order to invoke a device. This was clearly a mistake. People just want their HA voice assistant to understand them using common sense, without having to remember the friendly name of every device.

This upgrade reverts much of the HA specific pipeline, allows Sophia to operate as it was designed to, and should dramatically improve results, fluidity and comprehension. Regardless of your house layout and whether or not it contains an art studio, fitness room, jacuzzi, or outdoor sprinkler system, Sophia should now simply understand you and properly interpret your messages.

To gain an understanding of what Sophia is capable of, you may view some test sentences used to verify this upgrade at test sentences

Strict vs. Flexible Mode

Two modes of understanding are now available, strict and flexible. Sophia defaults to flexible, but can be changed via the Settings menu of the web interface.

  • Strict - Requires the exact friendly name in order to invoke the device.
  • Flexible - Gives Sophia leeway to determine the device(s) to invoke based on context.

There is also a new Tags column beside every device within the Entities menu of the web interface allowing you to define a comma delimited list of words for devices to provide Sophia more contextual information as to the device's use and location. For detailed information, please visit the updated documentation at Talking to Sophia.

Beta Testing Status

Thank you for your patience while this upgrade was being developed. Even though this is the era of vibe coding where releases are expected to ship within hours, and not days or weeks, Aquila Labs simply refuses to sacrifice quality and maintainability for expediency.

I'm confident the coming years will prove that quality matters more now than ever, and treating folks with the common respect and decency to only ship code that is fully understood, and can be properly serviced will pay off in dividends. You may view full details on how AI is deployed within the development lifecycle at: AI Coding Policy.

Beta testing should now be nearing completion, with the general release on the horizon. Thank you to all beta testers for your time and patience during this period, it's greatly appreciated, and won't be forgotten.

Moving Forward

Please continue submitting any problems or issues, along with any requests, suggestions or general feedback. You may e-mail [email protected] directly or post here on the forums. You can login to the forums with your nlu.to account.

Many upgrades are planned for the future -- from better compression and lower memory footprint to multi turn dialog conversations and more. Stay tuned.

Thank you again for your time and patience, and we'll be in touch shortly with even better software that simply helps. This is only the beginning.

NOTE: Github isn't working, this upgrade is first time compiling locally without Github runners. x86 will be fine, and please drop message if other architectures fail.