Neon AI - New Conversational Hardware - Wireless Audio Streaming - October 2023 Update

[NeonClary]

NEWS FOR NEON AI USERS & AI APPLICATIONS AND SYSTEMS DEVELOPERS

“TLDR ;-)”

• New Hardware Options in the Works for Neon AI Devices! Our dev team has added the high performance Orange Pi 5 to our supported platforms … and more to come. See a Demo
• Continuous Speech Processing
• Wireless Audio Streaming: Bluetooth, AirPlay and Spotify
• New Chatbot Support
• Factory Reset with 2-Button Press on Mycroft Mark II

October 2023 Update: Neon AI Open Source Conversational AI Platform Enhancements

Continuous Speech Processing - Speak Without a Pause

• New speech handler doesn’t need you to pause between “Hey Neon” and speaking a command.
• No need to wait for the ‘listening “oop” sound after saying “Hey Neon” before telling Neon to perform a request.

Thank you Mycroft AI and OVos for sharing your community development.

Wireless Audio Streaming !|20x20

Play audio on your Mark 2 via AirPlay, Bluetooth, Spotify, or KDE Connect.

• Bluetooth

Say “Hey Neon, pair Bluetooth”.

Select “neon” from the available connections on your device.

• AirPlay

Select “uxplay@neon”.

• Spotify

In Spotify Premium, go to Settings > “Connect to a Device” > select “Neon Mark II.”

Thank you community developer @mikejgray!

See the Skills Guide at this Google Doc link for more details.

“Hey Neon, I want to talk to FastChat”

• We’re now hosting our own instance of the FastChat LLM.
• It provides an open source alternative to ChatGPT.
• Say “Hey Neon, send me my chat” to get a log of your chat emailed to you.

Thank you Neon AI Ukrainian team!

New Low Cost, High Performance Hardware Platforms for Neon AI

We are pleased to bring you new hardware options for running the Neon AI OS and Personal Assistant voice skills. These are DIY system options right now, and we invite interested community members to participate in developing reasonably priced, high performance, end-user ready systems for homes and businesses.

Supported Devices

• Orange Pi - Supported DIY build, 2-4x performance advantage over the Mark II’s Raspberry Pi 4. Orange Pi provides the performance necessary for on-device STT, with TTS improvements in progress and we expect the same from the other hardware options in development.
• Intel x86 family processors - Supported, including server processors. Can achieve more than 10x the Mark II’s current performance. That performance increase makes on-device speech-to-text, text-to-speech and foreign language translations run smoothly.
• Mycroft Mark II / Raspberry Pi 4 - Supported.
• On-Premises Servers
• Cloud Integration

Under Development Devices

• Rock 5 - Performance advantage of 2-4x expected.
• Raspberry Pi 5 - Coming soon.

Please engage with the Neon AI community of developers on Element.

Photo: Running the Neon AI OS on an Orange Pi

Factory Reset with a Two-Button Press

• Neon OS now provides a method for resetting an installation to a clean state, even if it will not boot properly.
• To reset the Neon OS on your Mark II device to factory settings, first unplug it. Then hold down the left “Volume -” and circular “Action” buttons while plugging in the power. Neon will start up, and you can release the buttons after 5 seconds. Neon will reset itself, and boot as if it is the first time it was powered on.
• All configuration, data, network information, and user-installed software will be removed.

If you haven’t upgraded to Version 2.0 of the Neon AI OS yet, visit Upgrade Guide - Neon OS V2.0 for the Mark II | Neon AI

“Neon has been great to work with. The new version is much faster and more stable.” -Randy F.

Hints

• “Hey Neon, Enable LLM fallback skill” - when no other skill is identified to provide a response to your query, an LLM will do so.
• To disconnect & reconnect your Mark II from wifi without having to re-enter the password:

Swipe down for the menu, choose Wireless → On Device Setup and select your currently connected wireless network. It will disconnect automatically. Select it again to reconnect.

• When playing music, press the square stop icon on the music widget a second time to exit.
• Change the homescreen wallpaper in the system menu. Swipe down for the menu, choose Settings → Display → Wallpaper Settings
• “Hey Neon, Create a support ticket” to generate and email a log of recent activity.
• For V2.0 upgrade support options, visit Upgrade Guide - Neon OS V2.0 for the Mark II | Neon AI

Community Collaboration

Community Testers wanted: Visit the new Matrix chat room for testing requests. Developers working on advanced skills for the Neon OS will benefit from your testing and commenting on new features before release. To run advanced versions you can say: “Hey Neon, Enable pre-release updates”.

Community Developers wanted: Bounties available for skills, applications and documentation development. The Neon AI team is available to assist community developers.

Contact clary@neon.ai with your ideas and suggestions.

Resources

1. Quick Reference & Quick Start guides.
2. GitHub documentation at: About - NeonAI® Documentation
3. Advanced Neon developer updates are posted on our GitHub Discussions page

Please visit our Matrix community to continue the conversation and read more about the Neon OS.

LLMs and AI Chatbots in a Collaborative Conversation

Thank you for reading this update! Let us offer you a look at another of our AI projects, Chatbots Forum 2.0, where LLMs and Chatbots talk to each other.

ChatGPT, Llama, and Fast Chat are some of the LLMs getting all the press lately, and here on our forum, along with some AI chatbots from our version 1.0 they can discuss and vote on prompts submitted by users.

Users will soon be able to adjust the AI’s behaviors and ultimately even add their own AI participants to the mix. TTS & STT in multiple languages is also on the way. The explanation and documentation is on the main site with version 1.0, but we invite you to give the 2.0 beta version a spin and offer your input.

Wishing you success with your private personal digital assistant,

Clary Gasper

Neon AI Projects

Clary@Neon.AI

[pesone]

Does this only apply to a Orange Pi 5, or would a Orange Pi 5 Plus also work?

[NeonClary]

I think the answer is yes, but that would be a question for @NeonDaniel when he gets back from our holiday.

[pesone]

thanks for the info, I ordered a pi5,
and i am excited to try the orange pi master image

[NeonDaniel]

I’ve developed with an Orange Pi 5B but I believe anything using the RK3588 processor should work with those images. More testing to do and Rock5 support is on my short list too since it should be essentially the same.

Let me know if you run into any questions/issues and I can try to iron out any bugs .

[pesone]

yes, I ordered a standard orange pi 5 with a rk3588s.
I was not able to start the image from app/files/neon_images/orange_pi_5/master/
it is not loading the kernel.
any idea how to boot the neon.squashfs on a rk3588s?

[NeonClary]

@pesone Sorry to hear it’s not going smooth for you! I’m tagging @NeonDaniel here, he’s out sick but once he returns I’m sure he’ll be happy to help.

[pesone]

Thanks, I would be happy to try to compile it for the rk3588s, but I would need a little hint where to start,
otherwise I will order another device with rk3588.

[NeonDaniel]

@pesone

I was not able to start the image from app/files/neon_images/orange_pi_5/master/
it is not loading the kernel.
any idea how to boot the neon.squashfs on a rk3588s?

I’m not totally familiar with the OrangePi bootloader yet, but I believe the kernel load address is hard-coded so on the OPi 5B, it always checks the Micro SD slot and then onboard EMMC. Is your Pi booting a different OS, or not at all?

@pesone

Thanks, I would be happy to try to compile it for the rk3588s, but I would need a little hint where to start,
otherwise I will order another device with rk3588.

The build is done with debos recipes in this repository. I wonder if u-boot needs to be built with a different configuration here.

I’ve created an issue on GitHub to track Orange Pi 5 Plus support.

[AlertSubject]

@NeonClary

“Hey Neon, Enable LLM fallback skill” - when no other skill is identified to provide a response to your query, an LLM will do so.

I’d love to do so, however Neon seems to have a had time understanding LLM
I managed to get it to work after idk how many attempts now but maybe it would be a good idea to have an alternative command ready for this
such as “hey neon, enable language model fallback”
anything you can actually say that doesnt contain abbreviations

[NeonClary]

@AlertSubject

hey neon, enable language model fallback

I’ve put in a PR for that change, thanks for the excellent feedback and suggestion! Since it was in one of our own skills I was able to easily locate & update the intent file. Daniel will need to approve it, and it will go into a build soon - I expect you’ll see it in the beta first, then the stable version.

Those kind of changes are simple enough that I’ve been wondering if folks would like a tutorial on how they work. You wouldn’t need to be a skilled programmer to become a whiz at enhancing the language side of our interface.

(If anyone for whom this work is new decides to jump in with two feet, please make a draft PR and tag me for review first before submitting anything to Daniel.)

[AlertSubject]

@NeonClary

Those kind of changes are simple enough that I’ve been wondering if folks would like a tutorial on how they work. You wouldn’t need to be a skilled programmer to become a whiz at enhancing the language side of our interface.

Yes, that sounds good - would be nice to help with some smaller things like this myself

@NeonClary

I’ve put in a PR for that change

Please also adjust the documentation and the Paper instructions that you send out with the boot drives Otherwise people will still only see the old command being suggested