I'm starting a completely new AI project called E.V. — not an update to my previous Nova AI project.

The goal is to build something closer to a real desktop companion than another chatbot.

What I'm aiming for

• Always-available voice interaction instead of a traditional chat interface
• Natural conversations, including proactive conversation when appropriate
• Camera vision for understanding the user's environment
• Face and eye recognition
• Screen and object understanding
• Hand tracking with detailed hand landmarks and gesture recognition
• An interactive 3D mode controlled through hand gestures
• Create and manipulate 3D objects in the air
• Persistent memory across conversations and sessions
• Awareness of projects and the user's current working context
• Desktop control for apps, files, and workflows
• Permission controls before sensitive actions
• Verification and recovery when an action fails
• A floating, minimal UI that can be expanded, minimized, or closed
• A dedicated camera/interactive mode that can be activated entirely by voice

The long-term idea is that E.V. shouldn't just answer:

>"How do I do this?"

She should eventually be able to understand what I'm doing, what I'm looking at, what I'm trying to accomplish, and help me actually do it.

I'm deliberately starting this from scratch rather than adding endless features to my previous project.

It's still early, so I'm interested in hearing from people building similar systems:

What would make an AI actually feel like a desktop assistant rather than just another LLM interface?

Hey everyone,

For the past few weeks I've been working on a desktop AI assistant called Nova AI in Python. It started as a fun project to learn more about AI and programming, but it's slowly turning into something I'm really proud of.

Right now it can:

• Chat using an LLM

• Voice input and voice responses

• Remember the conversation during a session

• Run through a Flask-based desktop web interface

I'm currently redesigning the UI, fixing bugs, improving voice recognition, and making an installer so it's easier to use.

I was wondering:

• What features would you expect from a desktop AI assistant?

• Is there anything that would make you actually use or buy something like this?

• Anyone interested to Buy?

\- if anyone is interested, I can also share screenshots or a short demo video once the UI update is finished.

I'd genuinely appreciate any feedback. Thanks!