Over the past few months, I've been building AURA, an open-source, offline, voice-powered AI assistant that runs entirely on a local machine.
The goal isn't to compete with cloud AI assistants. It's to build something that's private, extensible, and fully owned by the user.
Current stack:
Ollama (local LLMs)
Whisper (Speech-to-Text)
Wake-word detection
ChromaDB (RAG/Memory)
Piper (Text-to-Speech)
Python
The core voice pipeline is now working end-to-end:
Wake Word → Speech-to-Text → Local LLM → RAG → Text-to-Speech
I'm actively improving areas like:
Better memory and retrieval
Multi-agent workflows
Safer execution of developer commands (Git, Docker, filesystem)
Cross-platform support (Linux/macOS)
Performance optimization
Documentation and testing
This is still an early-stage project, and I'd rather build it with the community than in isolation.
I'm looking for people who want to contribute—whether that's:
Writing code
Reporting bugs
Improving documentation
Testing on different operating systems
Suggesting features
Reviewing the architecture
Pointing out flaws I haven't considered
If you enjoy building local AI tools, voice interfaces, or developer-focused open-source projects, I'd love to have you involved.
The repository is here: https://github.com/aryanjsx/AURA
Every issue, discussion, PR, or piece of feedback helps move the project forward. Don't hold back—I'd much rather hear what's wrong with it now than after release.
Top comments (0)