Introducing VoxAI v1.7: The Local-First "Vibe Coding" Engine

[AlexC1991]

👋 Hi everyone,

I'm excited to share the latest release of VoxAI, an IDE built for a new way of working we call "Vibe Coding."

Most AI coding tools are just "smart autocomplete." They suggest lines, but you still have to drive. VoxAI is different. It's designed to be a fully autonomous agent that lives on your machine, reads your entire project, and executes high-level architectural intent while you direct the flow.

🌟 What's Different in v1.7?

1. 100% Local & Private (No Cloud Required)
   We've integrated a native inference engine. You can now drop standard .gguf models (Llama-3, Mistral, Gemma) directly into the
   models/
   folder and run the entire agent offline. No data ever leaves your machine.

2. "Siege Mode" (Autonomous Looping)
   In this new mode, the agent doesn't just write code once. If it encounters an error, it reads the traceback, creates a fix, and retries in a continuous loop until the task is complete. It's like having a senior engineer on your local machine who doesn't give up.

3. Deep Context (RAG)
   We've removed the complex networking setup. The new vector engine indexes your project locally in the background, allowing the agent to "remember" code from across your entire repository instantly.

🛠️ How to Try It
Download the latest release from the [Releases Page].
Run
setup_integration.bat
.
Drop a model into models/llm/ (or use your OpenAI key if you prefer cloud).
Start coding with start_IDE.bat.
💭 We Want Your Feedback!
This is a Beta release of the Agentic core. We're looking for feedback on:

Which local models perform best for you?
Are the "Safety Gates" in the chat too strict or too loose?
What features should the agent automate next?
Let us know in the comments below! 👇