IntroductionGetting Started

Initializing
Vextor Engine

Welcome to the next generation of local development. Vextor AI requires no cloud APIs, zero telemetry, and offers zero friction. Follow this guide to get the multi-process engine running on your machine.

01. System Prerequisites

Because Vextor bypasses legacy Node.js terminal environments, it compiles and runs its concurrent backend locally. You must ensure your machine has the required toolchains installed for the sidecar architecture to boot.

Rust (Cargo)
v1.70.0+
Required to compile the memory-safe AST parser.
Go Runtime
v1.21.0+
Powers the concurrent PTY and WebSocket server.
Node.js
v18.0.0+
Handles the Electron wrapper and React UI rendering.
!

Windows Users (WSL2 Required)

Vextor's concurrent Go PTY is deeply optimized for UNIX-like environments. To use Vextor on Windows, you must install and run the IDE from within a Windows Subsystem for Linux (WSL2) distribution.

02. Installation

Download the universal installer. This script automatically detects your operating system, downloads the pre-compiled Rust and Go binaries, and registers the vextor CLI globally.

macOS / Linux
$curl -fsSL https://vextor.ai/install.sh | bash

Verify the installation by checking the version:

$vextor --version
Vextor Engine v2.0.4 (Polyglot Edition)

03. Initializing a Workspace

Unlike traditional editors, Vextor does not just open a folder. It boots a local server cluster. Navigate to any existing repository and run the initialization command.

$cd my-awesome-project
$vextor .
[Vextor/Core] Spawning Rust AST Engine on port 8081... SUCCESS
[Vextor/PTY] Starting Concurrent Go Terminal... SUCCESS
[Vextor/UI] Launching Electron window... READY

The .vextorignore File

Upon first boot, Vextor will automatically generate a .vextorignore file in your root directory. This tells the Rust AST Parser which directories to skip (e.g., node_modules, target/, .git) to ensure background indexing remains instantaneous.

04. Linking the Local LLM

Out of the box, Vextor AI requires a localized LLM to drive the Natural Language Terminal and predictive autocomplete. We utilize GGUF formats integrated with llama.cpp for optimized CPU/GPU hybrid inference.

  1. Download a compatible GGUF model (e.g., Llama-3-8B-Instruct.gguf) from HuggingFace.
  2. Open the Vextor AI command palette (Ctrl+Shift+P or Cmd+Shift+P).
  3. Search for "Vextor: Link Local LLM" and select your downloaded .gguf file.
  4. Wait for the engine to allocate the model to your VRAM.
ℹ

Hardware Recommendation

For seamless terminal translations and code autocompletion, we highly recommend models quantized to Q4_K_M or Q5_K_M. This provides the perfect balance of instant reasoning speed and low RAM overhead (typically utilizing ~5GB of memory).

05. Your First AI Command

Now that your LLM is linked, let's test the Natural Language Terminal. Open the integrated terminal (`Ctrl + \``) and select the "CMD + AI" profile.

Instead of trying to remember complex shell syntax, simply type your intent in plain English into the Magic Bar.

✨"Find whatever process is running on port 3000 and kill it"
> npx kill-port 3000
Process on port 3000 killed successfully.

Congratulations. You are now running a completely offline, polyglot AI development environment.

Next
Polyglot Architecture
→