Local AI Setup

Overview

Vipr's desktop app can run AI inference locally using GGUF-format language models. Your code never leaves your machine.

Requirements

• Vipr Desktop (Pro license)
• 8GB+ RAM recommended
• A compatible GGUF model file

Setup

1. Open Vipr Desktop settings
2. Navigate to the AI section
3. Select or download a GGUF model
4. Vipr will load the model using node-llama-cpp

Supported models

Vipr offers a curated catalog of GGUF models with verified checksums, downloaded on demand:

• GPT-OSS 20B (recommended): OpenAI's open-weight model, strong coding quality on 16GB+ machines (~12GB download)
• Qwen3 Coder 30B A3B: the strongest local option, for 24GB+ machines (~18.6GB download)
• Qwen2.5 Coder 7B: lightweight option for 8–16GB machines (~4.7GB download)

Models that need more memory than your machine has are hidden automatically.

• Phi-3 Mini 3.8B (fastest, lower quality)