Local model optimization and setup

Choose and configure a runnable local model for your hardware.

Preview a hardware-aware recommendation and setup command:

arka model-optimizer recommend
arka model-optimizer setup --backend ollama

Apply the provider configuration explicitly with --apply:

arka model-optimizer setup --backend ollama --apply

Use vllm for a configured GPU server or lmstudio for its OpenAI-compatible desktop server. Arka does not download weights or start services implicitly. It prints the next command so setup remains explicit and reversible.

For deployment previews, Arka supports Vercel, Netlify, Railway, Render, Hugging Face Spaces, Cloudflare Workers, and GitHub Pages:

arka deploy --platform huggingface --json