LOCAL_AI_STACK / home
v2.2 · 2026-06-12NAME
LOCAL_AI_STACK
Local LLM guide for running private AI models on your own hardware
SYNOPSIS
local-ai --hardware <mac|windows> --ram <GB> --vram <GB> --model <name> --quant [q4_k_m|q8_0] --context<tokens>
DESCRIPTION
A practical local LLM guide for choosing between Ollama, LM Studio, and Open WebUI on Mac and Windows, with local-first workflows that still depend on your runtime, provider, embedding, storage, sync, and network settings.
Evidence stays separate from enthusiasm. Estimates are labeled. Local-only claims require the whole stack to stay local: model, documents, embeddings, storage, providers, and network settings.
TOOLS
#01
RAM/VRAM Calculator
Deterministic estimate for local LLM memory fit. Enter your hardware and model size — get a clear fit/no-fit result before downloading anything.
$check-hardware --ram <GB> --model <name>→
#02
Interactive Comparison
Compare 2–4 local AI tools side by side with filters, source links, and explicit caveats for every claim.
$compare --tools ollama lmstudio→
#03
Stack Recommender
Deterministic stack guidance based on your hardware, use case, and priorities. Reuses the RAM/VRAM calculator for consistency.
$recommend-stack --use-case chat --ram 16→
#04
Cloud vs Local Cost
Source-backed cost estimator covering tokens, hardware, electricity, storage, and amortization. Shows the breakeven point.
$cost-estimate --tokens-per-day 100k→