Use-case stack · privacy-minded makers

Local / offline AI

Run models on your own hardware for privacy, labs, and air-gapped benches.

Start with Ollama + a 7B–14B class model, then add Open WebUI if you want a browser UX for the lab.

Preferred / primary

Self-hosted LLM / chat

Ollama

Simplest path to run open models locally (CLI + local API).

In this stack: Model runner + API on the workstation or lab mini-PC.

Details Site

Self-hosted LLM / chat

Open WebUI

Self-hosted ChatGPT-like UI over Ollama / OpenAI-compatible backends.

In this stack: Browser UI for the lab / family server.

Details Site

Strong alternatives

App / desktop LLM / chat

LM Studio

Desktop app to download and run local GGUF models with a chat UI.

In this stack: Desktop-first local chat without Docker.

Details Site

Self-hosted LLM / chat

LocalAI

Drop-in OpenAI-compatible gateway for local models (LLM, audio, image backends).

In this stack: OpenAI-compatible gateway for mixed backends.

Details Site

Supporting tools

Self-hosted LLM / chat

llama.cpp

Run GGUF models efficiently on CPU/GPU — foundation under many local tools.

In this stack: Efficient GGUF core under many UIs.

Details Site

Self-hosted LLM / chat

vLLM

High-throughput open-source LLM serving engine for your own GPUs.

In this stack: When you have real GPUs and multi-user load.

Details Site

Self-hosted LLM / chat

text-generation-webui

Oobabooga’s classic local LLM web UI with many loader backends.

In this stack: Power-user UI with many loaders.

Details Site

Hybrid Research

AnythingLLM

Desktop/self-host RAG chat over your documents with multiple model backends.

In this stack: Local/desktop RAG over lab docs.

Details Site

Self-hosted Image

ComfyUI

Node-based local image/video workflows — maximum control for power users.

In this stack: Local image pipelines when you have GPU headroom.

Details Site

Self-hosted Audio

Whisper (OpenAI / faster-whisper)

Speech-to-text models you can run locally for captions and search.

In this stack: Offline transcription.

Details Site

Self-hosted Audio

Piper TTS

Fast local neural TTS — great for offline spoken notes and accessibility.

In this stack: Local speech for accessibility/notes.

Details Site

Hybrid Coding

Continue.dev

Open-source coding assistant that can point at local or cloud models.

In this stack: Point the IDE at local OpenAI-compatible endpoints.

Details Site

Stacks are editorial starting points — swap tools when policy, budget, or hardware demands it.