Skip to main content

browser-local-ai-webllm

Run local LLMs directly in the browser via WebLLM (WebGPU) with NO middleware app — no Ollama, no PocketPal, no PC. Covers lazy-loading the ESM lib, WebGPU/HTTPS requirements, device analysis (RAM/VRAM), model-fit ranking, the Chrome 8GB deviceMemory cap, and OpenAI-compatible call/stream. ALWAYS use when adding in-browser local AI, running models client-side without a server, or the user says: локальний AI у браузері, WebLLM, без посередників, модель прямо в застосунку, WebGPU AI, браузерний AI, офлайн модель у браузері, run model in browser, on-device browser inference. Also triggers for: device analysis for model fit, VRAM detection, prebuiltAppConfig, MLCEngine. DO NOT use for server-side inference, native mobile inference, or when Ollama/LM Studio is the intended runtime (those are external providers).

Jump to install

Source facts

Repository
PatriotAi/ai-lab
Last source activity
July 26, 2026 at 15:26
Detected SKILL.md language
Ukrainian
Stars
5
Forks
0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.