---
title: Llama Cpp Wasm
emoji: 🐨
colorFrom: green
colorTo: red
sdk: static
pinned: false
---

llm.js lab - static bundle
build commit: 1ab62ba
llama.cpp:    f7591a6a01f9273de03331732bdc0b39721168be
built:        2026-11-08

Serve this directory with any static server and open index.html:
    python3 -m http.server 8000 --directory .
Models are not bundled: add a GGUF by URL (e.g. Hugging Face) or pick one
from your disk. Everything else runs locally in the browser (wasm32, CPU).
No COOP/COEP headers are needed (single-threaded build).

This build adds, over the 0addf13 bundle:
  - tools: paste an OpenAI tools array, replies are split by the model's own
    PEG parser and calls shown as JSON with "Add tool response"
  - chat template params (chat_template_kwargs, e.g. enable_thinking)
  - conversation export as OpenAI JSON or Markdown