LlamaNexus WebUI

A standalone, smooth browser chat client for a LlamaNexus proxy - the same core experience as the VS Code extension, but a normal web app instead of an editor extension.

This is a client-server app, not a static site: read_file/write_file/patch_file/run_bash need real filesystem/shell access, which isn't possible from a browser tab alone. A small Node backend does that work (and talks to the LlamaNexus proxy on your behalf); the React frontend is a thin client that streams everything over a WebSocket.

Browser (React) <--WebSocket (live chat)--> Node backend --HTTP--> LlamaNexus proxy (LAN)
                <--HTTP/JSON (sessions, settings)-->

Status

All four planned phases are implemented: sessions/streaming chat/model listing/file/bash tool-calling with confirmations; Plan Mode; LAN server auto-discovery and run_agent/agent-server delegation; and Hugging Face model search, RPC worker management, the full granular settings panel, and the per-session model-tuning sidebar - matching the full feature set of the VS Code extension.

See Features for the full current feature list.

Requirements

• Docker
• A running LlamaNexus proxy reachable from this machine

Setup

Ships only as a Docker image - makershop/llamanexus-webui. See Installation for a quick docker run, or Docker for the full Compose reference.

Security

This backend executes run_bash and writes files on the strength of messages it receives over the network. It binds to 127.0.0.1 (loopback) by default - widen bindHost in Settings only on a network you trust, and only if you understand what that means.

In this section