Quick start

Quick start

Use this page after the source setup is complete and scripts/dev-doctor.py passes.

1. Launch LM Nexus

From the application repository root:

Linux / WSL:

.venv/bin/python run.py

Native Windows PowerShell:

.venv\Scripts\python.exe run.py

Open:

http://localhost:5173

2. Make sure the app is healthy

Before starting a model, check that the main UI loads and that the model/provider areas open normally.

If startup or a runtime fails, look at the visible logs and the Troubleshooting guide instead of starting more copies of the same process.

3. Choose how you want to run a model

For a local model managed by LM Nexus:

1. open Model Manager and choose or download a compatible model;
2. select the profile and runtime settings you want;
3. start the managed runtime;
4. wait for it to become healthy.

If you already have a self-hosted, custom, or API provider, connect that provider instead of starting another local runtime.

4. Start working

Once the model or provider is available, use it from Chat or from another part of LM Nexus that uses the configured provider.

The workspace also includes Files, Notes, Knowledge, Terminal, Git, agents, logs, and other module tools.

5. Connect directly to the runtime when you need to

A llama.cpp server started by LM Nexus exposes an OpenAI-compatible endpoint that power-user clients can use directly.

See Raw server access for the difference between talking straight to the runtime and using an LM Nexus workflow around it.

6. Stop what LM Nexus started

If LM Nexus started a local model/server, stop it through LM Nexus when you are done.

LM Nexus should only clean up processes it owns. It should not touch unrelated processes you started yourself.

Next

• Models
• Runtimes
• Providers
• Agents
• Workspace overview