Lumy Desktop AI Documentation
Downloads

AI setup

The AI assistant is the heart of Lumy. It can read, search, move, rename, archive, and delete files — but first it needs an AI model to "think" with.

You have two choices:

  • Local model (recommended): runs on your own PC. Private, free, works offline. No account needed.
  • Cloud model: runs on a provider's server. More powerful, but your chat text is sent to that provider.

Choose your provider

Lumy supports five providers:

Provider Where it runs Account needed? Best for
LM Studio Your PC (local) No Privacy, offline use
Ollama Your PC (local) No Privacy, offline use
OpenAI Cloud Yes Maximum quality
OpenRouter Cloud Yes Access to many models
Anthropic (Claude) Cloud Yes Advanced reasoning

Set up a local model (LM Studio or Ollama)

Local models keep everything on your machine. Nothing you type or ask is ever sent to the internet.

Step 1: Install a local server

Download and install either:

Step 2: Download a model

Inside LM Studio or Ollama, download a chat model. Any 7B–13B model works well. On an older or smaller machine, a 3B–7B model will feel faster.

Step 3: Start the server

  • LM Studio: open the "Local Server" tab and click Start.
  • Ollama: the server runs automatically once installed.

Step 4: Connect Lumy to the model

  1. Open Settings (the gear icon in the left sidebar).
  2. Scroll to the LLM Settings section.
  3. In the Provider dropdown, choose LM Studio or Ollama.
  4. The Endpoint is filled in automatically (LM Studio uses http://127.0.0.1:1234, Ollama uses http://127.0.0.1:11434).
  5. Click Test Connection. Lumy will check the server and list the models it finds.
  6. Pick your model from the dropdown.

If Test Connection succeeds, you are ready. Open the File Organizer and the chat status will show Online (green dot).

Set up a cloud model

  1. Open Settings → LLM Settings.
  2. Choose OpenAI, OpenRouter, or Anthropic as the provider.
  3. Enter your API key (you get this from the provider's website).
  4. The endpoint is usually filled in automatically.
  5. Leave the Model box empty to use your account's default, or type a specific model name.
  6. Click Test Connection to verify.

When using a cloud provider, the text of your chat messages — and the text read from documents you ask about — is sent to that provider's servers. Choose local models if you want everything to stay on your PC.

Common problems

  • The chat shows "Offline" — the model server is not running, or the endpoint is wrong. Check Settings → LLM Settings and click Test Connection.
  • No models appear — make sure you downloaded a model and the server is running before connecting.
  • "Test Connection" fails with a local server — check that the server (LM Studio or Ollama) is actually started.

See Troubleshooting for more help.