Local Models (LM Studio and Ollama)

Run chat and embedding models locally in Sciwand with LM Studio or Ollama - no API key, no subscription, no cost.

Run models on your own machine with no API key, no subscription, and no cost - for both chat and embeddings. Everything stays on your computer. Sciwand supports two local runtimes, LM Studio and Ollama, plus a Custom Endpoint for any other OpenAI-compatible server.

LM Studio

On your computer: install LM Studio, download a model (a chat model and/or an embedding model), and start its local server (the Developer / Local Server tab).

In Sciwand:

  • For chat: open Settings -> AI Models, click Manage, select LM Studio, and click Configure.
  • For embeddings: open Settings -> Embeddings, turn on Text Embeddings, click Select, choose LM Studio, and click Configure.

Set Server URL (default http://localhost:1234/v1) and click Save Changes. Sciwand detects the loaded models; enable the ones you want. For embeddings, pick the model and click Generate All.

Ollama

On your computer: install Ollama and pull a model - for example ollama pull llama3 (chat) or ollama pull nomic-embed-text (embeddings).

In Sciwand: open the model configurator - Manage on the AI Models tab (chat) or Select / Change on the Embeddings tab - then choose Ollama, click Configure, set Server URL (default http://localhost:11434), and click Save Changes. Enable your models.

Notes

  • No API key is required for either runtime.
  • A green status dot means Sciwand can reach the server. If it stays grey, confirm the server is running and the URL and port match.
  • For any other local or self-hosted server, use the Custom Endpoint option (available on both the AI Models and Embeddings tabs).
  • One embedding model is active at a time, and switching it clears existing vectors - see Connect an Embedding Model.