Models
Model providers, embeddings and the model that reads images.
Administrators manage the installation from the Admin area. At Admin → Models add one or more providers:
- Anthropic, or Anthropic-compatible proxies;
- OpenAI;
- OpenAI-compatible servers;
- Ollama, to run models on your own hardware.
Try a model in the playground before offering it to people.
Provider keys are encrypted with ORLOK_SECRET_KEY. Semantic search in the wiki uses the built-in
multilingual embedding model; you can choose an OpenAI-compatible embedding provider instead. You can
also choose a model that reads images, to transcribe scanned pages and describe figures: those images
are then sent to its provider.