Skip to main content
Obsidian Copilot is an AI plugin for Obsidian. It lets you chat with your notes, generate content, and run commands. Its custom-model support connects to the ScitiX Model Inference API through an OpenAI-compatible endpoint. The steps below reflect the plugin’s current settings UI. Newer versions may differ slightly. For available models, such as glm-5.2 and DeepSeek-V4-Flash, and pricing, see the model catalog.

​
Configure

Create an API Key on the API Keys page first.
  1. Open Settings → Copilot → Model, then click to add a model. The Add Custom Chat Model dialog opens.
  2. Fill in the required fields. Keep the following in mind:
    • Provider must be OpenAI Format. This is the option for a custom OpenAI-compatible endpoint. The named providers, such as OpenAI or Anthropic, target their official services.
    • The Base URL field is hinted as “leave it blank unless you are using a proxy”, but for a custom endpoint like ScitiX you must fill it with https://api.scitix.ai/model-api/v1.
    • Choose a model with native tool calling if you use Copilot features that rely on it.
  3. Tick the CORS checkbox under Model Capabilities. The desktop app requires CORS for a custom endpoint. Otherwise, the Test button reports “Connection successful, but requires CORS to be enabled”.
  4. Click Test to verify the connection. Expect “Model verification successful!”, then click Add Model.
Obsidian Copilot — Add Custom Chat Model dialog pointed at ScitiX

​
Verify

Open the Copilot chat pane, select glm-5.2 from the model selector at the bottom, and send a message. A normal reply confirms that the connection works.
The assistant may identify itself as “Obsidian Copilot” rather than the underlying model. That is the plugin’s own system prompt, not a routing issue; the model selector still shows the ScitiX model in use.
Obsidian Copilot — chatting with a ScitiX model