Skip to main content
AnythingLLM is an all-in-one desktop and self-hosted AI application with workspaces, document chat (RAG), and agents. Its Generic OpenAI provider can connect to the ScitiX Model Inference API, so ScitiX models can power chat and agent workflows. Supported models include glm-5.2 and DeepSeek-V4-Flash. For model pricing, see the model catalog.

​
Configure

You can configure AnythingLLM through the onboarding UI, the settings page, or server/.env for a self-hosted deployment.

​
1. Configure the Generic OpenAI provider

During onboarding, or later under Settings → AI Providers → LLM, choose Generic OpenAI as the LLM provider and fill in the required fields. When configuring workspace chat mode:
  • When the workspace chat mode is left on its default and the model supports native tool calling, AnythingLLM routes to its agent flow.
  • Set the workspace chat mode to chat for plain conversation, or keep the default for agent and tool use.
  • Choose a model with native tool calling if you rely on agents or document-tool features. glm-5.2 supports tool calling.
AnythingLLM — Generic OpenAI provider pointed at ScitiX

​
2. Configure server/.env for self-hosted deployments

For a self-hosted deployment, either bare-metal or Docker, set the same values in server/.env:
Use https://api.scitix.ai/model-api/v1 with /v1. AnythingLLM’s Generic OpenAI provider passes this base path to the OpenAI SDK, which appends /chat/completions.

​
Verify

Create a workspace, confirm that the model shown at the top is your ScitiX model, and send a message. A normal reply confirms that the connection works. AnythingLLM — chatting with a ScitiX model in a workspace