RAGFlow

RAGFlow is an open-source RAG engine built on deep document understanding. You can build knowledge bases from your documents and chat over them. Its OpenAI-API-Compatible provider points both chat and embedding models to the ScitiX Model Inference API. Deploy RAGFlow with Docker Compose by following the official guide, then open the web UI. The default URL is http://localhost. Supported chat models include glm-5.2 and DeepSeek-V4-Flash. The embedding model in this guide is Qwen/Qwen3-Embedding-8B. See the model catalog.

​

Configure

Create an API Key on the API Keys page, then open Model providers from the top-right avatar menu. Find OpenAI-API-Compatible and add it. Click List models to fetch the available models and confirm connectivity, then click Ok. Add the models you need and choose the correct model type for each. RAGFlow keeps chat and embedding models in separate lists. Add glm-5.2 as chat and Qwen/Qwen3-Embedding-8B as embedding. Choose a chat model with native tool calling if you use agent features. Back on the Model providers page, under Set default models, set LLM to glm-5.2 and Embedding to Qwen/Qwen3-Embedding-8B.

​

Verify

Create a chat assistant from Chat → Create an Assistant, or use the default one. Confirm that its model is glm-5.2, then send a message. A normal reply confirms that the connection works. For document RAG, create a Dataset with Qwen/Qwen3-Embedding-8B as the embedding model, upload and parse a document, then bind the dataset to a chat assistant.