ONLYOFFICE DocSpace 3.7 released: File generation & new providers in AI agents, smarter forms, room grouping, and more
ONLYOFFICE DocSpace 3.7 released

DeepSeek

DeepSeek is a family of AI models developed by DeepSeek Artificial Intelligence Co., Ltd., which specializes in natural language processing (NLP) and machine learning.

This is a step-by-step guide on installing and configuring DeepSeek for ONLYOFFICE Docs.

Connecting and configuring DeepSeek models

  1. Follow this guide for general configuration.
  2. Once the plugin is installed, go to the AI tab and click the Settings icon at the top of the panel.
  3. Choose the task you want to assign an AI model to, e.g., Chatbot or Summarization. In the drop-down list for that task, click Add AI model.
  4. In the window that appears, enter a name for the model, select DeepSeek as the provider, and enter the API URL and your API key.
    • The API URL for DeepSeek is https://api.deepseek.com.
    • The API key is issued by DeepSeek and can be generated on the DeepSeek official website.
    DeepSeek AI: dropdown list DeepSeek AI: dropdown list
  5. The model list will update. Select the model you want to use from the available options:
    • DeepSeek Chat / Deepseek-ai/DeepSeek-V3 — a general-purpose conversational model offering fluid, context-aware dialogue. Suitable for customer support, brainstorming, writing, coding, and real-time assistance in document creation. When connecting via the official DeepSeek API, this model is listed as DeepSeek Chat; when connecting through TogetherAI, it appears as Deepseek-ai/DeepSeek-V3.
    • DeepSeek Reasoner / Deepseek-ai/DeepSeek-R1 — a reasoning-focused model specialized in logical inference, multi-step problem solving, complex coding, and structured decision support. Particularly useful for deep analytical insights and mathematical reasoning. When connecting via the official DeepSeek API, this model is listed as DeepSeek Reasoner; when connecting through TogetherAI, it appears as Deepseek-ai/DeepSeek-R1.
  6. Select what you want to use the model for. DeepSeek models connected via the official API are best suited for Text and Coding help. Other task types such as Images, Embeddings, Audio processing, and Visual analysis are not supported through the official DeepSeek API endpoints and will result in errors if selected. For Content moderation and Real-time tasks, check your provider's documentation for support. If you are connecting through TogetherAI, available task types may differ, please refer to TogetherAI's documentation for details. Avoid selecting All unless you have verified that your provider supports every task type.
    DeepSeek AI: using models DeepSeek AI: using models
  7. Click OK. The model is now configured and ready to use.

Once set up, access the AI features through the AI tab or by selecting text and right-clicking to open the context menu.

Article with the tag:
Browse all tags