Skip to main content

​
DeepSeek configuration

OpenAI-compatible model configuration in Novaplan AI DeepSeek provides an OpenAI-compatible API that Novaplan AI can use for a language model. Novaplan helps connect it to your managed workspace. You need a DeepSeek account, an API key, an approved model, and network access from the Novaplan environment to DeepSeek.
DeepSeek retired the older deepseek-chat and deepseek-reasoner API model names in July 2026. Do not copy those IDs from an older setup guide. Check DeepSeek’s current model list before configuring or renewing the connection. The model and optional features must also be tested with the deployed Novaplan AI version.

​
Required fields

The Endpoint URL, API Key, and Model Name fields are required by the Novaplan form. DeepSeek’s published base URL does not include /v1/; do not assume the suffix is required just because the provider option is named OpenAI API Compatible. Novaplan should test the endpoint from the managed workspace before customers depend on it. To obtain a key, sign in to the DeepSeek Platform, open API Keys, create a key, and copy it immediately. Ask your administrator to confirm billing, allowed data use, and the model choice. Never place the key in documentation, screenshots, chats, or source code.

​
Optional features

DeepSeek’s current deepseek-flash API supports image input and thinking mode according to its API reference and thinking-mode guide. Enable Multimodal only after Novaplan confirms that the deployed integration sends image inputs in a supported format. The Reasoning checkbox should likewise be tested: DeepSeek exposes a separate thinking control, and the checkbox alone does not prove how Novaplan sends it. Model features and names change. Check the current DeepSeek documentation and the chosen model’s capabilities whenever you change the configuration.

​
Connect the model

  1. In Novaplan AI, choose OpenAI API Compatible as the provider type.
  2. Enter the DeepSeek Endpoint URL, API Key, and exact Model Name agreed with Novaplan.
  3. Set Multimodal and Reasoning only after testing the selected model and deployed Novaplan integration.
  4. Click Add Model and run a permitted text test. If image or thinking behavior matters, test those separately.
If connection fails, check the base URL, network access from the managed workspace, API key, account balance, model ID, and DeepSeek’s current API status. For image or thinking errors, compare the request behavior with DeepSeek’s current API reference. Contact Novaplan Support for integration-specific help.