Skip to main content

Connect a Third-Party Model to ZCode: DeepSeek Example

This is only one configuration example

ZCode is not limited to DeepSeek. Other providers with compatible OpenAI or Anthropic APIs can be added with the same four items: Base URL, API format, API Key, and model ID.

1. Open Model Settings​

Go to:

Gear icon → Settings → Model Settings

Select the existing DeepSeek provider, or click Add Provider to create one.

2. Enter the DeepSeek API Details​

Follow the provider's current documentation. This example uses:

SettingExample
ProviderDeepSeek
Base URLhttps://api.deepseek.com
API formatResponses (/responses)
API KeyYour DeepSeek API Key
Model IDdeepseek-v4-flash-vision-exp

See the DeepSeek Responses API documentation.

ZCode DeepSeek provider settings showing the Base URL, Responses API format, hidden API Key, and model IDs

3. Enable Image Input for the Vision Model​

When editing deepseek-v4-flash-vision-exp:

  • under Input Types, select Text and Image;
  • under Output Types, select Text;
  • click Save.

Without Image selected, screenshots uploaded in chat may not be passed to the model as visual input.

ZCode model editor for deepseek-v4-flash-vision-exp with Text and Image selected as inputs and Text selected as the output

4. Enable and Test the Model​

  1. Return to the DeepSeek provider page and confirm that it shows Enabled;
  2. create a new task;
  3. select DeepSeek/deepseek-v4-flash-vision-exp from the model menu;
  4. upload an image and ask a question to confirm that the image can be read.

Other Providers Use the Same Pattern​

Replace these values with the ones supplied by the provider you want to use:

Base URL
+ API format
+ API Key
+ model ID
+ input capabilities (text / image / video)

If a field is unclear, follow that provider's API documentation. Do not copy the DeepSeek URL or model ID into an unrelated provider.