Skip to main content

Run the playground against an OpenAI-compliant model provider/proxy

The LangSmith playground allows you to use any model that is compliant with the OpenAI API.

Deploy an OpenAI-compliant model​

Many providers offer OpenAI-compliant models or proxy services that wrap existing models with an OpenAI-compatible API. Some popular options include:

These tools allow you to deploy models with an API endpoint that follows the OpenAI specification. For implementation details, refer to the OpenAI API documentation.

Use the model in the LangSmith Playground​

Once you have deployed a model server, you can use it in the LangSmith Playground.

Configure the playground​

  1. Open the LangSmith Playground
  2. Change the provider to Custom Model Endpoint
  3. Enter your model's endpoint URL in the Base URL field
  4. Configure any additional parameters like API keys if required

Custom Model Endpoint

The playground uses ChatOpenAI from langchain-openai under the hood, automatically configuring it with your custom endpoint as the base_url.

Testing the connection​

Click Start to test the connection. If properly configured, you should see your model's responses appear in the playground. You can then experiment with different prompts and parameters.

Save your model configuration​

To reuse your custom model configuration in future sessions, learn how to save and manage your settings here.

Was this page helpful?


You can leave detailed feedback on GitHub.