GitHub Copilot + OAI Plugin

Use Cases

When using GitHub Copilot in programming tools such as VS Code and Cursor, if your installed OAI plugin supports configuring an OpenAI-compatible endpoint, you can forward model requests to New API or another compliant third-party relay service.

This approach is suitable for the following scenarios:

Prerequisites

Prepare the following before you begin:

1. VS Code, Cursor, or another programming tool that supports extensions is installed.
2. GitHub Copilot is installed, and you are signed in.
3. An OAI plugin that supports custom OpenAI-compatible endpoints is installed.
4. An API Key has been created in New API or a third-party relay service.
5. The available API endpoint and model name have been confirmed.

Common configuration options are as follows:

Install the Plugins

1. Install GitHub Copilot

Open the Extensions Marketplace in VS Code, then search for and install GitHub Copilot. After installation, follow the prompts to sign in to your GitHub account and verify that Copilot is working properly.

If you use a programming tool such as Cursor, Trae, or Windsurf, confirm that Copilot functionality is enabled on the corresponding extension or plugin page.

2. Install the OAI Plugin

Search for OAI in the Extensions Marketplace and install the OAI plugin you intend to use. Configuration option names may vary slightly between plugins, but they typically include the following fields:

• Base URL or API Base
• API Key
• Model
• Provider or Endpoint Type

If the plugin requires you to select an endpoint type, choose OpenAI, OpenAI Compatible, or a similar option.

Configure the Third-Party Relay API

1. Open the OAI Plugin Settings

In VS Code, you can open the settings as follows:

1. Press Ctrl + Shift + P or Command + Shift + P to open the Command Palette.
2. Search for the plugin name or OAI.
3. Select the Settings, Configure Provider, or similar command provided by the plugin.

Alternatively, open Settings, search for OAI, and locate the configuration options for the plugin.

2. Enter the API Endpoint

Set Base URL to the OpenAI-compatible endpoint of your relay service, for example:

https://api.example.com/v1

If you use a self-hosted New API deployment, replace the domain with your own service address. HTTPS is recommended, and you should ensure that the trailing path matches the plugin's requirements.

3. Enter the API Key

Enter the Token created by the relay service in the API Key field, for example:

sk-xxxxxxxxxxxxxxxx

Do not store the API Key in a public repository, screenshot, or shared configuration file. If the plugin supports the system keychain or Secret Storage, use it whenever possible.

4. Select a Model

Enter the name of a model actually supported by the relay service in the Model field, for example:

gpt-4o-mini

The model name must exactly match the relay service's model list. If the request fails, first check in the New API console whether the model is enabled, whether an available Channel exists, and whether the current Token has permission to access the model.

5. Save and Reload the Editor

After saving the settings, it is recommended that you reload the window:

1. Open the Command Palette.
2. Enter Developer: Reload Window.
3. Reopen a project file and test whether the OAI plugin responds correctly.

Test the API Call

Open any code file in the editor and send a simple request to the OAI plugin, for example:

解释当前函数的作用，并给出一个更清晰的命名建议。

Then check the New API or third-party relay service dashboard for the following:

• Whether a new request appears in the logs.
• Whether the correct model name was used.
• Whether the status code indicates success.
• Whether the Quota consumption matches expectations.

If the request appears in the dashboard logs, the plugin has successfully called the API through the relay service.

Troubleshooting

401 or Unauthorized

This usually means that the API Key is incorrect, the Token has been disabled, or extra spaces were introduced during copying. Generate a new Token and paste it into the plugin settings.

404 or model not found

This usually means that the model name is incorrect or the model is not enabled in the relay service. Copy the full model name from the model list and try again.

connection error

Check whether the API Base URL is reachable, starts with https://, requires the /v1 suffix, and whether your local network can connect to the relay service domain.

GitHub Copilot Itself Does Not Use the Relay Service

The built-in completion and chat services in the official GitHub Copilot extension typically use GitHub's official infrastructure. The OAI plugin only affects OpenAI-compatible requests initiated by the plugin itself and may not modify network requests made by the official Copilot extension.

The Plugin Does Not Have a Base URL Option

This indicates that the plugin may not support custom OpenAI-compatible endpoints. Use a plugin version that supports OpenAI Compatible, Custom Endpoint, or Base URL.

Recommended Configuration Checklist

After completing the configuration above, you can use the OAI plugin in your programming tools to call a third-party relay API and combine it with GitHub Copilot for routine code generation, explanation, refactoring, and test creation.