How to Use API Function Calling Using OpenAI Assistant

You can make third-party APIs available to your Assistant by providing details about the endpoints, parameters, and a description about how the model should use it. It is similar to Custom Actions in OpenAI GPTs, but it functions within the OpenAI Assistant.

Understanding How Function Calling Works

Check OpenAI Assistant Function Calling to learn how it works.

Adding API Function Calling to Assistant

Runbear supports API calls in OpenAI Assistant. This can be achieved by simply adding the __api__ property under parameters.properties in the Function JSON.

__api__

**api** object consists of the following properties:

• type: This should be object.
• value: This should be an object with the following properties:
  • url: URL of the API endpoint
  • method (optional): HTTP method to use (GET, POST, PUT, DELETE, etc.)
  • headers (optional): HTTP headers to send

URL String Interpolation

You can interpolate parameters into the URL. For example, to interpolate a resource parameter into the URL, use the {resource} string within the URL.

Example

This example shows how to add a REST API to your Assistant Function. The function retrieves information of people in Star Wars.

{
  "name": "search_starwars",
  "parameters": {
    "type": "object",
    "properties": {
      "search": {
        "type": "string",
        "description": "Search keyword"
      },
      "resource": {
        "type": "string",
        "description": "Type of searching resource. The value should be one of 'people', 'planets' or 'starships'."
      },
      "__api__": {
        "type": "object",
        "value": {
          "url": "https://swapi.dev/api/{resource}/",
          "method": "GET",
          "headers": {
            "Content-Type": "application/json"
          }
        }
      }
    },
    "required": ["search", "resource"]
  },
  "description": "Search about the keyword from Starwars"
}

This enables OpenAI Assistant to call the API endpoint with extracted parameters.