Can ChatGPT plugin is designed to support a single API endpoint. This endpoint can accept a variety of requests, which can be specified using optional parameters or through string matching with predefined enumerations. Additionally, there’s flexibility to include extra details through other open parameter types. Once the API receives a request, it can internally route it to the appropriate internal API call, fetch the response, and relay it back to the GPT. The response from this API is expected to be a un-structured JSON object.
For instance, consider an operation named “Multiplication.” This single operation can be set up to handle tasks like addition, subtraction, and square root based on the description provided. If a request aligns with the description, GPT can direct it to the “/Multiply” endpoint. Here, the parameter key can be either “Add” | “Subtract” | “Sqrt” and the associated values could be an array of numbers. The resulting response could be a numerical value or a combination of strings formatted as a JSON object.
Will this approach fulfill the objective? Certainly, there will be significant logic to manage on the API side. If we’re dealing with around 30 to 40 or even more downstream APIs to achieve all the user concerns, will is this a viable method? And will this the optimal way to proceed?