Unity request to openAI API - returns empty text

I’m new to web requests but after some time I finally was able to create a request in Unity (C#) without getting back errors.

I am using the following code and got a response (down below) with, as you can see, empty text.

As ChatGPT can’t help me out here atm, I kindly ask for some help from the community.

Best regards

        private void ChatGPTRequest()
            var API_URL = "https://api.openai.com/v1/completions";
            var API_KEY = myKey;

            string prompt = "What day is today?";

            string requestData = "{\"model\": \"text-davinci-003\", " +
                                 "\"prompt\": \"" + prompt + "\", " +
                                 "\"max_tokens\": 7, " +
                                 "\"temperature\": 0.7, " +
                                 "\"top_p\": 1, " +
                                 "\"n\": 1, " +
                                 "\"stream\": false, " +
                                 "\"logprobs\": null, " +
                                 "\"stop\": \"\\n\"}";

            UnityWebRequest request = UnityWebRequest.Post(API_URL, "");
            byte[] bodyRaw = Encoding.UTF8.GetBytes(requestData);
            request.uploadHandler = new UploadHandlerRaw(bodyRaw);

            request.downloadHandler = new DownloadHandlerBuffer();
            request.disposeUploadHandlerOnDispose = true;
            request.disposeCertificateHandlerOnDispose = true;
            request.disposeDownloadHandlerOnDispose = true;

            request.SetRequestHeader("Authorization", "Bearer " + API_KEY);
            request.SetRequestHeader("Content-Type", "application/json");

            request.SendWebRequest().completed += operation =>
                if (request.result == UnityWebRequest.Result.ConnectionError || request.result == UnityWebRequest.Result.DataProcessingError)
                    string responseInfo = request.downloadHandler.text;
                    Debug.Log("Response: " + responseInfo); 

Response: {“id”:“cmpl-70vtw0ROFo2d9EFNK5wXjwRvMinCA”,“object”:“text_completion”,“created”:1680456928,“model”:“text-davinci-003”,“choices”:[{“text”:“”,“index”:0,“logprobs”:null,“finish_reason”:“stop”}],“usage”:{“prompt_tokens”:5,“total_tokens”:5}}

I forget if max_tokens includes the prompt or not. But if the prompt is using up 5 tokens, and it can only use seven total, maybe its giving up, instead of trying to respond with only the two tokens it has left. I’d try bumping it up to like 20 tokens

Or, maybe your prompt is causing it to start with a newline, but you are also te lling it to stop after a newline. But, because of the response it returns, I dont think it is a problem with your actual request. it would be givingg and error otherwise, like “bad formatted json” or something

1 Like