WEBVTT 1 00:00:00.000 --> 00:00:02.560 Welcome back! 2 00:00:02.560 --> 00:00:08.840 In this video, you'll learn to make your very first request to the OpenAI API! 3 00:00:08.840 --> 00:00:08.880 4 00:00:08.880 --> 00:00:08.880 5 00:00:08.880 --> 00:00:15.760 Recall that we can request the use of a model from OpenAI by making a request to their API. 6 00:00:15.760 --> 00:00:15.760 7 00:00:15.760 --> 00:00:15.760 8 00:00:15.760 --> 00:00:21.600 Depending on the model or services required, APIs have different access points for users. 9 00:00:21.600 --> 00:00:21.600 10 00:00:21.600 --> 00:00:22.320 11 00:00:22.320 --> 00:00:24.880 These access points are called endpoints. 12 00:00:24.880 --> 00:00:25.760 13 00:00:25.760 --> 00:00:25.760 14 00:00:25.760 --> 00:00:29.120 Endpoints are a lot like doors in a hospital. 15 00:00:29.120 --> 00:00:34.720 Depending on the treatment required, patients use different doors to reach different departments. 16 00:00:34.720 --> 00:00:34.720 17 00:00:34.720 --> 00:00:35.680 18 00:00:35.680 --> 00:00:43.800 API endpoints, like many hospital departments, may also require authentication before accessing services. 19 00:00:43.800 --> 00:00:43.800 20 00:00:43.800 --> 00:00:43.840 21 00:00:43.840 --> 00:00:50.920 API authentication is usually in the form of providing a unique key containing a random assortment of characters. 22 00:00:50.920 --> 00:00:50.920 23 00:00:50.920 --> 00:00:51.920 24 00:00:51.920 --> 00:01:00.000 It's important to note that many APIs, including the OpenAI API, have costs associated with using their services. 25 00:01:00.000 --> 00:01:00.000 26 00:01:00.000 --> 00:01:00.840 27 00:01:00.840 --> 00:01:07.720 For OpenAI, these costs are dependent on the model requested and on the size of the model input and output. 28 00:01:07.720 --> 00:01:09.000 29 00:01:09.000 --> 00:01:09.000 30 00:01:09.000 --> 00:01:15.160 A crucial part of working with APIs is navigating API documentation, which provides 31 00:01:15.160 --> 00:01:22.240 details on which endpoints to use, their functionality, and how to set up authentication. 32 00:01:22.240 --> 00:01:22.240 33 00:01:22.240 --> 00:01:22.240 34 00:01:22.240 --> 00:01:24.080 Throughout the course, we'll provide 35 00:01:24.080 --> 00:01:30.440 these API details, but we also recommend checking out OpenAI's excellent API documentation. 36 00:01:30.440 --> 00:01:30.440 37 00:01:30.440 --> 00:01:31.560 38 00:01:31.560 --> 00:01:37.320 Usage of OpenAI's API requires authentication, so to use the API and 39 00:01:37.320 --> 00:01:42.880 complete the exercises in this course, you'll need to set up an account. 40 00:01:42.880 --> 00:01:42.880 41 00:01:42.880 --> 00:01:42.880 42 00:01:42.880 --> 00:01:47.800 OpenAI often provides free trial credit to new users, which will be more than 43 00:01:47.800 --> 00:01:53.240 sufficient for completing this course; however, for some countries, you may need to add credit. 44 00:01:53.240 --> 00:01:53.240 45 00:01:53.240 --> 00:01:53.240 46 00:01:53.240 --> 00:01:58.920 Then, for authentication, you'll need to create a secret key and copy it. 47 00:01:58.920 --> 00:02:06.920 DataCamp doesn't store any API keys used in this course, so you can copy it directly into the exercises. 48 00:02:06.920 --> 00:02:07.720 49 00:02:07.720 --> 00:02:07.720 50 00:02:07.720 --> 00:02:10.040 Let's make our first API request! 51 00:02:10.040 --> 00:02:10.040 52 00:02:10.040 --> 00:02:10.040 53 00:02:10.040 --> 00:02:14.560 There are several ways to interact with an API, but in this 54 00:02:14.560 --> 00:02:20.080 course, we'll use OpenAI's own Python library, due to its friendly syntax. 55 00:02:20.080 --> 00:02:21.000 56 00:02:21.000 --> 00:02:21.000 57 00:02:21.000 --> 00:02:27.600 We assign our API key to openai-dot-api_key, which is used to authenticate our request. 58 00:02:27.600 --> 00:02:27.600 59 00:02:27.600 --> 00:02:27.600 60 00:02:27.600 --> 00:02:30.480 Now for the request code. 61 00:02:30.480 --> 00:02:34.320 We create a request to the Completions endpoint, which is used for 62 00:02:34.320 --> 00:02:40.080 completing a text prompt, by calling the create method on openai-dot-Completion. 63 00:02:40.080 --> 00:02:41.160 64 00:02:41.160 --> 00:02:41.160 65 00:02:41.160 --> 00:02:45.440 Inside this method, we specify the model and the prompt to send to it. 66 00:02:45.440 --> 00:02:50.200 We'll discuss prompts in greater detail later in the course. 67 00:02:50.200 --> 00:02:50.200 68 00:02:50.200 --> 00:02:50.200 69 00:02:50.200 --> 00:02:52.480 Let's take a look at the API response. 70 00:02:52.480 --> 00:02:52.480 71 00:02:52.480 --> 00:02:53.400 72 00:02:53.400 --> 00:02:56.200 The API returns a JSON response. 73 00:02:56.200 --> 00:03:01.960 JSON stands for JavaScript Object Notation; it's a nested data 74 00:03:01.960 --> 00:03:05.960 format that resembles a Python dictionary with keys and values. 75 00:03:05.960 --> 00:03:13.880 The response has several keys: "choices" at the top, and "created", "id", "model", and others at the bottom. 76 00:03:13.880 --> 00:03:13.880 77 00:03:13.880 --> 00:03:13.880 78 00:03:13.880 --> 00:03:18.760 We can see that the text response to our prompt is nested inside the 79 00:03:18.760 --> 00:03:24.800 "choices" key; let's start to unnest this object by selecting the value from that key. 80 00:03:24.800 --> 00:03:24.800 81 00:03:24.800 --> 00:03:24.800 82 00:03:24.800 --> 00:03:30.400 Recall that we can extract the value from a dictionary key using square brackets. 83 00:03:30.400 --> 00:03:30.400 84 00:03:30.400 --> 00:03:30.400 85 00:03:30.400 --> 00:03:35.960 This returns another nested object, but we've gotten much closer to the text. 86 00:03:35.960 --> 00:03:42.560 Notice from the square brackets at the beginning and end, that this is actually a list with a single element. 87 00:03:42.560 --> 00:03:42.560 88 00:03:42.560 --> 00:03:42.560 89 00:03:42.560 --> 00:03:46.520 Let's extract the first element to dig further. 90 00:03:46.520 --> 00:03:46.520 91 00:03:46.520 --> 00:03:46.520 92 00:03:46.520 --> 00:03:54.400 Ok - now we're left with a standard dictionary, and our text response is located underneath the "text" key. 93 00:03:54.400 --> 00:03:54.400 94 00:03:54.400 --> 00:03:54.400 95 00:03:54.400 --> 00:03:57.680 Let's use dictionary subsetting once more. 96 00:03:57.680 --> 00:03:57.680 97 00:03:57.680 --> 00:03:57.680 98 00:03:57.680 --> 00:04:01.680 There we have it - our text response as a string! 99 00:04:01.680 --> 00:04:06.560 We started off with a complex object, but by applying our Python 100 00:04:06.560 --> 00:04:10.720 programming fundamentals, we were able to get there one step at a time. 101 00:04:10.720 --> 00:04:10.720 102 00:04:10.720 --> 00:04:10.720 103 00:04:10.720 --> 00:04:16.120 Time to practice making your own requests!