WEBVTT 00:00.030 --> 00:01.920 -: Okay, so after you've set up billing 00:01.920 --> 00:03.990 and also you've got your API keys, 00:03.990 --> 00:05.160 the next thing we're gonna have a look at 00:05.160 --> 00:06.600 is the playground's in the top right. 00:06.600 --> 00:08.250 Click on playground, you'll see 00:08.250 --> 00:09.630 that you can pick a project. 00:09.630 --> 00:11.430 I'm gonna go to the default projects 00:11.430 --> 00:13.800 and inside of here on the left, you have prompts, 00:13.800 --> 00:16.560 real-time assistance and text speech. 00:16.560 --> 00:19.020 We're gonna look at the prompt section in detail 00:19.020 --> 00:22.203 so I can specifically pick which type of model I want. 00:23.040 --> 00:25.923 I can give it certain tools, so for example, web search. 00:28.200 --> 00:31.020 And then I can say, you are going to help 00:31.020 --> 00:32.613 with customer support. 00:33.570 --> 00:37.230 And we'll then type on the bottom right 00:37.230 --> 00:38.550 with the chat to prompt. 00:38.550 --> 00:40.920 So I'm gonna say, can you research 00:40.920 --> 00:45.920 the best Mexican restaurant inside of London? 00:47.190 --> 00:49.320 And because we've attached one tool, 00:49.320 --> 00:52.740 which is the web search tool, you can see that the agent 00:52.740 --> 00:55.320 or chat bot is gonna basically go and assist 00:55.320 --> 00:57.090 and search the web and it's found 00:57.090 --> 00:58.500 a bunch of different results. 00:58.500 --> 01:01.590 And these are some things that might be useful. 01:01.590 --> 01:04.890 So you can choose the model, you can choose the tool, 01:04.890 --> 01:07.110 you can set a system message here, 01:07.110 --> 01:09.840 which is basically an instructional message 01:09.840 --> 01:12.060 that you should add to make sure that your chat bots 01:12.060 --> 01:14.850 or your LLM-based requests 01:14.850 --> 01:17.130 have a standardized way of applying. 01:17.130 --> 01:19.740 You can also, in the top right, click on the code 01:19.740 --> 01:21.630 and you can copy this code in lots 01:21.630 --> 01:24.900 of different programming languages such as Curl, Python, 01:24.900 --> 01:27.900 node.js, or as well, you can copy the JSON 01:27.900 --> 01:30.270 and see the output of the response. 01:30.270 --> 01:32.910 I'll also make a note that you can switch seamlessly 01:32.910 --> 01:35.400 between the chat completions API, 01:35.400 --> 01:38.340 and also, you can switch to the responses API. 01:38.340 --> 01:40.050 And you can do that in the top right 01:40.050 --> 01:43.230 and you can see that it keeps roughly everything the same. 01:43.230 --> 01:45.240 If you go to the copy on the code section, 01:45.240 --> 01:46.410 then we get the syntax 01:46.410 --> 01:49.200 for the chat.completions endpoint. 01:49.200 --> 01:51.690 As well as that, if we go back to the responses, 01:51.690 --> 01:55.140 you can have a look at the history of how you created this 01:55.140 --> 01:58.470 and what happened within that 30-day history. 01:58.470 --> 02:00.450 You can also compare, so you can have lots 02:00.450 --> 02:03.720 of different models and you can run these simultaneously. 02:03.720 --> 02:05.670 How are you today? 02:05.670 --> 02:08.790 If you hit command, enter, it will submit the prompt. 02:08.790 --> 02:10.650 And you can also add attachments here. 02:10.650 --> 02:12.870 So you can add a link to image, upload an image, 02:12.870 --> 02:14.370 or upload a file. 02:14.370 --> 02:17.100 So there's lots of things that are useful for this. 02:17.100 --> 02:21.660 I find it's a great place to demo a new use case 02:21.660 --> 02:23.460 or a new part of an application. 02:23.460 --> 02:25.230 You can play around with the prompts, 02:25.230 --> 02:26.700 you can easily change the model 02:26.700 --> 02:27.810 and you can see which type 02:27.810 --> 02:30.180 of model gives you the best response. 02:30.180 --> 02:33.000 As well as that, you can also click on this cog over here 02:33.000 --> 02:36.090 in the middle of the screen, which is a dial icon. 02:36.090 --> 02:39.030 And you can change the temperature, the maximum number 02:39.030 --> 02:42.960 of tokens that the app model can produce, and the top P. 02:42.960 --> 02:45.630 For the temperature, you can think of the temperature 02:45.630 --> 02:47.730 as a controlled randomness. 02:47.730 --> 02:52.080 So layering the results will have less random completions. 02:52.080 --> 02:54.900 If we set the temperature to very high, 02:54.900 --> 02:57.090 then it will produce random results. 02:57.090 --> 03:00.300 The temperature, you will get more deterministic outputs 03:00.300 --> 03:03.510 and standard outputs by having the temperature to zero. 03:03.510 --> 03:05.850 If you have the temperature at two, 03:05.850 --> 03:08.460 the model will be a lot more creative. 03:08.460 --> 03:10.770 You can use this when you are generating copy 03:10.770 --> 03:12.840 or social media text, play around 03:12.840 --> 03:14.580 with the temperature parameter. 03:14.580 --> 03:16.980 And this allows you to easily do that 03:16.980 --> 03:19.590 without doing it inside of code. 03:19.590 --> 03:22.980 You can also output different types of structured data. 03:22.980 --> 03:27.000 So we can output JSON objects or JSON schema, 03:27.000 --> 03:28.770 and you've got some examples here. 03:28.770 --> 03:31.650 Like for example, if we're doing moderation, 03:31.650 --> 03:33.780 we can attach an example of that 03:33.780 --> 03:37.293 and say, the cat sat on the mat, 03:38.250 --> 03:40.380 and you'll see that it actually provides us 03:40.380 --> 03:42.270 with structured data out 03:42.270 --> 03:45.930 via that text.format being the JSON schema. 03:45.930 --> 03:47.190 You can easily switch this back 03:47.190 --> 03:50.790 to text by going into the cog icon and clicking on text. 03:50.790 --> 03:52.380 There are other areas of the playground 03:52.380 --> 03:53.640 that you might want to experiment. 03:53.640 --> 03:56.640 So for example, if you are doing an OpenAI real-time, 03:56.640 --> 03:59.070 a voice to voice agent, you'll want 03:59.070 --> 04:00.720 to look at the real-time agent 04:00.720 --> 04:02.940 and that allows you to add on your microphone. 04:02.940 --> 04:05.400 And then you can start a chat with ChatGPT, 04:05.400 --> 04:07.950 and then ChatGPT will be able to then answer you 04:07.950 --> 04:09.540 and generate questions for you. 04:09.540 --> 04:10.920 So I can click generate 04:10.920 --> 04:14.700 and then the assistant will actually respond in real-time. 04:14.700 --> 04:16.770 So this is something that you can easily do 04:16.770 --> 04:18.600 and test out as well. 04:18.600 --> 04:20.940 I wouldn't recommend looking at the assistance API 04:20.940 --> 04:23.580 because that is going to be deprecated soon. 04:23.580 --> 04:25.890 And then text to speech is a way for you 04:25.890 --> 04:28.170 to basically put in some text 04:28.170 --> 04:31.770 and then OpenAI's backend will create real lifelike speech 04:31.770 --> 04:34.500 for you that you can then use in your applications. 04:34.500 --> 04:36.660 These are all the different things in the playground. 04:36.660 --> 04:38.370 I would suspect that you'll spend a lot 04:38.370 --> 04:40.170 of time in the prompts area 04:40.170 --> 04:42.480 because this is where you will do a lot of the work. 04:42.480 --> 04:44.460 And you can also save your prompts as well. 04:44.460 --> 04:47.040 So after you're done, you can click on the save 04:47.040 --> 04:49.620 and we could say, this is a customer support 04:49.620 --> 04:51.960 plus web search agent. 04:51.960 --> 04:54.180 And you can also allow anyone as well 04:54.180 --> 04:56.460 to have access to that specific prompt. 04:56.460 --> 04:59.520 And then you can share that with someone by. 04:59.520 --> 05:02.160 If you just have this URL, you can see this is the preset 05:02.160 --> 05:03.870 that they would go in the top here. 05:03.870 --> 05:06.120 So that's everything inside of the playground. 05:06.120 --> 05:07.170 In the next video, 05:07.170 --> 05:09.390 we're just gonna have a brief look, all the different areas 05:09.390 --> 05:12.090 of the dashboard and why you might decide to look at those. 05:12.090 --> 05:13.940 Cool, I'll see you in the next video.