1 00:00:03,590 --> 00:00:12,890 So these are very interesting and important chapter where we are going to explore how to a use a long 2 00:00:12,890 --> 00:00:17,030 chain application and an application with an API. 3 00:00:17,030 --> 00:00:22,400 So we can start communicating our application with the external world. 4 00:00:22,400 --> 00:00:22,970 Right. 5 00:00:22,970 --> 00:00:28,640 And to connect it and integrate it with other, other pieces of software. 6 00:00:28,640 --> 00:00:39,230 So what we are going to do in this exercise is to connect a basic drag application with a fast API. 7 00:00:39,680 --> 00:00:40,430 API. 8 00:00:40,580 --> 00:00:49,070 So fast API is a probably the most, uh, used, uh, framework in the Python ecosystem. 9 00:00:49,340 --> 00:00:56,210 Uh, you can you have different possibilities, but this one is one of the most popular in the launching 10 00:00:56,210 --> 00:00:56,780 world. 11 00:00:57,110 --> 00:01:03,530 So you will see that we will use the fast API module of long chain to create the API. 12 00:01:03,530 --> 00:01:08,150 And we will use Uvicorn to create the local server. 13 00:01:09,670 --> 00:01:21,190 The creation of the Rag application is a familiar to you, so we will start loading the dot env file 14 00:01:21,190 --> 00:01:26,980 the credentials from the file in order to be able to communicate with the OpenAI API. 15 00:01:26,980 --> 00:01:29,890 And then we will create the Rag application. 16 00:01:29,890 --> 00:01:32,680 We will create the LM instance. 17 00:01:32,680 --> 00:01:34,810 We will import the loader. 18 00:01:34,810 --> 00:01:40,570 We will load a text file that is in the data directory. 19 00:01:40,570 --> 00:01:46,600 This is an article an article about startups and stuff about startups. 20 00:01:46,600 --> 00:01:54,010 And we uh the next thing we do is we split this document into small chunks. 21 00:01:54,010 --> 00:02:03,580 Then we, uh, transform the chunks into embeddings, and we load the content into a file vector database. 22 00:02:03,790 --> 00:02:09,250 Then we create a retrieval QA chain. 23 00:02:09,250 --> 00:02:15,550 And with this chain we can start making questions to our private document. 24 00:02:16,510 --> 00:02:24,400 So how do we connect these basic rack application with fast API. 25 00:02:25,090 --> 00:02:27,370 So the first thing we need is to install. 26 00:02:27,370 --> 00:02:32,260 If you don't have it to install the fast API package you know how to do it. 27 00:02:32,260 --> 00:02:38,590 If you are using terminal, you just write pip install fast API and if you want to do it, you want 28 00:02:38,590 --> 00:02:41,320 to do it via a Jupyter notebook. 29 00:02:41,320 --> 00:02:46,330 You can write the exclamation mark before and then you execute the cell. 30 00:02:46,330 --> 00:02:47,290 And that's it. 31 00:02:47,290 --> 00:02:50,680 You have here the pound sign because we want to uncomment this. 32 00:02:50,710 --> 00:02:54,400 We don't want to do this operation because we have done it already. 33 00:02:54,670 --> 00:03:01,150 So then you import the fast API and the Http exception modules. 34 00:03:01,150 --> 00:03:02,680 You create the application. 35 00:03:02,680 --> 00:03:05,590 And in this case we are creating an endpoint. 36 00:03:05,590 --> 00:03:07,060 It's a post endpoint. 37 00:03:07,060 --> 00:03:15,250 And you see that this endpoint is using the chain we just created in order to get the uh data. 38 00:03:15,250 --> 00:03:22,810 So once you have that you install you can install Uvicorn to proceed with the local server. 39 00:03:22,810 --> 00:03:28,600 And once you have that, this is the way you configure it using Jupyter Notebook. 40 00:03:28,600 --> 00:03:34,600 If you are using, for example, a Visual Studio Code editor instead of a Jupyter notebook, you can 41 00:03:34,600 --> 00:03:36,220 use this other approach. 42 00:03:36,220 --> 00:03:38,350 You have any problem with this process? 43 00:03:38,350 --> 00:03:46,510 With the server, you just go to the usual places ChatGPT and Google or StackOverflow and ask about 44 00:03:46,510 --> 00:03:51,040 any doubt or problem you may have in the process, and you will find, uh, the response there. 45 00:03:51,040 --> 00:03:58,360 But if everything is successful, and I think it is going to be because this is quite simple to do, 46 00:03:58,480 --> 00:04:07,540 uh, then you will have a situation like this where you can start the local server in this, uh, address 47 00:04:07,540 --> 00:04:08,650 and here. 48 00:04:10,510 --> 00:04:12,280 Vuiyasawa said this. 49 00:04:14,740 --> 00:04:23,320 You will see something like this and you can click here on post and try it out and you can enter here. 50 00:04:23,320 --> 00:04:24,490 Your question. 51 00:04:24,640 --> 00:04:31,180 Uh, it would be like what is this article? 52 00:04:33,140 --> 00:04:36,650 About in less than. 53 00:04:38,900 --> 00:04:40,670 100 words. 54 00:04:42,830 --> 00:04:43,820 Execute. 55 00:04:46,920 --> 00:04:48,540 And here you have the response. 56 00:04:49,300 --> 00:04:58,240 Okay, so the most important thing of this chapter is to understand that our LM application can end 57 00:04:58,240 --> 00:04:59,890 up in an API. 58 00:04:59,890 --> 00:05:07,300 So we can start plugging our LM applications with other external applications to do. 59 00:05:08,100 --> 00:05:10,140 A whole lot of different things. 60 00:05:10,440 --> 00:05:16,380 Remember that you can go to the documentation to know more, or to experiment further with this approach, 61 00:05:16,380 --> 00:05:26,760 and to find the special use case that is, uh, best for you, the particular API, API, or approach 62 00:05:26,760 --> 00:05:29,340 that works better for your use case.