1 00:00:03,450 --> 00:00:03,990 Okay. 2 00:00:03,990 --> 00:00:04,710 So. 3 00:00:06,220 --> 00:00:12,910 In this chapter we are going to talk about a basic long chain operations. 4 00:00:13,090 --> 00:00:23,320 So the goal of this chapter is not to go into details, is just show you how easy it is to work with 5 00:00:23,320 --> 00:00:24,400 a long chain. 6 00:00:24,880 --> 00:00:27,010 So a. 7 00:00:28,650 --> 00:00:33,720 From now on you are going to see the screen in these two sides. 8 00:00:33,810 --> 00:00:42,270 In this side we are going to follow the presentation in order to to have, you know, the main concepts 9 00:00:42,270 --> 00:00:49,350 we are talking about and in this other side is where we are going to have the code of our application. 10 00:00:49,350 --> 00:00:52,290 As you can see, this is a Jupyter notebook. 11 00:00:52,290 --> 00:00:55,350 And in the Jupyter notebook you see a couple of things. 12 00:00:55,350 --> 00:00:58,830 You have markdown and you have code. 13 00:00:58,830 --> 00:01:01,500 So a Jupyter notebook. 14 00:01:02,040 --> 00:01:03,240 Let's move. 15 00:01:03,240 --> 00:01:08,190 This Jupyter notebook is something you. 16 00:01:09,240 --> 00:01:13,920 And you create, you will see that we will have the terminal open. 17 00:01:13,920 --> 00:01:18,420 As you see, you can see that I have a virtual environment here. 18 00:01:18,420 --> 00:01:25,860 And we are also having a ChatGPT here as an assistant for us. 19 00:01:25,860 --> 00:01:31,680 So for example you can come here, you can use GPT 3.5. 20 00:01:31,680 --> 00:01:39,930 This is the free version or GPT four which is the plus uh, subscription of OpenAI. 21 00:01:39,930 --> 00:01:49,290 And this allows you for a very, very, uh, convenient price, uh, to use different, uh, functionality. 22 00:01:49,290 --> 00:01:56,010 The default is extremely good because it knows about coding is really amazing. 23 00:01:56,010 --> 00:01:59,520 It's like GPT 3.5 on steroids. 24 00:01:59,520 --> 00:02:01,980 But you also have, you know, different possibilities. 25 00:02:01,980 --> 00:02:04,230 Like to browse with Bing. 26 00:02:04,230 --> 00:02:11,310 And this is where you can find answers to, uh, current matters, you know, things that happened yesterday 27 00:02:11,310 --> 00:02:11,970 or whatever. 28 00:02:11,970 --> 00:02:12,480 Right. 29 00:02:12,480 --> 00:02:19,410 So for example, if you enter here how to install Pyenv. 30 00:02:21,840 --> 00:02:25,110 You are going to see that immediately. 31 00:02:25,110 --> 00:02:31,440 GPT four is going to, uh, give you the detailed instructions for doing so. 32 00:02:31,440 --> 00:02:38,070 So super easy companion to have a in the terminal. 33 00:02:38,070 --> 00:02:46,380 You know that I have several windows of the terminal open in one of the, uh, one of the windows. 34 00:02:46,380 --> 00:02:51,000 I, I entered Jupyter notebook. 35 00:02:51,690 --> 00:02:57,240 If you click here, you are going to see a screen like this. 36 00:02:58,450 --> 00:03:02,920 So here we have a lot of Jupyter notebooks already created. 37 00:03:02,920 --> 00:03:06,070 You can create a new Jupyter notebook doing this. 38 00:03:06,070 --> 00:03:07,360 Notebook new. 39 00:03:08,400 --> 00:03:09,900 And you will see something like this. 40 00:03:09,900 --> 00:03:12,000 We want this with Python three. 41 00:03:12,030 --> 00:03:17,550 You will see that the Jupyter notebook is a is formed by cells. 42 00:03:17,550 --> 00:03:18,600 This is a cell. 43 00:03:18,600 --> 00:03:22,110 In the cell you can have, you know, code. 44 00:03:27,600 --> 00:03:30,240 In order to execute this code. 45 00:03:30,240 --> 00:03:38,250 Once you are here, you, uh, press shift and enter, and then you execute the code. 46 00:03:38,250 --> 00:03:44,790 And here, if you want to have markdown instead of code, you can have a. 47 00:03:48,620 --> 00:03:49,430 Things like that. 48 00:03:49,460 --> 00:03:49,760 Right. 49 00:03:49,760 --> 00:03:53,510 So this is not going to be executed. 50 00:03:53,540 --> 00:03:56,210 This is just a markdown like like text. 51 00:03:56,210 --> 00:04:03,200 So if you go to ChatGPT and ask about Jupyter Notebook markdown, you will have a complete guide or 52 00:04:03,200 --> 00:04:05,150 super easy to to use. 53 00:04:05,450 --> 00:04:12,500 And this is very, very useful to have markdown with your code because it's going to allow you to understand 54 00:04:12,500 --> 00:04:17,060 the code you're writing, and especially if you are learning or teaching, it's a super useful tool. 55 00:04:17,180 --> 00:04:23,660 And a in the code you can come here and for example comment out this. 56 00:04:23,720 --> 00:04:26,330 Now you do shift enter. 57 00:04:26,330 --> 00:04:29,090 This is not executed because it's commented out. 58 00:04:29,090 --> 00:04:35,120 If you come back and you remove the pound sign and shift enter again, you will execute. 59 00:04:35,870 --> 00:04:37,250 You come here and click in. 60 00:04:37,250 --> 00:04:39,440 Plus you create a new cell. 61 00:04:39,440 --> 00:04:40,340 You click here. 62 00:04:40,340 --> 00:04:41,030 You are removed. 63 00:04:41,030 --> 00:04:42,170 This is super easy. 64 00:04:42,170 --> 00:04:45,440 This is super easy to use Jupyter notebook. 65 00:04:45,440 --> 00:04:55,070 So in this, uh, notebook we have five operations in in long chain to show you how easy it is. 66 00:04:55,490 --> 00:05:02,300 The first thing you see here is how we are invoking the dot env file. 67 00:05:02,300 --> 00:05:09,710 So if you remember in the previous chapter we talk about this hidden file where we have our passwords, 68 00:05:09,710 --> 00:05:11,510 you know and private information. 69 00:05:11,510 --> 00:05:15,860 So this dot m file is something like this. 70 00:05:15,860 --> 00:05:20,900 We have a file that is called dot env. 71 00:05:21,620 --> 00:05:28,730 And we have inside of this file which we can create this file with whatever editor like the typical 72 00:05:28,730 --> 00:05:30,920 notes editor you have in your computer. 73 00:05:30,920 --> 00:05:39,350 Important thing is that this dot m file is in the same directory as the directory you have here. 74 00:05:39,350 --> 00:05:42,800 So here in this folder I have the dot m. 75 00:05:42,800 --> 00:05:47,300 But since it's a hidden file uh, you you cannot see it here. 76 00:05:47,300 --> 00:05:58,070 So in this file I have a variable called OpenAI API key in capital letters equal to, you know, and 77 00:05:58,070 --> 00:06:03,320 here I copy and paste my, you know, OpenAI key. 78 00:06:03,530 --> 00:06:06,290 I close this file, okay. 79 00:06:06,290 --> 00:06:11,870 And uh, and I invoke it from the Jupyter notebook. 80 00:06:11,870 --> 00:06:16,580 So what I'm doing here is import a module called OS. 81 00:06:16,730 --> 00:06:23,960 And from dot m the module dot m import these two uh components. 82 00:06:23,960 --> 00:06:31,970 So this thing the module dot m, uh, if you remember in a previous chapter, the way we import this 83 00:06:31,970 --> 00:06:36,830 module is actually, uh, in importing Python. 84 00:06:37,690 --> 00:06:41,560 A m instead of a m by itself. 85 00:06:41,560 --> 00:06:47,050 So is pip python a dot m okay. 86 00:06:47,560 --> 00:06:53,110 So once you have that imported, if you don't have it, you will see an error message. 87 00:06:53,110 --> 00:06:55,420 Then you go to terminal and you import. 88 00:06:55,420 --> 00:07:02,410 It is like pip install python this dot m. 89 00:07:05,250 --> 00:07:05,940 Okay. 90 00:07:07,080 --> 00:07:08,100 So. 91 00:07:09,180 --> 00:07:13,380 I say pit instead of pip is pip okay? 92 00:07:13,380 --> 00:07:15,600 Then you execute it and you will install. 93 00:07:15,600 --> 00:07:17,820 I have it installed so I'm not going to do it again. 94 00:07:17,850 --> 00:07:30,000 Okay, so once we have that now this Jupyter notebook has my OpenAI API key save in this variable. 95 00:07:30,000 --> 00:07:30,630 Okay. 96 00:07:30,630 --> 00:07:42,360 This is the name that OpenAI require us to have a in order to, uh, use the OpenAI API key. 97 00:07:42,360 --> 00:07:45,420 So first operation we are going to do. 98 00:07:47,020 --> 00:07:51,700 How to connect with an LLM. 99 00:07:51,700 --> 00:07:56,590 In this case, we are going to connect with OpenAI with ChatGPT. 100 00:07:56,620 --> 00:07:57,250 Okay. 101 00:07:57,250 --> 00:08:02,860 We are going to ask a question to ChatGPT from our application. 102 00:08:02,860 --> 00:08:09,190 So the first thing we do is we call the module OpenAI inside of lang chain. 103 00:08:09,190 --> 00:08:13,930 And with this module we are going to create a new instance of an LLM. 104 00:08:13,930 --> 00:08:17,770 In this case we are going to call it my creative LLM. 105 00:08:18,430 --> 00:08:24,520 And as you will see, we are going to include a temperature of 0.9. 106 00:08:24,520 --> 00:08:25,150 Why? 107 00:08:25,150 --> 00:08:31,870 Because uh, we want this instance of LLM to be creative, very creative. 108 00:08:31,870 --> 00:08:40,179 The creativity or the temperature of the OpenAI ChatGPT is from 0 to 1. 109 00:08:40,179 --> 00:08:46,240 So 0.9 is a very high creativity and zero is a very low. 110 00:08:46,240 --> 00:08:53,890 We are going to use low creativity when we want to have very accurate information, and we are going 111 00:08:53,890 --> 00:08:56,890 to use high creativity when we want. 112 00:08:57,190 --> 00:09:05,710 We don't matter about it if if it is a fake content or a hallucination or whatever, we don't care as 113 00:09:05,710 --> 00:09:08,710 long as it is interesting or beautiful for us. 114 00:09:08,710 --> 00:09:09,010 Right? 115 00:09:09,010 --> 00:09:15,220 So we are creating a brand name or we are creating a story for our children or whatever. 116 00:09:15,220 --> 00:09:21,460 We want to have a high temperature, high creativity, or if we want to have factual data, we want 117 00:09:21,460 --> 00:09:24,760 to have low creativity, low temperature. 118 00:09:24,760 --> 00:09:30,940 So this is the question we want to, uh, to send to the. 119 00:09:32,050 --> 00:09:33,520 Chat gpt lm. 120 00:09:33,790 --> 00:09:38,110 So we save this question in a variable. 121 00:09:38,410 --> 00:09:41,770 And then the only thing we do is we. 122 00:09:43,160 --> 00:09:50,480 Send this question to the instance of the LM we have created and print the response. 123 00:09:50,990 --> 00:09:57,830 So this is the response that ChatGPT sent to our application. 124 00:09:59,000 --> 00:10:03,170 As you can see here, all this code is written for you. 125 00:10:03,200 --> 00:10:08,510 I don't want to waste time writing, writing and executing these cells. 126 00:10:08,540 --> 00:10:13,910 This is a pre-written, uh, for you so we can go a little bit faster. 127 00:10:13,910 --> 00:10:21,050 So this is the response that the GPT, OpenAI LM has sent back to us. 128 00:10:21,050 --> 00:10:25,850 So to the question, what are the five best things to do in life? 129 00:10:25,850 --> 00:10:29,840 ChatGPT says, live life of purpose and meaning, blah, blah, blah. 130 00:10:29,990 --> 00:10:30,620 Excellent. 131 00:10:30,620 --> 00:10:31,130 Okay. 132 00:10:31,860 --> 00:10:37,530 So you see how easy it is to connect our application with ChatGPT. 133 00:10:37,530 --> 00:10:40,770 And we can do that with any other foundational. 134 00:10:40,860 --> 00:10:49,020 So we will do the same when we are connecting with llama too, or anthropic or whatever Elm foundation 135 00:10:49,020 --> 00:10:50,310 elm we want to use. 136 00:10:50,460 --> 00:10:57,060 Second thing, I want to show you how to wrap the user prompt with a prompt template. 137 00:10:57,090 --> 00:11:03,060 So we are going to do is we are going to create the prompt template with a variable to include the user's 138 00:11:03,060 --> 00:11:03,750 prompt. 139 00:11:04,590 --> 00:11:07,290 Second, we are going to get the user's prompt. 140 00:11:07,290 --> 00:11:14,490 Then combine the user's prompt with the prompt template, pass the combined prompt to the LM and finally 141 00:11:14,490 --> 00:11:15,240 get the answer. 142 00:11:15,240 --> 00:11:20,850 So what we have done here is we have this question. 143 00:11:20,850 --> 00:11:22,050 Is the prompt okay. 144 00:11:22,050 --> 00:11:25,110 The prompt is what we tell the LM. 145 00:11:25,470 --> 00:11:32,310 So in the second example we are going to do what we are most frequently going to do in our application, 146 00:11:32,310 --> 00:11:35,250 which is in the application. 147 00:11:35,250 --> 00:11:42,780 The LM engineer is going to create a prompt template. 148 00:11:42,780 --> 00:11:52,470 The prompt template is not the full prompt, it's just a part of the prompt that has some parts that 149 00:11:52,470 --> 00:11:55,590 are going to be filled by the user. 150 00:11:55,590 --> 00:12:04,320 So as you can see here, we are using the prompt template module to create my prompt template with this 151 00:12:04,320 --> 00:12:05,190 content. 152 00:12:05,190 --> 00:12:08,430 And in this content we are going to include a variable. 153 00:12:08,430 --> 00:12:14,430 So the template is going to be what are the three best things to do in a destination. 154 00:12:14,430 --> 00:12:18,030 But the destination is not provided here. 155 00:12:18,030 --> 00:12:19,560 It's just a variable. 156 00:12:19,560 --> 00:12:22,770 And this variable will be provided by the user. 157 00:12:22,770 --> 00:12:28,800 So this user can provide this variable from a form or a mobile or an email or whatever. 158 00:12:28,800 --> 00:12:31,320 But in this case we are going to enter it manually. 159 00:12:31,320 --> 00:12:34,950 We say this input variable is Barcelona. 160 00:12:35,460 --> 00:12:43,920 So the prompt we are going to send the LM is going to be what are the three best things to do in Barcelona 161 00:12:43,920 --> 00:12:44,400 okay. 162 00:12:44,400 --> 00:12:45,510 Or Barcelona. 163 00:12:45,900 --> 00:12:49,710 So what we are doing here is we are telling okay. 164 00:12:51,250 --> 00:12:54,010 Enter this prompt template. 165 00:12:54,720 --> 00:12:57,060 With this variable. 166 00:12:57,900 --> 00:13:02,190 And send it to the creative LM. 167 00:13:02,860 --> 00:13:08,080 You have created and print the response. 168 00:13:08,080 --> 00:13:13,330 So the LM ChatGPT in this case is going to send back this response. 169 00:13:13,330 --> 00:13:16,330 So what are the three best things to do in Barcelona. 170 00:13:16,360 --> 00:13:18,520 First, explore the Sagrada Familia. 171 00:13:18,550 --> 00:13:20,530 Second, wondering La Ramblas. 172 00:13:20,560 --> 00:13:22,780 Third, take a day trip to the park. 173 00:13:23,020 --> 00:13:23,590 Blah blah. 174 00:13:24,040 --> 00:13:24,640 Excellent. 175 00:13:24,640 --> 00:13:25,600 Good answer. 176 00:13:26,170 --> 00:13:31,600 Third thing, I want to show you how to combine instructions in chains. 177 00:13:31,780 --> 00:13:38,710 So long chain as its name because of it is very good doing chains. 178 00:13:38,710 --> 00:13:46,630 We will see that chains are just a collection or a sequence of instructions we group together. 179 00:13:47,140 --> 00:13:53,350 So what we are going to do here is we are going to do the first two things we previously did. 180 00:13:53,350 --> 00:13:57,070 So initialize an instance of the foundation lm. 181 00:13:57,070 --> 00:14:01,060 Create the prompt template with a variable to include the user's prompt. 182 00:14:01,090 --> 00:14:02,560 Get the user's prompt. 183 00:14:02,560 --> 00:14:04,720 All this we have already made. 184 00:14:04,720 --> 00:14:12,040 Now we are going to add the fourth step, which is combine the user's prompt with the prompt template. 185 00:14:13,700 --> 00:14:14,510 We will do. 186 00:14:14,510 --> 00:14:20,690 Uh, we will create a chain combining the LM and the Prom template, and we will run the chain. 187 00:14:20,690 --> 00:14:23,390 So this we already did it. 188 00:14:23,390 --> 00:14:28,790 We create we have an instance of the LM, we create a Prom template. 189 00:14:28,970 --> 00:14:35,510 And, uh, in this case, the template is going to be what are the three most remarkable achievements 190 00:14:35,510 --> 00:14:37,340 of a politician? 191 00:14:37,340 --> 00:14:42,170 We don't know which politician is going to be the variable that the user is going to include. 192 00:14:42,530 --> 00:14:49,700 And here what we are saying is create a chain instead of a printing. 193 00:14:49,700 --> 00:14:54,980 Here, all the things we were printing in the previous exercise. 194 00:14:54,980 --> 00:14:57,680 Here we are printing the chain. 195 00:14:57,680 --> 00:15:01,520 We are printing the result of running the chain with this variable. 196 00:15:01,520 --> 00:15:05,660 So the politician that the user is giving us is Winston Churchill. 197 00:15:07,180 --> 00:15:14,530 And here we have created the chain with the component LM chain we have imported. 198 00:15:14,530 --> 00:15:22,840 So here what we are saying is okay, use this component to create a chain that is going to use the LM 199 00:15:22,840 --> 00:15:26,590 you have created and the prompt template you have created. 200 00:15:26,830 --> 00:15:35,320 And here send this chain to the foundation LM to ChatGPT in this case and print the response. 201 00:15:35,320 --> 00:15:42,790 So in this case the prompt we are sending is what are the three most remarkable achievements of Churchill 202 00:15:43,360 --> 00:15:48,160 Downs is going to be first one leading Britain through the World War two. 203 00:15:48,160 --> 00:15:53,140 Second, establishing the special relationship between the United Kingdom and the United States. 204 00:15:53,140 --> 00:15:56,860 And third, winning the Nobel Prize in Literature. 205 00:15:56,890 --> 00:15:59,110 Okay, very good answer. 206 00:16:02,470 --> 00:16:09,100 The next thing I want to review, just to see how easy it is to do things with long chain, is to create 207 00:16:09,100 --> 00:16:10,090 an agent. 208 00:16:10,240 --> 00:16:15,790 So agents are one of the most interesting things with LM applications. 209 00:16:15,790 --> 00:16:19,810 In this case you will see how fast we can create an agent. 210 00:16:19,990 --> 00:16:27,280 But we are going to do is to load external tools, initialize an agent over a foundation LM in this 211 00:16:27,280 --> 00:16:35,530 case OpenAI, and set a goal for the agent and let it decide what tools to use, uh, to achieve it. 212 00:16:35,530 --> 00:16:35,830 Okay. 213 00:16:35,830 --> 00:16:37,330 This is the purpose of the agent. 214 00:16:37,330 --> 00:16:40,450 The agent is going to reason by itself. 215 00:16:40,450 --> 00:16:44,680 So we are going to to in this case, we are going to create an agent. 216 00:16:44,680 --> 00:16:50,650 And we are going to tell him, okay, here you have a tool, uh, to search on Google. 217 00:16:50,650 --> 00:16:53,620 And here you have a math calculator. 218 00:16:54,580 --> 00:16:56,650 Use whatever you need. 219 00:16:56,980 --> 00:16:58,210 You decide. 220 00:16:58,210 --> 00:17:05,109 So if we ask a mathematical question like, okay, you know, calculate the square root of whatever 221 00:17:05,380 --> 00:17:10,150 this agent is going to say, okay, I'm going to use the math calculator for this. 222 00:17:10,420 --> 00:17:17,680 And if we make a question which is I don't know how old are the pyramids in Egypt, the agent is going 223 00:17:17,680 --> 00:17:22,990 to say, okay, for this question, I'm going to use the Google search in order to find the results. 224 00:17:22,990 --> 00:17:23,619 Okay. 225 00:17:23,619 --> 00:17:24,579 So. 226 00:17:25,510 --> 00:17:31,330 We achieve this functionality with these six lines. 227 00:17:31,900 --> 00:17:40,390 So what we are doing here and this, all these six lines are commented out in order to comment it out. 228 00:17:40,390 --> 00:17:47,260 All the code in a in a cell in Mac you use command plus design. 229 00:17:47,290 --> 00:17:53,320 Okay, so if you use this, all the code you have in in the cell is commented out. 230 00:17:53,470 --> 00:17:55,510 What we are saying in this code why? 231 00:17:55,510 --> 00:17:57,670 Why have we commented out this code? 232 00:17:57,670 --> 00:18:06,310 Because as I say here, operations with agents are more expensive in tokens. 233 00:18:06,400 --> 00:18:19,390 So if you know the previous operations have a cost of $0.0003 for you, this one can cost you ten times, 234 00:18:19,390 --> 00:18:20,320 uh, more. 235 00:18:20,320 --> 00:18:23,020 So it's still very low. 236 00:18:23,020 --> 00:18:25,570 But I just wanted to tell you this is more expensive. 237 00:18:25,780 --> 00:18:27,850 How can you control. 238 00:18:27,850 --> 00:18:29,710 How much are you spending? 239 00:18:29,710 --> 00:18:40,210 So in your OpenAI account, you have this usage, uh, tab where you can see how much have you spent 240 00:18:40,210 --> 00:18:40,960 per day. 241 00:18:41,140 --> 00:18:47,140 And you can do a lot of operations and still spend, you know, less than $1 per day. 242 00:18:47,140 --> 00:18:53,890 So it's good for you, you know, to control on a daily basis at the at the beginning, what is your 243 00:18:53,890 --> 00:18:56,680 cost and your expenses in ChatGPT? 244 00:18:56,710 --> 00:18:57,310 Okay. 245 00:18:57,850 --> 00:19:08,050 This is is not exactly real time, but is it can have a delay of one minute in order to to update the 246 00:19:08,050 --> 00:19:10,330 cost that you are, uh, you are having. 247 00:19:10,330 --> 00:19:13,810 So in this case, what we are saying is load the tools. 248 00:19:13,960 --> 00:19:20,800 Initially this is a module two load module to load the tools, a module to initialize the agent, and 249 00:19:20,800 --> 00:19:22,870 a module to initialize the LM. 250 00:19:22,870 --> 00:19:26,140 And these three lines of code make it all. 251 00:19:26,140 --> 00:19:29,380 The first thing we do is load two tools. 252 00:19:29,380 --> 00:19:35,560 Three tools, actually the Google search, the calculator and the LM. 253 00:19:36,370 --> 00:19:42,610 Then we are initializing an agent with the previous tools and the LM. 254 00:19:42,610 --> 00:19:48,610 And the type of agent we are going to use is this type you have here you have several types of agents. 255 00:19:48,610 --> 00:19:49,510 We will see it. 256 00:19:49,510 --> 00:19:57,670 And then with this line of code agent dot run, we are executing the agent with this particular question. 257 00:19:57,670 --> 00:20:01,900 So here as you can see in this prompt we have two questions. 258 00:20:01,900 --> 00:20:07,690 Who is the current leader of Japan and what is the largest prime number blah blah blah. 259 00:20:07,870 --> 00:20:12,760 So for the first question the agent is going to use the Google search. 260 00:20:12,760 --> 00:20:16,450 And for the second question it is going to use the math calculator. 261 00:20:16,450 --> 00:20:19,870 And this decision is made by the agent itself. 262 00:20:20,230 --> 00:20:22,810 That's the powerful thing of agents. 263 00:20:23,740 --> 00:20:31,990 And the last thing I want to cover is the memory how to add memory to the LM. 264 00:20:31,990 --> 00:20:40,630 So in this case we are going to use a chain, a predefined chain called conversation chain to record 265 00:20:40,630 --> 00:20:44,650 the conversation in the memory of the LM instance. 266 00:20:44,800 --> 00:20:53,920 So with just these lines we well we are importing a couple of modules OpenAI and conversation chain. 267 00:20:53,920 --> 00:20:57,880 With the first one we are creating a LM instance. 268 00:20:57,880 --> 00:21:05,650 We call it my no creative LM because the temperature, the creativity level is very low and we are going 269 00:21:05,650 --> 00:21:07,120 to create a chain. 270 00:21:08,160 --> 00:21:10,380 With that LM. 271 00:21:11,430 --> 00:21:17,100 This chain is created with the component conversation chain. 272 00:21:17,100 --> 00:21:24,870 And because of that it is going to have a very interesting functionality already included. 273 00:21:24,870 --> 00:21:31,500 So this is one of the predefined chains that long chain has created for us. 274 00:21:31,590 --> 00:21:34,560 And this is going to save us a lot of work. 275 00:21:34,560 --> 00:21:35,070 Okay. 276 00:21:35,070 --> 00:21:41,700 So with this chain that we call my conversation we can start a conversation with the LM. 277 00:21:41,970 --> 00:21:43,440 So hi there. 278 00:21:43,560 --> 00:21:52,560 The LM is going to answer first showing us the prompt, the default prompt that the conversation chain 279 00:21:52,560 --> 00:21:54,240 has created. 280 00:21:54,240 --> 00:21:58,110 And then it's going to tell us okay, this is what you are saying. 281 00:21:58,110 --> 00:21:58,650 Hi there. 282 00:21:58,650 --> 00:22:00,300 And this is what I respond. 283 00:22:00,300 --> 00:22:01,110 Hi there. 284 00:22:01,110 --> 00:22:02,340 How can I help you? 285 00:22:02,580 --> 00:22:04,980 Next thing we tell we tell the LM. 286 00:22:05,340 --> 00:22:06,330 My name is Julio. 287 00:22:06,360 --> 00:22:09,540 What is your name response my name is Ivan. 288 00:22:09,540 --> 00:22:10,800 It's nice to meet you, Julio. 289 00:22:10,830 --> 00:22:11,400 Third thing. 290 00:22:11,430 --> 00:22:13,500 Do you remember my name? 291 00:22:14,910 --> 00:22:16,770 Yes, I remember your name is Julio. 292 00:22:16,770 --> 00:22:24,750 This is just a very basic example to show you how easy it is to create an LM application with memory. 293 00:22:24,750 --> 00:22:29,760 And the last thing you have here is we are printing my conversations. 294 00:22:29,760 --> 00:22:31,170 We are printing these. 295 00:22:31,860 --> 00:22:33,330 Chain you have created. 296 00:22:33,330 --> 00:22:36,810 We are printing just to see what is inside there. 297 00:22:36,810 --> 00:22:40,110 You see that inside here, this variable, my conversation. 298 00:22:40,110 --> 00:22:41,670 There are a lot of things. 299 00:22:42,120 --> 00:22:43,470 There are memory. 300 00:22:43,470 --> 00:22:51,390 There are things like human message, I message, etc. etc. we will see what is this about in the next 301 00:22:51,390 --> 00:22:51,840 chapter. 302 00:22:51,840 --> 00:23:00,210 So the purpose of this initial chapter regarding long chain is just to show you how easy it is to work 303 00:23:00,210 --> 00:23:03,450 with this framework and how powerful it is. 304 00:23:04,080 --> 00:23:11,310 What we have covered here are super simple things we haven't gone through details you know, into details, 305 00:23:11,310 --> 00:23:18,900 but you see how easy it is, uh, to communicate with, uh, the, the long chain code, to use the 306 00:23:18,900 --> 00:23:23,640 long chain code and to integrate, you know, different tools and different functionality. 307 00:23:23,640 --> 00:23:26,880 We will go into details in the next chapters.