1 00:00:00,000 --> 00:00:01,000 Hello guys! 2 00:00:01,000 --> 00:00:07,000 In this video we are going to create an amazing end to end multi AI agent application using land graph 3 00:00:07,000 --> 00:00:08,000 and astrolabe. 4 00:00:08,000 --> 00:00:13,000 Now if you don't know about land graph, Landgraaf actually helps you to create AI agents in such a 5 00:00:13,000 --> 00:00:15,000 way that you'll be able to manage their state. 6 00:00:15,000 --> 00:00:17,000 The communication between the agents. 7 00:00:17,000 --> 00:00:22,000 It provides you the entire workflow to probably create this amazing AI applications. 8 00:00:22,000 --> 00:00:27,000 So in this video also, we are going to probably discuss about an end to end project. 9 00:00:27,000 --> 00:00:31,000 First of all, we will go ahead and discuss about the architecture, what we are actually trying to 10 00:00:31,000 --> 00:00:32,000 implement. 11 00:00:32,000 --> 00:00:36,000 And then step by step, we'll go ahead and implement this with the help of land graph. 12 00:00:36,000 --> 00:00:40,000 So here the architecture of the project looks like this. 13 00:00:40,000 --> 00:00:46,000 Let's say that I'm trying to create a chatbot application which has this multi AI agent feature which 14 00:00:46,000 --> 00:00:49,000 will be able to communicate with each other and it will be able to do multiple task. 15 00:00:50,000 --> 00:00:54,000 So whenever a user query comes, there will be a start node in land graph. 16 00:00:54,000 --> 00:00:57,000 And this start node will further send it to router. 17 00:00:57,000 --> 00:01:01,000 Router basically means whenever I'm probably putting up a user query. 18 00:01:01,000 --> 00:01:08,000 We need to route it whether this user query can be fulfilled after requesting from a vector database, 19 00:01:08,000 --> 00:01:13,000 or whether it can be fulfilled from an external tool like wiki search. 20 00:01:13,000 --> 00:01:16,000 Now here, just for an example, I'm considering wiki search. 21 00:01:16,000 --> 00:01:22,000 But uh, in real world application you can even implement different different external tools you like, 22 00:01:22,000 --> 00:01:23,000 right? 23 00:01:23,000 --> 00:01:24,000 So wiki search is one. 24 00:01:24,000 --> 00:01:29,000 You can also use serp API, you know, to probably do Google search itself. 25 00:01:29,000 --> 00:01:29,000 Right. 26 00:01:29,000 --> 00:01:33,000 So in this example I'm just going to consider one. 27 00:01:33,000 --> 00:01:38,000 Uh, and similarly you can implement with respect to other third party external tools. 28 00:01:38,000 --> 00:01:44,000 Then uh, if I probably consider this side, you will be able to see that it will do a vector DB search. 29 00:01:44,000 --> 00:01:47,000 Now this vector db is basically getting created by Astra db. 30 00:01:47,000 --> 00:01:51,000 We will try to create this entire vector database in the cloud. 31 00:01:51,000 --> 00:01:51,000 Okay. 32 00:01:52,000 --> 00:01:53,000 In the cloud. 33 00:01:53,000 --> 00:01:58,000 And initially inside this vector database we will be putting some information. 34 00:01:58,000 --> 00:02:04,000 The information will be a website page from which we can specifically query right. 35 00:02:04,000 --> 00:02:08,000 We will take some URLs and we will read the entire data inside that particular website. 36 00:02:08,000 --> 00:02:13,000 And we will store that all in in the form of vectors in the vector database. 37 00:02:13,000 --> 00:02:14,000 That is Astra DB. 38 00:02:14,000 --> 00:02:15,000 Right. 39 00:02:15,000 --> 00:02:17,000 And this will be in the cloud platform. 40 00:02:17,000 --> 00:02:19,000 Um, basically it is deployed in the cloud. 41 00:02:19,000 --> 00:02:25,000 And I will try to show you the entire configuration and how powerful and amazing Astra DB is. 42 00:02:25,000 --> 00:02:30,000 Okay, after we do this specific thing, then the next step will be that we will get the response from 43 00:02:30,000 --> 00:02:38,000 this, and then we can integrate along with LLM along with some prompt engineering like what we really 44 00:02:38,000 --> 00:02:40,000 want to do with this particular response. 45 00:02:40,000 --> 00:02:42,000 And finally we should be able to get our response. 46 00:02:42,000 --> 00:02:44,000 And the final node is basically n. 47 00:02:44,000 --> 00:02:48,000 So this is the entire architecture of this project. 48 00:02:48,000 --> 00:02:49,000 In terms of line graph. 49 00:02:49,000 --> 00:02:53,000 We are going to be we are we are going to use the line graph where we are going to implement this entire 50 00:02:53,000 --> 00:02:54,000 thing. 51 00:02:54,000 --> 00:02:54,000 Right. 52 00:02:55,000 --> 00:02:59,000 So without wasting any time, let's go ahead and implement this entire thing. 53 00:02:59,000 --> 00:03:00,000 Okay. 54 00:03:00,000 --> 00:03:04,000 So first of all you really need to know about data stacks. 55 00:03:04,000 --> 00:03:07,000 So data stacks actually provides you the access of Astra DB. 56 00:03:07,000 --> 00:03:13,000 So in order to access them, uh, you know, you can actually do vector search for real world journey 57 00:03:13,000 --> 00:03:14,000 applications. 58 00:03:14,000 --> 00:03:16,000 And it is really, really powerful. 59 00:03:16,000 --> 00:03:17,000 Okay. 60 00:03:17,000 --> 00:03:22,000 And already I have made some videos related to Astra in my previous tutorials also. 61 00:03:22,000 --> 00:03:24,000 So first of all, just go ahead and sign in. 62 00:03:24,000 --> 00:03:30,000 So once you probably sign in then you will be able to see this specific page okay. 63 00:03:30,000 --> 00:03:32,000 Um if you have not signed in. 64 00:03:32,000 --> 00:03:34,000 So first of all you need to sign up over there. 65 00:03:34,000 --> 00:03:38,000 Here you will be able to probably create databases. 66 00:03:38,000 --> 00:03:38,000 Okay. 67 00:03:38,000 --> 00:03:43,000 So in order to create the databases you can actually see over here I will just go ahead and click on 68 00:03:43,000 --> 00:03:45,000 Create Serverless database. 69 00:03:45,000 --> 00:03:46,000 Okay. 70 00:03:47,000 --> 00:03:48,000 Now there are two options over here. 71 00:03:48,000 --> 00:03:50,000 Serverless vector database. 72 00:03:50,000 --> 00:03:51,000 Serverless non vector database. 73 00:03:52,000 --> 00:03:57,000 But as you all know we are planning to probably create serverless vector database. 74 00:03:57,000 --> 00:03:57,000 Okay. 75 00:03:57,000 --> 00:03:59,000 And the reason is very simple. 76 00:03:59,000 --> 00:04:05,000 Whatever data we have or whatever, uh, whatever data in the form of text we have, we are going to 77 00:04:05,000 --> 00:04:06,000 convert that into vector. 78 00:04:06,000 --> 00:04:10,000 And we are going to save in this particular vector database itself. 79 00:04:10,000 --> 00:04:12,000 So first of all I'll click this. 80 00:04:12,000 --> 00:04:14,000 I will give my vector database name. 81 00:04:14,000 --> 00:04:18,000 Let's say this will be test database right. 82 00:04:19,000 --> 00:04:22,000 And here I will go ahead and select any of the region. 83 00:04:22,000 --> 00:04:25,000 And let me just go ahead and click on Create Database. 84 00:04:25,000 --> 00:04:30,000 So as soon as I probably create it it will take some amount of time. 85 00:04:30,000 --> 00:04:32,000 So let this get created. 86 00:04:32,000 --> 00:04:37,000 Till then I will just wait probably to create this entire database and then we will proceed with the 87 00:04:37,000 --> 00:04:38,000 next step. 88 00:04:38,000 --> 00:04:41,000 So yes, let me just go ahead and click this and let me start it. 89 00:04:41,000 --> 00:04:42,000 Okay. 90 00:04:42,000 --> 00:04:43,000 So this is getting created. 91 00:04:43,000 --> 00:04:45,000 It is going to take some amount of time. 92 00:04:45,000 --> 00:04:49,000 So once it is getting created I will probably go with the next step. 93 00:04:49,000 --> 00:04:53,000 So finally guys here you can see that my database has got created. 94 00:04:53,000 --> 00:04:55,000 That is text database a test database. 95 00:04:55,000 --> 00:04:57,000 It is a vector database. 96 00:04:57,000 --> 00:05:02,000 And please make sure that you keep a note of this database ID, because we will be requiring in order 97 00:05:02,000 --> 00:05:03,000 to access this database. 98 00:05:03,000 --> 00:05:07,000 Then the next step will be that we will go ahead and create our tokens. 99 00:05:07,000 --> 00:05:13,000 Now in order to create a token, first of all, you need to probably select the role that you are probably 100 00:05:13,000 --> 00:05:14,000 going ahead with. 101 00:05:14,000 --> 00:05:16,000 So I need administrator user. 102 00:05:16,000 --> 00:05:18,000 I will just go ahead and click on Generate Token. 103 00:05:18,000 --> 00:05:23,000 Now this token will be important because with the help of this particular token I will go ahead and 104 00:05:23,000 --> 00:05:25,000 access my database. 105 00:05:25,000 --> 00:05:25,000 Right. 106 00:05:25,000 --> 00:05:29,000 So if you also want to know you can also download this JSON file. 107 00:05:29,000 --> 00:05:32,000 So let me just go ahead and click and download the JSON file. 108 00:05:32,000 --> 00:05:32,000 Okay. 109 00:05:33,000 --> 00:05:35,000 So now I have my access token. 110 00:05:35,000 --> 00:05:36,000 Now my database is ready. 111 00:05:36,000 --> 00:05:40,000 Now it's time that we go ahead and start our execution of the code. 112 00:05:40,000 --> 00:05:40,000 Okay. 113 00:05:40,000 --> 00:05:45,000 So, uh, before we go ahead, step by step, I will try to show it to you. 114 00:05:45,000 --> 00:05:49,000 Till then, just let me hide my face so that you'll be able to clearly see the code itself. 115 00:05:50,000 --> 00:05:50,000 Okay. 116 00:05:50,000 --> 00:05:53,000 So here I'm going to use Google Colab. 117 00:05:53,000 --> 00:05:54,000 So I will change the runtime. 118 00:05:54,000 --> 00:05:57,000 Let me just use any GPU that I want. 119 00:05:57,000 --> 00:05:59,000 So I'll just go ahead and click on save. 120 00:06:00,000 --> 00:06:03,000 Um, so once I save it, my runtime will be on. 121 00:06:03,000 --> 00:06:04,000 Okay. 122 00:06:04,000 --> 00:06:05,000 Now what I'm actually going to do. 123 00:06:05,000 --> 00:06:09,000 First of all, we will go ahead and install some libraries. 124 00:06:09,000 --> 00:06:15,000 So first of all I require library like lang chain install lang chain. 125 00:06:16,000 --> 00:06:19,000 Then we are going to install lang graph. 126 00:06:19,000 --> 00:06:23,000 Then along with this we are also going to install cashier okay. 127 00:06:23,000 --> 00:06:30,000 So this cashier will be important because I am going to this will be helpful for me to initialize the 128 00:06:30,000 --> 00:06:31,000 Astra DB database itself. 129 00:06:31,000 --> 00:06:32,000 Right. 130 00:06:32,000 --> 00:06:37,000 So I will show you uh, because for initializing Cassandra DB we require this library. 131 00:06:37,000 --> 00:06:38,000 I'll show you as we go ahead. 132 00:06:38,000 --> 00:06:40,000 How why are we using this? 133 00:06:40,000 --> 00:06:44,000 Okay, so here you can see the installation has already taken it. 134 00:06:44,000 --> 00:06:45,000 It has already started. 135 00:06:45,000 --> 00:06:46,000 Okay. 136 00:06:46,000 --> 00:06:48,000 Then the next step after this. 137 00:06:48,000 --> 00:06:57,000 So what we are basically going to do is that we are going to import cashier And then we will provide 138 00:06:57,000 --> 00:07:03,000 the connection of the of the Astra DB. 139 00:07:03,000 --> 00:07:04,000 Okay. 140 00:07:06,000 --> 00:07:09,000 So in order to connect to the Astra DB, we will be using this. 141 00:07:09,000 --> 00:07:15,000 First of all, the important parameter that I actually require is my Astra DB application underscore 142 00:07:15,000 --> 00:07:16,000 token. 143 00:07:16,000 --> 00:07:16,000 Okay. 144 00:07:16,000 --> 00:07:21,000 Now this particular token we can get it from over here. 145 00:07:21,000 --> 00:07:24,000 So here you can see this token information is basically present. 146 00:07:24,000 --> 00:07:28,000 So I'm just going to copy this and I'm going to paste it over here. 147 00:07:28,000 --> 00:07:29,000 Okay. 148 00:07:29,000 --> 00:07:31,000 So here I'm just going to go ahead and paste it. 149 00:07:31,000 --> 00:07:36,000 And with the help of this particular token we will be able to connect to our Astra DB right. 150 00:07:36,000 --> 00:07:41,000 Then we are going to probably use our Astra db id. 151 00:07:41,000 --> 00:07:41,000 Okay. 152 00:07:42,000 --> 00:07:44,000 That is our database ID okay. 153 00:07:44,000 --> 00:07:46,000 So we will be using this. 154 00:07:46,000 --> 00:07:48,000 We have created this particular variable. 155 00:07:48,000 --> 00:07:54,000 Um let's let me go and see this So here is my test database. 156 00:07:54,000 --> 00:07:57,000 And here is my database ID. 157 00:07:57,000 --> 00:07:58,000 Okay. 158 00:07:58,000 --> 00:08:01,000 So this basically becomes my database ID itself right. 159 00:08:02,000 --> 00:08:05,000 Now in order to initialize this I require this two information. 160 00:08:05,000 --> 00:08:13,000 So I will go ahead and write Cachio dot init which is nothing but sorry dot init and will initialize 161 00:08:13,000 --> 00:08:14,000 this particular database. 162 00:08:14,000 --> 00:08:21,000 So token is equal to astra db app store db application token. 163 00:08:21,000 --> 00:08:31,000 And the second parameter will basically be my database database id which will be equal to Astra Astra 164 00:08:31,000 --> 00:08:33,000 db id okay. 165 00:08:33,000 --> 00:08:37,000 So here you can see once I execute this this will get initialized. 166 00:08:37,000 --> 00:08:40,000 And here I have got some warnings but it's okay. 167 00:08:40,000 --> 00:08:44,000 Now we are able to connect to our Astra DB right. 168 00:08:44,000 --> 00:08:45,000 Perfect. 169 00:08:45,000 --> 00:08:50,000 Now the next step will be that since we need to we need to develop the land graph itself. 170 00:08:50,000 --> 00:08:58,000 First of all, I will also install one more very important thing that is pip install lang chain underscore 171 00:08:58,000 --> 00:08:59,000 community. 172 00:08:59,000 --> 00:08:59,000 Okay. 173 00:09:00,000 --> 00:09:04,000 So let me go to go ahead and install this entire application over here. 174 00:09:04,000 --> 00:09:13,000 Once this is installed uh, I still need to install some more things like uh, lang chain grok, uh, 175 00:09:13,000 --> 00:09:16,000 lang chain hub, uh, lang Lang graphs. 176 00:09:16,000 --> 00:09:18,000 There are some more libraries that we are just going to install. 177 00:09:18,000 --> 00:09:24,000 So for that I will just go ahead and copy and paste over here and I'll talk about what all things we 178 00:09:24,000 --> 00:09:25,000 are basically going to install. 179 00:09:26,000 --> 00:09:26,000 Okay. 180 00:09:26,000 --> 00:09:31,000 So here you can see I'll be installing lang chain underscore community lang underscore community I think 181 00:09:31,000 --> 00:09:33,000 it is already been installed. 182 00:09:33,000 --> 00:09:35,000 We will be installing tick token. 183 00:09:35,000 --> 00:09:37,000 I'll talk about this. 184 00:09:37,000 --> 00:09:39,000 Uh we are going to use open source LM models. 185 00:09:39,000 --> 00:09:42,000 So for that we are going to use this Lang chain grok. 186 00:09:42,000 --> 00:09:48,000 Then you have lantern hub chrome adb lang chain lang graph lang chain hugging face. 187 00:09:48,000 --> 00:09:53,000 I will not require chrome adb because I'm specifically going to use this Datastax DB, right? 188 00:09:53,000 --> 00:09:56,000 So line graph line chain hugging face is basically there. 189 00:09:56,000 --> 00:09:59,000 So now let me just go ahead and install all these libraries. 190 00:09:59,000 --> 00:10:05,000 These are some of the basic requirements that I require in order to start my process. 191 00:10:05,000 --> 00:10:06,000 Okay. 192 00:10:06,000 --> 00:10:13,000 Now if you have been seeing my entire lecture series, uh, in this course itself, you'll be able to 193 00:10:13,000 --> 00:10:17,000 see that, um, I have already discussed about Lang Chain. 194 00:10:17,000 --> 00:10:20,000 I have discussed about recursive character text splitter. 195 00:10:20,000 --> 00:10:25,000 So first of all, what I'm actually going to do is that I'm going to read some kind of websites. 196 00:10:25,000 --> 00:10:30,000 You know, from that I will try to read the text data and convert that into vector and store it in a 197 00:10:30,000 --> 00:10:32,000 vector database which is present in Astro DB. 198 00:10:32,000 --> 00:10:33,000 Okay. 199 00:10:33,000 --> 00:10:35,000 So this is basically getting installed. 200 00:10:35,000 --> 00:10:36,000 It will take some amount of time. 201 00:10:36,000 --> 00:10:37,000 Let's wait till then. 202 00:10:38,000 --> 00:10:40,000 Then till then I will go ahead and write my other code. 203 00:10:40,000 --> 00:10:44,000 So here I'm going to use a text splitter which is called as recursive character text splitter. 204 00:10:44,000 --> 00:10:49,000 And since I told you that the data ingestion, I need to read my data from different websites. 205 00:10:49,000 --> 00:10:55,000 So I'll be loading underscore community dot document underscore loaders import web based loader okay. 206 00:10:56,000 --> 00:11:00,000 Then the URL uh which all URLs I'm specifically trying to read. 207 00:11:00,000 --> 00:11:03,000 So these are my three URLs that I'm going to use. 208 00:11:03,000 --> 00:11:09,000 One is related to agent, one is related to prompt engineering and one is related to adversarial attack 209 00:11:09,000 --> 00:11:10,000 LLM okay. 210 00:11:11,000 --> 00:11:15,000 Now what we are basically going to do is that we are going to load the URL. 211 00:11:15,000 --> 00:11:21,000 So for this I will go ahead and write docs is equal to web based web based loader. 212 00:11:21,000 --> 00:11:26,000 I will give my URL dot load instead of just using zero. 213 00:11:26,000 --> 00:11:28,000 I will use for URL in URLs. 214 00:11:28,000 --> 00:11:33,000 And this is basically going to read each and every URL and probably go ahead with the next step. 215 00:11:33,000 --> 00:11:36,000 Then I have automatically. 216 00:11:36,000 --> 00:11:37,000 This is the amazing thing. 217 00:11:38,000 --> 00:11:41,000 Google Colab provides you all the necessary things that is required. 218 00:11:41,000 --> 00:11:44,000 Okay, so here I'm going to use a text splitter. 219 00:11:44,000 --> 00:11:48,000 Now instead of using every time recursive character text splitter. 220 00:11:48,000 --> 00:11:50,000 This time I will show you something in a different way. 221 00:11:50,000 --> 00:11:55,000 So I am going to use something called as from underscore tech token Encoder. 222 00:11:55,000 --> 00:11:59,000 And here we are going to provide the chunk size of 500. 223 00:11:59,000 --> 00:12:03,000 And there will be also a chunk overlap. 224 00:12:03,000 --> 00:12:07,000 Let's say that I'm not interested in putting any kind of overlap. 225 00:12:07,000 --> 00:12:09,000 So I'll keep it to zero okay. 226 00:12:09,000 --> 00:12:14,000 So this way we will be able to do the splitting. 227 00:12:14,000 --> 00:12:19,000 Uh, so one assignment that I really want to give you is that just go ahead and explore about this particular 228 00:12:19,000 --> 00:12:22,000 function why this function is specifically used okay. 229 00:12:22,000 --> 00:12:28,000 The answer is given over here only if you just highlight this text splitter that uses tick token encoder 230 00:12:28,000 --> 00:12:30,000 to count length okay. 231 00:12:30,000 --> 00:12:35,000 So tick token encoder using uses encoding uh name as GPT two. 232 00:12:35,000 --> 00:12:36,000 You can just go ahead and see the definition. 233 00:12:36,000 --> 00:12:42,000 There is a slight difference between recursive character text splitter using this tick token encoders 234 00:12:42,000 --> 00:12:42,000 okay. 235 00:12:43,000 --> 00:12:50,000 Then I'm going to probably do the splitting my documents split, I will go ahead and define this variable. 236 00:12:50,000 --> 00:12:56,000 And I here I'm going to use my document split split underscore documents with respect to docs okay. 237 00:12:56,000 --> 00:13:02,000 Now before I go ahead and probably do this, you know, I will also go ahead and write one more line 238 00:13:02,000 --> 00:13:02,000 of code. 239 00:13:02,000 --> 00:13:05,000 That is my document list. 240 00:13:05,000 --> 00:13:12,000 Here I'm going to consider item for sub list and docs. 241 00:13:12,000 --> 00:13:18,000 I'm going to explore each and every docs for item and sub list. 242 00:13:18,000 --> 00:13:18,000 Okay. 243 00:13:19,000 --> 00:13:28,000 Let me explain about this specific code what exactly this does okay so here I'm saying hey for item 244 00:13:28,000 --> 00:13:28,000 in sub list. 245 00:13:28,000 --> 00:13:31,000 Then we are iterating to sub list in docs okay. 246 00:13:31,000 --> 00:13:33,000 And we are just getting the item okay. 247 00:13:33,000 --> 00:13:38,000 So if you just go ahead and print the document list, you'll be able to see that we'll get all the items 248 00:13:38,000 --> 00:13:39,000 that is displayed okay. 249 00:13:39,000 --> 00:13:45,000 So once I go ahead and execute this you'll be able to see okay let's execute this. 250 00:13:45,000 --> 00:13:47,000 It is going to take some amount of time. 251 00:13:47,000 --> 00:13:56,000 So here it is giving us an error saying that the list object has no attribute page underscore content. 252 00:13:56,000 --> 00:14:02,000 Let me just go ahead and check this out and let me see what is the error that is probably coming up 253 00:14:02,000 --> 00:14:02,000 over here. 254 00:14:02,000 --> 00:14:03,000 So okay. 255 00:14:03,000 --> 00:14:07,000 It should be documents doc underscore list. 256 00:14:07,000 --> 00:14:07,000 Okay. 257 00:14:07,000 --> 00:14:14,000 So if you want to just go ahead and print this doc list you'll be able to see what it prints okay. 258 00:14:15,000 --> 00:14:16,000 Okay. 259 00:14:16,000 --> 00:14:17,000 Again I'll get an error. 260 00:14:17,000 --> 00:14:19,000 So it should be doc underscore list. 261 00:14:20,000 --> 00:14:24,000 So in Doc underscore list you'll be able to see what item is basically getting displayed. 262 00:14:24,000 --> 00:14:29,000 Okay so here is all my information with respect to this. 263 00:14:29,000 --> 00:14:29,000 See. 264 00:14:31,000 --> 00:14:34,000 And then I'm taking this entire information I am splitting it. 265 00:14:34,000 --> 00:14:34,000 Okay. 266 00:14:34,000 --> 00:14:38,000 So this is my entire metadata that I have right. 267 00:14:38,000 --> 00:14:40,000 Then we do the text splitter with respect to chunk size. 268 00:14:40,000 --> 00:14:45,000 And then finally we do the we get the we do the split with the help of text underscore Splitter. 269 00:14:45,000 --> 00:14:45,000 Okay. 270 00:14:46,000 --> 00:14:49,000 Now once I have this dock underscore split. 271 00:14:49,000 --> 00:14:54,000 So if you go ahead and see this this is my dock underscore split okay. 272 00:14:55,000 --> 00:14:59,000 Now if I just go ahead and display this this is my entire splits that you'll be able to see. 273 00:14:59,000 --> 00:15:00,000 Right. 274 00:15:00,000 --> 00:15:01,000 Perfect. 275 00:15:01,000 --> 00:15:07,000 Now I need to take this content and convert this into vectors by using some kind of embeddings. 276 00:15:07,000 --> 00:15:13,000 So for this I am going to use from lang chain underscore hugging face okay. 277 00:15:13,000 --> 00:15:14,000 Hugging face. 278 00:15:14,000 --> 00:15:19,000 And from the hugging face we are going to import hugging face embeddings. 279 00:15:20,000 --> 00:15:26,000 Inside this embeddings I'm going to use hugging face embeddings. 280 00:15:28,000 --> 00:15:30,000 Hugging face embeddings. 281 00:15:30,000 --> 00:15:39,000 And we are going to use a free embedding technique that is available as an open source all dash mini 282 00:15:39,000 --> 00:15:44,000 lm okay L6V2 two. 283 00:15:44,000 --> 00:15:44,000 Okay. 284 00:15:44,000 --> 00:15:49,000 So this name if you see this is already available in the hugging face. 285 00:15:49,000 --> 00:15:52,000 So let's see this. 286 00:15:52,000 --> 00:15:54,000 And here you can see it is a sentence transformer. 287 00:15:54,000 --> 00:16:00,000 And this basically converts it maps sentences and paragraphs to a 384 dimensional dense vector. 288 00:16:00,000 --> 00:16:01,000 Okay. 289 00:16:01,000 --> 00:16:05,000 But for this you also require one very important thing. 290 00:16:05,000 --> 00:16:08,000 If I go and click on settings you'll be requiring a token. 291 00:16:08,000 --> 00:16:08,000 Right. 292 00:16:08,000 --> 00:16:12,000 So this many number of tokens is basically created right in my. 293 00:16:12,000 --> 00:16:16,000 You can go ahead and create a token because this token will be specifically used over here. 294 00:16:16,000 --> 00:16:21,000 You can see in the secret key I've added a token which is called as hugging face right. 295 00:16:21,000 --> 00:16:23,000 Hugging face API token. 296 00:16:23,000 --> 00:16:25,000 So it will be asking for this particular token itself right. 297 00:16:25,000 --> 00:16:27,000 So or h f underscore token. 298 00:16:27,000 --> 00:16:29,000 This token also you can basically specify. 299 00:16:29,000 --> 00:16:33,000 So create a token name provide a token name whatever name you want. 300 00:16:33,000 --> 00:16:39,000 And you should just provide permissions and go ahead and click create this for creating. 301 00:16:39,000 --> 00:16:43,000 You can see over here uh I will just go ahead and write read write. 302 00:16:43,000 --> 00:16:43,000 Okay. 303 00:16:43,000 --> 00:16:45,000 So let's say I want to read. 304 00:16:45,000 --> 00:16:47,000 So here I will just provide my token. 305 00:16:47,000 --> 00:16:51,000 Let's say my token name is something like h f token. 306 00:16:51,000 --> 00:16:53,000 Okay I will go ahead and create the token. 307 00:16:53,000 --> 00:16:57,000 Just copy this token over here and paste it right. 308 00:16:57,000 --> 00:17:00,000 This particular key name h f underscore token and paste it over here. 309 00:17:00,000 --> 00:17:03,000 I've already done that so I'm not going to do that. 310 00:17:03,000 --> 00:17:08,000 I will just go ahead and execute this because this is going to ask me the token information over here. 311 00:17:08,000 --> 00:17:11,000 And I just need to provide the grant access. 312 00:17:11,000 --> 00:17:15,000 So here you can see once this gets executed my token will be ready. 313 00:17:15,000 --> 00:17:18,000 See it is asking for grant access of h f underscore token. 314 00:17:18,000 --> 00:17:22,000 And here you can see it is now able to read the token itself. 315 00:17:22,000 --> 00:17:23,000 Okay. 316 00:17:24,000 --> 00:17:26,000 Now the next step over here. 317 00:17:26,000 --> 00:17:30,000 What we are basically going to do is that we are going to insert all the vectors that we have actually 318 00:17:30,000 --> 00:17:36,000 converted, uh, by using this embedding technique, okay, into our vector database. 319 00:17:36,000 --> 00:17:44,000 So first of all I will go ahead and write from lang chain lang chain Dot vector stores. 320 00:17:44,000 --> 00:17:45,000 Dot. 321 00:17:45,000 --> 00:17:46,000 Cassandra. 322 00:17:47,000 --> 00:17:47,000 Okay. 323 00:17:47,000 --> 00:17:49,000 Import. 324 00:17:49,000 --> 00:17:50,000 Cassandra. 325 00:17:50,000 --> 00:17:51,000 Okay. 326 00:17:51,000 --> 00:17:53,000 So we are going to import Cassandra. 327 00:17:53,000 --> 00:17:58,000 And then we are going to probably create our Astra underscore vector. 328 00:17:58,000 --> 00:18:02,000 Underscore store is equal to Cassandra. 329 00:18:02,000 --> 00:18:07,000 And inside this Cassandra we are going to use our embeddings. 330 00:18:07,000 --> 00:18:12,000 Embedding is equal to embeddings. 331 00:18:14,000 --> 00:18:18,000 And then I have my table underscore name is equal to. 332 00:18:18,000 --> 00:18:22,000 Let me give a table name like QA mini demo. 333 00:18:22,000 --> 00:18:25,000 Okay I'll just go ahead and write this. 334 00:18:25,000 --> 00:18:27,000 I will keep my session as none. 335 00:18:28,000 --> 00:18:35,000 Now what this is basically going to do is that it is just going to insert my data in the database that 336 00:18:35,000 --> 00:18:36,000 I've actually given. 337 00:18:36,000 --> 00:18:37,000 Okay. 338 00:18:37,000 --> 00:18:45,000 So this will basically be none Okay, so almost done. 339 00:18:45,000 --> 00:18:46,000 Any parameters has been left. 340 00:18:46,000 --> 00:18:49,000 Okay, I have to remove this bracket. 341 00:18:49,000 --> 00:18:49,000 Okay. 342 00:18:50,000 --> 00:18:51,000 Now this is done over here. 343 00:18:51,000 --> 00:18:54,000 So let me just go ahead and execute this. 344 00:18:55,000 --> 00:18:59,000 Now this will get connected to my database along with this embeddings. 345 00:18:59,000 --> 00:18:59,000 Okay. 346 00:18:59,000 --> 00:19:02,000 And the table that I'm actually going to create is this particular table. 347 00:19:02,000 --> 00:19:06,000 Now inside this table I need to just go ahead and store my data. 348 00:19:06,000 --> 00:19:07,000 Right. 349 00:19:07,000 --> 00:19:14,000 So first of all what I will do is that I will go ahead and import from lang chain dot indexes. 350 00:19:15,000 --> 00:19:15,000 Indexes. 351 00:19:15,000 --> 00:19:23,000 Dot vector store I'm going to import vector store index rapper not creator index rapper. 352 00:19:23,000 --> 00:19:29,000 Let's say because this will be a rapper on top of the vector store that we have created, the extra 353 00:19:30,000 --> 00:19:31,000 extra vector store. 354 00:19:31,000 --> 00:19:36,000 This will be my vector store itself to probably I can use it to query my vector database. 355 00:19:36,000 --> 00:19:39,000 Okay, so let me delete this. 356 00:19:39,000 --> 00:19:41,000 okay vector store index wrapper. 357 00:19:41,000 --> 00:19:52,000 Now I can use this Astra underscore vector underscore store dot add underscore documents. 358 00:19:54,000 --> 00:19:57,000 And here I'm going to basically use doc split. 359 00:19:57,000 --> 00:19:57,000 Okay. 360 00:19:58,000 --> 00:20:02,000 Then I can go ahead and print inserted 361 00:20:04,000 --> 00:20:10,000 modulus I headlines how many records I'm basically inserting. 362 00:20:10,000 --> 00:20:13,000 I'm just putting that particular information over here. 363 00:20:13,000 --> 00:20:20,000 And here I can use percentile length of the doc split. 364 00:20:21,000 --> 00:20:26,000 So this is basically how the insertion how many records has been inserted okay. 365 00:20:26,000 --> 00:20:34,000 So I'm just going to print this entire thing or inserted print inserted inserted records. 366 00:20:34,000 --> 00:20:39,000 Let me just go ahead and see whether I've done any syntax issue over here Oh yeah. 367 00:20:39,000 --> 00:20:42,000 So docs underscore split. 368 00:20:43,000 --> 00:20:44,000 Perfect, right. 369 00:20:44,000 --> 00:20:46,000 So now it is fine. 370 00:20:46,000 --> 00:20:53,000 Now I'm going to use this extra extra vector D vector index. 371 00:20:53,000 --> 00:21:00,000 And after insertion if I really want to access this I will be using this vector store index wrapper. 372 00:21:00,000 --> 00:21:02,000 And I will be using this extra vector store. 373 00:21:02,000 --> 00:21:03,000 Okay. 374 00:21:03,000 --> 00:21:05,000 So let us go ahead and execute. 375 00:21:05,000 --> 00:21:07,000 First of all it will just go ahead and insert it. 376 00:21:07,000 --> 00:21:08,000 Uh let's see. 377 00:21:08,000 --> 00:21:11,000 Vector store index wrapper from Lang cannot import the name. 378 00:21:12,000 --> 00:21:15,000 The reason is very simple here I made a spelling mistake. 379 00:21:15,000 --> 00:21:17,000 Okay, so here also I made a spelling mistake. 380 00:21:17,000 --> 00:21:19,000 It should be s capital. 381 00:21:19,000 --> 00:21:20,000 So now let's go ahead and execute this. 382 00:21:20,000 --> 00:21:24,000 This line of code is basically going to insert all the documents. 383 00:21:24,000 --> 00:21:25,000 Right. 384 00:21:25,000 --> 00:21:29,000 And then on top of this I am creating a wrapper using vector store index wrapper. 385 00:21:29,000 --> 00:21:32,000 Because of this wrapper you will be able to interact with the vector database. 386 00:21:32,000 --> 00:21:36,000 Okay so here you have this entire thing okay. 387 00:21:36,000 --> 00:21:38,000 And now let me show the next step. 388 00:21:38,000 --> 00:21:44,000 It is obviously true that you cannot just use directly vector database or this particular variable to 389 00:21:44,000 --> 00:21:45,000 access the vector database. 390 00:21:45,000 --> 00:21:49,000 Yes, you can use it, but on top of it we will try to create it as a retriever. 391 00:21:49,000 --> 00:21:56,000 So here I'm going to define my retriever is equal to Astra vector store as retriever okay. 392 00:21:56,000 --> 00:22:00,000 And then I can go ahead and ask any question that I want. 393 00:22:00,000 --> 00:22:05,000 Retriever retriever dot invoke. 394 00:22:05,000 --> 00:22:11,000 Let's say I will go ahead and ask what is agent okay. 395 00:22:12,000 --> 00:22:13,000 And let me just go ahead and execute it. 396 00:22:13,000 --> 00:22:14,000 There's an error. 397 00:22:14,000 --> 00:22:17,000 No worries okay. 398 00:22:17,000 --> 00:22:18,000 What is an agent. 399 00:22:18,000 --> 00:22:21,000 And this is my entire answer right. 400 00:22:21,000 --> 00:22:27,000 Building agents with this this this C is basically coming from the vector database itself when I probably 401 00:22:27,000 --> 00:22:28,000 use this retriever. 402 00:22:28,000 --> 00:22:28,000 Right. 403 00:22:28,000 --> 00:22:32,000 And retriever is basically created on top of vector store. 404 00:22:32,000 --> 00:22:33,000 Right. 405 00:22:33,000 --> 00:22:36,000 Which is basically referring to my vector database Perfect. 406 00:22:36,000 --> 00:22:41,000 So if I go ahead and see this part, I have actually created it, right? 407 00:22:41,000 --> 00:22:45,000 This entire part of Xtradb configuration inserting the data. 408 00:22:45,000 --> 00:22:51,000 And if you also go ahead and see over here, right over here, if I just go ahead and execute it, see, 409 00:22:51,000 --> 00:22:53,000 see the amount of data that has been put up. 410 00:22:53,000 --> 00:22:54,000 Right. 411 00:22:54,000 --> 00:22:58,000 So here is my QA mini demo default key space right. 412 00:22:58,000 --> 00:23:00,000 All the data is basically available over here. 413 00:23:00,000 --> 00:23:03,000 I'm again reloading it so that you'll be able to see this. 414 00:23:03,000 --> 00:23:06,000 If I go ahead and click this you will be able to create a collection. 415 00:23:06,000 --> 00:23:09,000 You'll be able to see this all the information over here right. 416 00:23:09,000 --> 00:23:11,000 And this is great. 417 00:23:11,000 --> 00:23:12,000 Just now you did a query. 418 00:23:12,000 --> 00:23:14,000 And these are all the information that you are able to see. 419 00:23:15,000 --> 00:23:15,000 Perfect. 420 00:23:16,000 --> 00:23:22,000 Now the next step uh, obviously if this is completed I need to create a wiki search. 421 00:23:22,000 --> 00:23:29,000 And for creating wiki search, I really need to go ahead and probably, uh, create and use an external 422 00:23:29,000 --> 00:23:30,000 tool that is called as wiki search. 423 00:23:30,000 --> 00:23:31,000 Okay. 424 00:23:31,000 --> 00:23:36,000 Now before going and creating this wiki search tool, what I'm actually going to do is that I'm going 425 00:23:36,000 --> 00:23:37,000 to define a router function. 426 00:23:37,000 --> 00:23:38,000 Okay. 427 00:23:38,000 --> 00:23:43,000 And if you are seeing my landgraaf videos, I think you should be knowing how to probably go ahead and 428 00:23:43,000 --> 00:23:44,000 create it. 429 00:23:44,000 --> 00:23:47,000 So let's start working on our land graph application. 430 00:23:47,000 --> 00:23:51,000 So here I'm going to basically write land graph application. 431 00:23:51,000 --> 00:23:58,000 And here are three important uh things we are basically going to import okay. 432 00:23:58,000 --> 00:24:05,000 One is one is from typing import literal from land underscore code or prompts. 433 00:24:05,000 --> 00:24:07,000 We are going to use chat prompt template. 434 00:24:07,000 --> 00:24:10,000 And from lang underscore core we are going to use base model and field. 435 00:24:10,000 --> 00:24:16,000 This will be important because when we are basically using or creating our root query we have to inherit 436 00:24:16,000 --> 00:24:17,000 this base model. 437 00:24:18,000 --> 00:24:22,000 So here let me just go ahead and create my data model. 438 00:24:22,000 --> 00:24:26,000 So here I'm writing grass root query base model. 439 00:24:26,000 --> 00:24:29,000 Here I've used data source as two important things. 440 00:24:29,000 --> 00:24:30,000 As a literal right. 441 00:24:30,000 --> 00:24:31,000 One is vector store, one is vector. 442 00:24:31,000 --> 00:24:32,000 See wiki search. 443 00:24:33,000 --> 00:24:39,000 Because this root query class needs to root on any one of them, whether it should go to vector store 444 00:24:39,000 --> 00:24:40,000 or wiki search. 445 00:24:40,000 --> 00:24:43,000 So given a user question, choose to root it to Wikipedia or a vector store. 446 00:24:43,000 --> 00:24:47,000 Okay, so that is the reason we are basically using this particular root query. 447 00:24:47,000 --> 00:24:49,000 So let me just go ahead and execute this. 448 00:24:49,000 --> 00:24:52,000 Because after this we are also going to create some functions. 449 00:24:52,000 --> 00:24:56,000 Now if you are familiar with already working with charts and all. 450 00:24:56,000 --> 00:25:04,000 So here what I'm actually going to do is that I will go ahead and initialize my, uh, use my grok API. 451 00:25:04,000 --> 00:25:08,000 So for grok I will just go to grok.com okay. 452 00:25:09,000 --> 00:25:12,000 Inside this grok dotcom you'll need to sign up or sign in. 453 00:25:12,000 --> 00:25:17,000 And here you can see that it provides you most of the open source model in the form of APIs. 454 00:25:17,000 --> 00:25:18,000 Right. 455 00:25:18,000 --> 00:25:20,000 You need to use the access token. 456 00:25:20,000 --> 00:25:22,000 And you should be able to use any of this. 457 00:25:22,000 --> 00:25:24,000 Even llama 3.1 is also available. 458 00:25:24,000 --> 00:25:24,000 Okay. 459 00:25:24,000 --> 00:25:27,000 So first of all I'll click on Start Building. 460 00:25:27,000 --> 00:25:30,000 I will go ahead and see my grok API. 461 00:25:30,000 --> 00:25:33,000 So I will go ahead and create my API token. 462 00:25:33,000 --> 00:25:34,000 So here you can see API keys. 463 00:25:34,000 --> 00:25:36,000 You can just go ahead and click on API key. 464 00:25:36,000 --> 00:25:39,000 Give us some name and you should be able to create it. 465 00:25:39,000 --> 00:25:42,000 And once you get the secret key you need to update this in the secret key. 466 00:25:42,000 --> 00:25:46,000 So here you can see I have already updated it in the name of grok underscore API key. 467 00:25:46,000 --> 00:25:48,000 And this is my API key right. 468 00:25:48,000 --> 00:25:56,000 So with the same name you can go ahead and call it right now from this grok I'm going to use my model. 469 00:25:56,000 --> 00:25:59,000 That is nothing but llama 3.1 70. 470 00:25:59,000 --> 00:26:01,000 Be versatile or it be instant. 471 00:26:01,000 --> 00:26:03,000 Any one of them we can specifically use. 472 00:26:03,000 --> 00:26:05,000 I will use this okay. 473 00:26:05,000 --> 00:26:07,000 And let's use this. 474 00:26:07,000 --> 00:26:12,000 So before using this first of all we need to import or get our grok API. 475 00:26:12,000 --> 00:26:19,000 So here I will go ahead and execute this um before understanding one thing that if I really want to 476 00:26:19,000 --> 00:26:25,000 access any secret keys from the Google Colab, I need to import from Google Colab, import user data, 477 00:26:25,000 --> 00:26:27,000 and then I'm using User data dot grok API. 478 00:26:27,000 --> 00:26:34,000 So if I just go ahead and print my Grok underscore API underscore key. 479 00:26:34,000 --> 00:26:37,000 You should be able to see my API key right. 480 00:26:38,000 --> 00:26:40,000 Grant access. 481 00:26:40,000 --> 00:26:42,000 And here you can see. 482 00:26:42,000 --> 00:26:47,000 Now if you go ahead and see the secret key the same API key same API key, it is basically referring 483 00:26:47,000 --> 00:26:48,000 to. 484 00:26:48,000 --> 00:26:48,000 Right. 485 00:26:48,000 --> 00:26:52,000 So here is my entire API key that you will be able to see. 486 00:26:52,000 --> 00:26:55,000 Then we will go ahead and initialize our LM model. 487 00:26:55,000 --> 00:27:00,000 So for initializing it I will write LM chat grok. 488 00:27:01,000 --> 00:27:01,000 Okay. 489 00:27:01,000 --> 00:27:05,000 And here I'm going to use grok underscore API. 490 00:27:05,000 --> 00:27:10,000 Underscore key is equal to grok underscore API underscore key. 491 00:27:10,000 --> 00:27:11,000 Okay. 492 00:27:11,000 --> 00:27:14,000 And second parameter I need to give my model underscore name. 493 00:27:14,000 --> 00:27:19,000 This model underscore name we will go ahead and initialize with llama 3.1. 494 00:27:19,000 --> 00:27:22,000 Let's see whether I should be able to copy this or not. 495 00:27:23,000 --> 00:27:25,000 Llama 3.1 70 Be. 496 00:27:25,000 --> 00:27:28,000 Let's say I'll go ahead and write over here. 497 00:27:28,000 --> 00:27:32,000 Llama 3.1. 498 00:27:33,000 --> 00:27:33,000 Okay. 499 00:27:33,000 --> 00:27:34,000 Dash. 500 00:27:34,000 --> 00:27:35,000 Dash 3.1. 501 00:27:37,000 --> 00:27:38,000 Dash 3.1. 502 00:27:38,000 --> 00:27:40,000 Dash 70. 503 00:27:40,000 --> 00:27:41,000 B 70. 504 00:27:41,000 --> 00:27:44,000 B and this is last is versatile. 505 00:27:44,000 --> 00:27:45,000 Okay. 506 00:27:45,000 --> 00:27:47,000 Please make sure that you have the name same. 507 00:27:47,000 --> 00:27:49,000 Otherwise it will not be able to call. 508 00:27:49,000 --> 00:27:50,000 Okay. 509 00:27:50,000 --> 00:27:56,000 Now if I have this what I am going to basically do is that I will just go ahead and use this LM model. 510 00:27:56,000 --> 00:27:57,000 Okay. 511 00:27:58,000 --> 00:28:01,000 We'll just go ahead and print this and see whether we are able to call it or not. 512 00:28:01,000 --> 00:28:05,000 So here you can see I'm able to call this particular model which is amazing. 513 00:28:05,000 --> 00:28:05,000 Okay. 514 00:28:05,000 --> 00:28:12,000 Now one thing that you should know is that our router you know our router this router is really important 515 00:28:12,000 --> 00:28:15,000 because based on a user query I need to give it to vector store or wiki search. 516 00:28:15,000 --> 00:28:15,000 Okay. 517 00:28:15,000 --> 00:28:20,000 So now what we are basically going to do is that we are going to integrate this route query with our 518 00:28:20,000 --> 00:28:21,000 LM model also. 519 00:28:21,000 --> 00:28:22,000 Now how do I do that. 520 00:28:22,000 --> 00:28:23,000 You know. 521 00:28:23,000 --> 00:28:28,000 So see at the end of the day whatever output that I'm actually getting, it should be in this particular 522 00:28:28,000 --> 00:28:28,000 format. 523 00:28:28,000 --> 00:28:38,000 So here I'm going to create a variable called a structured LM underscore router is equal to lm dot width 524 00:28:38,000 --> 00:28:39,000 structured. 525 00:28:39,000 --> 00:28:44,000 So there inside the LM there is something called as which structured output. 526 00:28:44,000 --> 00:28:47,000 And I'm going to use my root query over here. 527 00:28:47,000 --> 00:28:47,000 Okay. 528 00:28:47,000 --> 00:28:49,000 I'm basically going to call my root query class. 529 00:28:50,000 --> 00:28:50,000 Okay. 530 00:28:50,000 --> 00:28:52,000 Now this is fine. 531 00:28:52,000 --> 00:28:52,000 Perfect. 532 00:28:52,000 --> 00:28:58,000 Uh now let's go ahead and check whether everything will work or not with respect to our root query. 533 00:28:58,000 --> 00:28:58,000 Okay. 534 00:28:58,000 --> 00:29:00,000 Root query basically means what? 535 00:29:00,000 --> 00:29:04,000 Whenever we get any user query, it should be able to tell whether it is a wiki search or wiki vector 536 00:29:04,000 --> 00:29:05,000 DB search. 537 00:29:05,000 --> 00:29:06,000 Right. 538 00:29:06,000 --> 00:29:08,000 And for that we have also used one lm model. 539 00:29:08,000 --> 00:29:11,000 And we have used one root query inside it. 540 00:29:11,000 --> 00:29:11,000 Right. 541 00:29:11,000 --> 00:29:15,000 So first of all, what I'm actually going to do is that I'm going to use one prompt. 542 00:29:15,000 --> 00:29:15,000 Okay. 543 00:29:15,000 --> 00:29:17,000 So let's say this is my prompt. 544 00:29:17,000 --> 00:29:19,000 In this prompt I'm writing. 545 00:29:19,000 --> 00:29:23,000 Hey you are an expert at routing a user question to vector Store or Wikipedia. 546 00:29:23,000 --> 00:29:27,000 The vector store contains document related to agents, prompt engineering, and adversarial attacks. 547 00:29:27,000 --> 00:29:30,000 Use the vector store for the question on this topic. 548 00:29:30,000 --> 00:29:31,000 Otherwise, use wiki search. 549 00:29:31,000 --> 00:29:34,000 Okay, so that is what we have written over here. 550 00:29:34,000 --> 00:29:34,000 Okay. 551 00:29:34,000 --> 00:29:37,000 And that is the reason we have used this chat prompt template from message. 552 00:29:37,000 --> 00:29:41,000 We have used system system because this is the instruction that I need to give to my LM model. 553 00:29:41,000 --> 00:29:47,000 And here we are giving the question as with respect to whatever human question is basically given, 554 00:29:47,000 --> 00:29:52,000 then we will go ahead and create a variable which is called as question router. 555 00:29:52,000 --> 00:29:57,000 Here I'm going to use the same root prompt and structured LM router, because this entire task will 556 00:29:57,000 --> 00:29:59,000 be done by an LM model itself. 557 00:29:59,000 --> 00:30:00,000 Right. 558 00:30:00,000 --> 00:30:10,000 So once I execute this this is done right now I will go ahead and write print question underscore router 559 00:30:12,000 --> 00:30:18,000 dot invoke okay dot invoke question underscore router dot invoke. 560 00:30:18,000 --> 00:30:19,000 Let's let's test it. 561 00:30:19,000 --> 00:30:20,000 Okay. 562 00:30:20,000 --> 00:30:24,000 So I will just go ahead and write a question. 563 00:30:24,000 --> 00:30:27,000 Let's say the question is what is an agent. 564 00:30:27,000 --> 00:30:29,000 So if I'm writing, what is an agent? 565 00:30:29,000 --> 00:30:36,000 Obviously my answer should be it should route it to what vector store, because my vector store has 566 00:30:36,000 --> 00:30:38,000 already all this particular information. 567 00:30:38,000 --> 00:30:41,000 But what if I write a question which looks like this? 568 00:30:42,000 --> 00:30:43,000 Who is Shah Rukh Khan? 569 00:30:43,000 --> 00:30:45,000 Shah Rukh Khan is a famous actor in India, right? 570 00:30:45,000 --> 00:30:49,000 So if I go ahead and execute it, you'll be able to see that it is going to Wikipedia search, because 571 00:30:49,000 --> 00:30:52,000 that kind of information is even not present in vector store. 572 00:30:52,000 --> 00:30:57,000 So I don't need to probably push this particular query to my vector store itself. 573 00:30:57,000 --> 00:30:59,000 It should probably go to the wiki search. 574 00:30:59,000 --> 00:31:03,000 So this part of vector store I have already created, right? 575 00:31:03,000 --> 00:31:05,000 I've created a retriever. 576 00:31:05,000 --> 00:31:09,000 Now what is pending is basically for this Wikipedia search right now. 577 00:31:09,000 --> 00:31:11,000 How do I probably go ahead and create this Wikipedia search. 578 00:31:11,000 --> 00:31:16,000 So again in long chain you have this third party tool in Wikipedia. 579 00:31:16,000 --> 00:31:19,000 And for that we need to install this right. 580 00:31:19,000 --> 00:31:21,000 We, we have something like arXiv or Wikipedia. 581 00:31:21,000 --> 00:31:27,000 But I'll just go ahead and install Wikipedia and I'll also install wiki lang underscore community okay. 582 00:31:27,000 --> 00:31:29,000 so here you can see it is already been satisfied. 583 00:31:29,000 --> 00:31:34,000 So it is getting downloaded okay so let's go community. 584 00:31:34,000 --> 00:31:35,000 We have already installed it I guess. 585 00:31:35,000 --> 00:31:36,000 No worries. 586 00:31:36,000 --> 00:31:39,000 Okay now for Wikipedia. 587 00:31:39,000 --> 00:31:46,000 I will probably go ahead and import because here we are working with external tools right lang underscore 588 00:31:46,000 --> 00:31:47,000 community dot utilities. 589 00:31:47,000 --> 00:31:52,000 I'm going to import Wikipedia rapper Wikipedia API wrapper. 590 00:31:52,000 --> 00:32:05,000 Now along with this I'm also going to import from lang chain underscore community dot tools import Wikipedia 591 00:32:06,000 --> 00:32:06,000 query. 592 00:32:06,000 --> 00:32:07,000 Done right. 593 00:32:07,000 --> 00:32:09,000 We're going to import both this. 594 00:32:09,000 --> 00:32:12,000 Now first of all I will go ahead and create my API wrapper. 595 00:32:13,000 --> 00:32:19,000 And this entire thing will basically be responsible in probably running my entire queries. 596 00:32:19,000 --> 00:32:19,000 Right. 597 00:32:19,000 --> 00:32:24,000 So whenever I give a query when it is going to the Wikipedia search, this API wrapper will make sure 598 00:32:24,000 --> 00:32:28,000 that it will do that And here I will just go ahead and define my variable wiki. 599 00:32:28,000 --> 00:32:29,000 Okay. 600 00:32:29,000 --> 00:32:32,000 Inside this I can also go ahead and write my how many results I want. 601 00:32:32,000 --> 00:32:39,000 Let's say if I go ahead and write top k results is equal to one and the max to max document content 602 00:32:39,000 --> 00:32:42,000 character Max, I say that hey, I need 200 characters, okay? 603 00:32:43,000 --> 00:32:45,000 And then I will be able to use this Wikipedia. 604 00:32:45,000 --> 00:32:51,000 So once I execute this, you will be able to see that if I go ahead and ask any question. 605 00:32:51,000 --> 00:32:52,000 Wiki dot. 606 00:32:52,000 --> 00:32:56,000 If I go ahead and write wiki dot run. 607 00:32:57,000 --> 00:32:58,000 Okay. 608 00:32:58,000 --> 00:33:00,000 And let's say if I go ahead and write. 609 00:33:02,000 --> 00:33:09,000 Tell me tell me about Shahrukh Khan. 610 00:33:10,000 --> 00:33:15,000 Okay, so I should be able to get all the response from the Wikipedia itself, right? 611 00:33:15,000 --> 00:33:18,000 So all the information is in Indian actor, film producer who works in Indian cinema. 612 00:33:18,000 --> 00:33:19,000 So 200 characters. 613 00:33:19,000 --> 00:33:20,000 You will be able to see this. 614 00:33:20,000 --> 00:33:25,000 So this in short is basically querying the Wikipedia itself. 615 00:33:25,000 --> 00:33:30,000 Right now my turn is basically to create a graph. 616 00:33:30,000 --> 00:33:31,000 Right. 617 00:33:31,000 --> 00:33:34,000 So we will go ahead and create graph. 618 00:33:34,000 --> 00:33:37,000 Or we'll go ahead and create AI agents. 619 00:33:37,000 --> 00:33:41,000 A AI agent application using line graph. 620 00:33:42,000 --> 00:33:47,000 Because if you understand from our architecture we have also developed this. 621 00:33:47,000 --> 00:33:48,000 We have developed this. 622 00:33:48,000 --> 00:33:49,000 We have also developed router. 623 00:33:49,000 --> 00:33:52,000 Now it's like we need to connect this in the form of graph right. 624 00:33:52,000 --> 00:33:54,000 Like how this graph is entirely there. 625 00:33:54,000 --> 00:33:56,000 So let's go ahead and do that. 626 00:33:56,000 --> 00:33:58,000 So agents application using line graph. 627 00:33:58,000 --> 00:34:02,000 So first of all I will go ahead and create one class. 628 00:34:02,000 --> 00:34:04,000 Let me just go ahead and create one class. 629 00:34:04,000 --> 00:34:09,000 So here I have from typing import list typing extension import Typekit. 630 00:34:09,000 --> 00:34:14,000 So I'm going to create a graph state which will be my class which is inheriting type dictionary. 631 00:34:14,000 --> 00:34:18,000 And this will be in the form of key value pairs like question is equal to STR generation is equal to 632 00:34:18,000 --> 00:34:21,000 x, str and documents list of str okay. 633 00:34:21,000 --> 00:34:23,000 So I'm basically creating three variables. 634 00:34:23,000 --> 00:34:28,000 This three variables will be responsible in managing the entire state of the AI agents that I've actually 635 00:34:28,000 --> 00:34:29,000 created. 636 00:34:29,000 --> 00:34:31,000 So it represents the state of a graph. 637 00:34:31,000 --> 00:34:32,000 Attributes are this many things. 638 00:34:32,000 --> 00:34:32,000 Okay. 639 00:34:32,000 --> 00:34:35,000 So for right now just keep a note of that. 640 00:34:35,000 --> 00:34:36,000 Okay. 641 00:34:36,000 --> 00:34:38,000 You'll be able to understand once we start this. 642 00:34:38,000 --> 00:34:39,000 Okay. 643 00:34:39,000 --> 00:34:42,000 Now let's do one thing. 644 00:34:42,000 --> 00:34:48,000 Whenever let's say whenever I ask or whenever a query is basically coming, whenever we give that query 645 00:34:48,000 --> 00:34:53,000 to the vector database, then for that particular vector database, what we really need to do is that 646 00:34:53,000 --> 00:34:55,000 we need to probably retrieve from the vector database. 647 00:34:55,000 --> 00:34:57,000 And we need to give the response okay. 648 00:34:57,000 --> 00:35:02,000 Whenever we get a response that will basically be stored in this documents, okay. 649 00:35:02,000 --> 00:35:07,000 In this document variable and whatever question user is basically asking it will get stored over here. 650 00:35:07,000 --> 00:35:08,000 Right. 651 00:35:08,000 --> 00:35:14,000 So here I'm going to probably go ahead and create from line chain I'm going to import dot schema. 652 00:35:16,000 --> 00:35:20,000 So dot schema I'm going to import something called as document. 653 00:35:20,000 --> 00:35:21,000 Okay. 654 00:35:21,000 --> 00:35:25,000 Then I will write Definition retrieve. 655 00:35:26,000 --> 00:35:27,000 Definition, retrieve. 656 00:35:27,000 --> 00:35:30,000 Let's say I'm going to take some state over here. 657 00:35:30,000 --> 00:35:35,000 So this state that we have defined it will be a type of this graph state which will have this three 658 00:35:35,000 --> 00:35:35,000 information. 659 00:35:35,000 --> 00:35:36,000 Okay. 660 00:35:36,000 --> 00:35:40,000 And when we are calling this retrieve we are just going to give that particular parameter. 661 00:35:40,000 --> 00:35:46,000 So here what we are basically going to do I will just write down some of the comments so that you'll 662 00:35:46,000 --> 00:35:48,000 be able to understand what is happening inside this function. 663 00:35:48,000 --> 00:35:53,000 So here you can see State dictionary the current graph state and returns a state dictionary itself. 664 00:35:53,000 --> 00:36:03,000 Okay so here what I'm actually going to write I'll write print print retrieve. 665 00:36:06,000 --> 00:36:08,000 Retrieve. 666 00:36:08,000 --> 00:36:10,000 Dash dash dash. 667 00:36:10,000 --> 00:36:12,000 Okay I'm just going to write this print. 668 00:36:12,000 --> 00:36:17,000 And then let me quickly go ahead and write my question. 669 00:36:18,000 --> 00:36:25,000 Question is equal to obviously the question of state which should be coming from this, because the 670 00:36:25,000 --> 00:36:28,000 state, which is a type of graph state, will be keeping in a node of the question. 671 00:36:28,000 --> 00:36:30,000 So we are going to probably consider the state over here. 672 00:36:30,000 --> 00:36:33,000 And then we are going to go to the next line. 673 00:36:33,000 --> 00:36:38,000 Uh, here we are specifically going to do the retrieval or for doing the retrieval I will go ahead and 674 00:36:38,000 --> 00:36:45,000 write documents is equal to retriever dot invoke. 675 00:36:45,000 --> 00:36:49,000 And here I'm going to use the question okay. 676 00:36:49,000 --> 00:36:51,000 Now let me just go ahead and return. 677 00:36:53,000 --> 00:36:54,000 Return. 678 00:36:57,000 --> 00:37:07,000 The documents colon documents whatever documents we are going to specifically get comma. 679 00:37:08,000 --> 00:37:12,000 So this should be my question over here okay. 680 00:37:13,000 --> 00:37:13,000 Oops. 681 00:37:14,000 --> 00:37:15,000 Questions okay. 682 00:37:15,000 --> 00:37:21,000 And the second parameter that I'm going to retrieve is basically my question which will be equal to 683 00:37:21,000 --> 00:37:23,000 this particular questions. 684 00:37:23,000 --> 00:37:25,000 So this is what we are specifically returning. 685 00:37:25,000 --> 00:37:26,000 Okay. 686 00:37:26,000 --> 00:37:28,000 And just through this you will be able to see this. 687 00:37:28,000 --> 00:37:29,000 Okay. 688 00:37:29,000 --> 00:37:34,000 So overall I hope you are able to understand what we are basically doing with this retrieve. 689 00:37:34,000 --> 00:37:39,000 Retrieve basically means we are just going to query it from the retriever, and we are going to just 690 00:37:39,000 --> 00:37:39,000 display it. 691 00:37:39,000 --> 00:37:40,000 Okay. 692 00:37:40,000 --> 00:37:46,000 So I hope everybody was able to understand okay there was some indentation issues. 693 00:37:46,000 --> 00:37:47,000 No worries. 694 00:37:47,000 --> 00:37:50,000 I will just use this. 695 00:37:50,000 --> 00:37:50,000 Let's see. 696 00:37:50,000 --> 00:37:52,000 It should be fixed. 697 00:37:52,000 --> 00:37:52,000 Okay. 698 00:37:52,000 --> 00:37:53,000 Perfect. 699 00:37:53,000 --> 00:37:55,000 So now it gets executed. 700 00:37:55,000 --> 00:37:58,000 So retrieve function is basically done. 701 00:37:58,000 --> 00:38:01,000 Now similarly I will go ahead and do it for the Wikipedia search. 702 00:38:01,000 --> 00:38:04,000 So for that I can just copy and do this. 703 00:38:04,000 --> 00:38:08,000 So for Wikipedia search there will be a functionality where I take the same state. 704 00:38:08,000 --> 00:38:11,000 I will write print Wikipedia over here. 705 00:38:11,000 --> 00:38:11,000 No need to print. 706 00:38:12,000 --> 00:38:12,000 Hello. 707 00:38:12,000 --> 00:38:13,000 I'll take the question. 708 00:38:13,000 --> 00:38:14,000 I'll print the question over here. 709 00:38:14,000 --> 00:38:20,000 Then we'll go ahead and write wiki dot invoke query of question and then whatever docs I'm actually 710 00:38:20,000 --> 00:38:21,000 getting that is my wiki results. 711 00:38:21,000 --> 00:38:27,000 I'll convert that into a document and I hope I have imported the document also. 712 00:38:27,000 --> 00:38:30,000 So this should be imported now, right? 713 00:38:30,000 --> 00:38:31,000 And now this. 714 00:38:31,000 --> 00:38:35,000 Finally, I'm probably doing the Wikipedia search and converting it to a document and returning back 715 00:38:35,000 --> 00:38:36,000 all the results. 716 00:38:36,000 --> 00:38:36,000 Okay. 717 00:38:36,000 --> 00:38:38,000 So perfect. 718 00:38:38,000 --> 00:38:39,000 Uh, so what? 719 00:38:39,000 --> 00:38:42,000 All things have been implemented, we have implemented this functionality. 720 00:38:42,000 --> 00:38:46,000 We have implemented this functionality, we have implemented this functionality, even the LLM part 721 00:38:46,000 --> 00:38:49,000 also we have implemented along with the prompt engineering. 722 00:38:49,000 --> 00:38:51,000 So in short why do we say multi agents. 723 00:38:51,000 --> 00:38:53,000 Because here they are multi agents. 724 00:38:53,000 --> 00:38:54,000 One is communicating with the astrolabe. 725 00:38:55,000 --> 00:38:56,000 One is retrieving the result. 726 00:38:56,000 --> 00:38:56,000 From there. 727 00:38:56,000 --> 00:39:00,000 One is doing the Wikipedia search and all of them are interacting with each other. 728 00:39:00,000 --> 00:39:06,000 So finally uh now it's time that we go ahead and probably do all these things. 729 00:39:06,000 --> 00:39:13,000 But one thing, uh, that we missed out, right, uh, over here is with respect to the routing functionality, 730 00:39:13,000 --> 00:39:17,000 because that routing functionality needs to be also defined as a function. 731 00:39:17,000 --> 00:39:17,000 Right? 732 00:39:17,000 --> 00:39:21,000 So here I will just go ahead and define this. 733 00:39:21,000 --> 00:39:22,000 Let's go ahead and do this. 734 00:39:22,000 --> 00:39:24,000 And I think you should be able to get this. 735 00:39:24,000 --> 00:39:28,000 I'll be getting state of question I'll say question router dot invoke this one okay. 736 00:39:28,000 --> 00:39:31,000 And where do you get the question router. 737 00:39:31,000 --> 00:39:31,000 So question router. 738 00:39:31,000 --> 00:39:33,000 We have already created it earlier right. 739 00:39:33,000 --> 00:39:37,000 If you remember we have created over here question router. 740 00:39:37,000 --> 00:39:39,000 So we are going to use that question router over here. 741 00:39:39,000 --> 00:39:43,000 And here only you will be able to see that we are just writing this particular question. 742 00:39:43,000 --> 00:39:48,000 And I'm saying if source dot data source is wiki search then we return wiki search. 743 00:39:48,000 --> 00:39:49,000 Otherwise it is a vector store. 744 00:39:49,000 --> 00:39:50,000 We return vector store. 745 00:39:51,000 --> 00:39:55,000 So all the functionalities, the chunk of functionalities of all the nodes has been created. 746 00:39:56,000 --> 00:40:03,000 Okay, now it's time that we start creating our entire land graph flow, right? 747 00:40:03,000 --> 00:40:14,000 So I'll write from land graph dot graph import and then you have this state graph. 748 00:40:16,000 --> 00:40:18,000 Then you have this the start. 749 00:40:18,000 --> 00:40:18,000 Okay. 750 00:40:19,000 --> 00:40:22,000 Start now. 751 00:40:22,000 --> 00:40:24,000 The first thing is that I will start my workflow. 752 00:40:24,000 --> 00:40:25,000 I will go ahead and write. 753 00:40:25,000 --> 00:40:28,000 Workflow is equal to state graph. 754 00:40:29,000 --> 00:40:32,000 And here I'm going to use my graph state. 755 00:40:32,000 --> 00:40:35,000 Whatever graph state I've actually defined that is my default class okay. 756 00:40:36,000 --> 00:40:39,000 Then we will go ahead and define the nodes. 757 00:40:39,000 --> 00:40:41,000 What are nodes you have over here. 758 00:40:41,000 --> 00:40:43,000 Main important nodes. 759 00:40:43,000 --> 00:40:45,000 One is the wiki search node and one is the vector db. 760 00:40:45,000 --> 00:40:54,000 Now in order to create the nodes in a line graph, I will just go ahead and write this one workflow 761 00:40:54,000 --> 00:40:57,000 dot add node wiki search. 762 00:40:57,000 --> 00:41:00,000 Then you have this retrieve function. 763 00:41:00,000 --> 00:41:04,000 What is the function name you can see over here retrieve. 764 00:41:04,000 --> 00:41:04,000 Okay. 765 00:41:05,000 --> 00:41:09,000 So I will just go ahead and write this name because the spelling is wrong over there. 766 00:41:09,000 --> 00:41:12,000 So I will go ahead and define this. 767 00:41:12,000 --> 00:41:12,000 Okay. 768 00:41:12,000 --> 00:41:13,000 No worries. 769 00:41:13,000 --> 00:41:17,000 So here uh, one node name is wiki search. 770 00:41:17,000 --> 00:41:19,000 The functionality is this. 771 00:41:19,000 --> 00:41:20,000 Then I have add node. 772 00:41:20,000 --> 00:41:22,000 I'm adding retrieve which is the functionality. 773 00:41:22,000 --> 00:41:23,000 Is this okay? 774 00:41:23,000 --> 00:41:26,000 This is basically do the web search, that is Wikipedia search. 775 00:41:26,000 --> 00:41:28,000 And the other one is basically to do the vector search. 776 00:41:28,000 --> 00:41:29,000 Okay. 777 00:41:29,000 --> 00:41:31,000 Now I'm going to build the entire graph. 778 00:41:31,000 --> 00:41:32,000 This is amazing. 779 00:41:32,000 --> 00:41:33,000 This is just amazing. 780 00:41:33,000 --> 00:41:34,000 Okay. 781 00:41:34,000 --> 00:41:38,000 Workflow dot add conditional logic. 782 00:41:38,000 --> 00:41:43,000 So here I'm also going to go ahead and add conditional logics because this conditional logics will be 783 00:41:43,000 --> 00:41:44,000 really important. 784 00:41:44,000 --> 00:41:44,000 So sorry. 785 00:41:45,000 --> 00:41:47,000 Add conditional edges okay. 786 00:41:47,000 --> 00:41:48,000 Should not be logic. 787 00:41:48,000 --> 00:41:51,000 So first of all we start with start. 788 00:41:51,000 --> 00:41:54,000 Then we go with our root question right. 789 00:41:54,000 --> 00:41:57,000 So here I'll just go ahead and define my root question. 790 00:41:57,000 --> 00:42:01,000 And you know in root question what is basically going to happen. 791 00:42:01,000 --> 00:42:03,000 It is going to go in this two terms right. 792 00:42:03,000 --> 00:42:12,000 One is one is wiki search wiki search colon. 793 00:42:12,000 --> 00:42:16,000 And then I have this wiki underscore search. 794 00:42:16,000 --> 00:42:16,000 Okay. 795 00:42:17,000 --> 00:42:19,000 And the second one is nothing but vector store. 796 00:42:19,000 --> 00:42:23,000 The vector store will go to my which node. 797 00:42:24,000 --> 00:42:27,000 See this is this will basically go to this particular node. 798 00:42:27,000 --> 00:42:28,000 This node is defined over here. 799 00:42:28,000 --> 00:42:31,000 Vector store should go to my retrieve node. 800 00:42:31,000 --> 00:42:34,000 So here I'm going to basically go ahead and define my retrieve node. 801 00:42:35,000 --> 00:42:36,000 Done. 802 00:42:36,000 --> 00:42:37,000 Now this is my root question. 803 00:42:37,000 --> 00:42:38,000 Okay. 804 00:42:38,000 --> 00:42:40,000 So start we have started it over here. 805 00:42:40,000 --> 00:42:46,000 We have put a conditional edges saying that when we are routing whenever this function is written it 806 00:42:46,000 --> 00:42:50,000 should either go to wiki search that is this or this particular node that is retrieve. 807 00:42:50,000 --> 00:42:51,000 Okay. 808 00:42:51,000 --> 00:43:01,000 Then after completing this, what is the next step is that we go ahead and add workflow dot add underscore. 809 00:43:01,000 --> 00:43:03,000 Edge is nothing from from retrieve. 810 00:43:03,000 --> 00:43:04,000 Where do we go after? 811 00:43:04,000 --> 00:43:12,000 Probably if we complete this retrieve where do we go see over here after we do the retrieve. 812 00:43:12,000 --> 00:43:13,000 Retrieve is basically done. 813 00:43:13,000 --> 00:43:15,000 From here we go to our end stage. 814 00:43:15,000 --> 00:43:15,000 Right? 815 00:43:16,000 --> 00:43:17,000 So end stage will go ahead and connect it. 816 00:43:17,000 --> 00:43:20,000 So here I'm going to basically connect my end stage. 817 00:43:20,000 --> 00:43:27,000 And then I will also go ahead and write workflow dot add wiki search should also be going to the end 818 00:43:27,000 --> 00:43:29,000 stage right now. 819 00:43:29,000 --> 00:43:31,000 Finally we do compile this entire graph. 820 00:43:31,000 --> 00:43:39,000 So I will write app is equal to workflow dot compile okay. 821 00:43:39,000 --> 00:43:42,000 And now this entire compile will happen. 822 00:43:42,000 --> 00:43:43,000 Root question is there. 823 00:43:43,000 --> 00:43:46,000 Let's see okay I have to give comma over here. 824 00:43:46,000 --> 00:43:47,000 Then all it will work okay. 825 00:43:48,000 --> 00:43:50,000 So let me just go ahead and execute this. 826 00:43:50,000 --> 00:43:53,000 So root question here two options I have given them. 827 00:43:53,000 --> 00:43:56,000 Then over here that I have this entire workflow. 828 00:43:56,000 --> 00:43:58,000 And all the compilation is specifically happening. 829 00:43:58,000 --> 00:44:05,000 Now in order to just display this entire graph, I can just use this image and display from IPython 830 00:44:05,000 --> 00:44:05,000 display. 831 00:44:05,000 --> 00:44:06,000 Right. 832 00:44:06,000 --> 00:44:11,000 And this is what is the function app dot get graph dot draw mermaid png. 833 00:44:11,000 --> 00:44:13,000 So this I probably show from the documentation. 834 00:44:13,000 --> 00:44:17,000 So here you can see C start vector store retrieve. 835 00:44:17,000 --> 00:44:18,000 Otherwise Wikipedia. 836 00:44:18,000 --> 00:44:20,000 And in between there is something called as router. 837 00:44:20,000 --> 00:44:24,000 Also that router is not getting displayed but the entire flow will happen in that way. 838 00:44:24,000 --> 00:44:25,000 Okay. 839 00:44:25,000 --> 00:44:30,000 Now in order to execute and see how things are basically happening, you can actually check this out, 840 00:44:31,000 --> 00:44:31,000 okay. 841 00:44:31,000 --> 00:44:34,000 And you will be definitely able to see this okay. 842 00:44:35,000 --> 00:44:38,000 So I will just go ahead and print one of the code. 843 00:44:38,000 --> 00:44:44,000 So here you can see what is agent I'm asking the input I can add multiple inputs if I want. 844 00:44:44,000 --> 00:44:46,000 So I'll say for output in app dot stream input. 845 00:44:46,000 --> 00:44:48,000 So key value output items I'm printing the key. 846 00:44:48,000 --> 00:44:54,000 And then we are printing the document of zero dot dictionary dot metadata dot description. 847 00:44:54,000 --> 00:44:56,000 It should be able to get the answer. 848 00:44:56,000 --> 00:45:01,000 So once I execute this here you can see building agents with LM as it is control. 849 00:45:01,000 --> 00:45:06,000 So when I ask this question this is saying hey route the question to rag that is retrieve. 850 00:45:07,000 --> 00:45:09,000 This is the what is I have printed it out there. 851 00:45:09,000 --> 00:45:10,000 Right. 852 00:45:10,000 --> 00:45:13,000 And finally I am able to call this node retrieve function. 853 00:45:13,000 --> 00:45:15,000 And finally I'm getting the output. 854 00:45:15,000 --> 00:45:21,000 Similarly, if I go ahead and ask some other question, let's say I ask a basic questions like Avengers 855 00:45:21,000 --> 00:45:21,000 or something. 856 00:45:21,000 --> 00:45:23,000 Some movies right here. 857 00:45:23,000 --> 00:45:28,000 My question is Avengers, and if I just go ahead and print the same documents you should be able to 858 00:45:28,000 --> 00:45:31,000 see now it is routing to Wikipedia search. 859 00:45:31,000 --> 00:45:32,000 Right? 860 00:45:32,000 --> 00:45:36,000 So this was about this amazing end to end Rag application. 861 00:45:36,000 --> 00:45:43,000 Land graph applications are basically a multi agents where we have added conditional edges. 862 00:45:43,000 --> 00:45:47,000 We have seen how each and every node functionalities will be created and many more things. 863 00:45:47,000 --> 00:45:50,000 So I hope you like this particular video. 864 00:45:50,000 --> 00:45:52,000 Definitely go ahead and check it out. 865 00:45:52,000 --> 00:45:54,000 With respect to astrolabe, because this was quite amazing. 866 00:45:54,000 --> 00:45:57,000 Because of this we are able to do this also. 867 00:45:57,000 --> 00:46:02,000 And the vector database search that we are able to do with the help of retriever. 868 00:46:02,000 --> 00:46:03,000 It was possible. 869 00:46:03,000 --> 00:46:07,000 And again here we have used grok and the entire application has been basically created. 870 00:46:07,000 --> 00:46:07,000 Right. 871 00:46:07,000 --> 00:46:09,000 So this was it from my side. 872 00:46:09,000 --> 00:46:10,000 I hope you like this particular video. 873 00:46:10,000 --> 00:46:11,000 I will see you in the next video. 874 00:46:11,000 --> 00:46:12,000 Thank you. 875 00:46:12,000 --> 00:46:12,000 Take care.