1 00:00:00,000 --> 00:00:01,000 Hello guys. 2 00:00:01,000 --> 00:00:03,000 So we are going to continue the discussion with respect to Lang Chen. 3 00:00:03,000 --> 00:00:07,000 And now I'm going to show you how we can go ahead and build a chat bot. 4 00:00:07,000 --> 00:00:13,000 Already in our previous video we have seen that how to build a simple LM application with LCL. 5 00:00:13,000 --> 00:00:19,000 And we have discussed about Lang Lang chain expression, uh, language where when we were able to, 6 00:00:19,000 --> 00:00:25,000 you know, uh, may, uh, combine the chains together, all the chain components and we were able to 7 00:00:25,000 --> 00:00:25,000 do the prediction. 8 00:00:25,000 --> 00:00:26,000 Right. 9 00:00:26,000 --> 00:00:29,000 So let's go ahead and let's start working with chat bots. 10 00:00:29,000 --> 00:00:35,000 Now in this video we will go over an example of how to design and implement an LLM powered chat bot. 11 00:00:35,000 --> 00:00:39,000 This chat bot will be able to have a conversation and remember previous interaction. 12 00:00:39,000 --> 00:00:44,000 Note that this chatbot that we build will only use the language model to have a conversation. 13 00:00:44,000 --> 00:00:49,000 Okay, there are several other related concepts that you may be looking for, like conversational rag, 14 00:00:49,000 --> 00:00:53,000 like how to enable a chat bot experience over an external source of data. 15 00:00:53,000 --> 00:00:55,000 We will probably discuss this in the upcoming videos. 16 00:00:55,000 --> 00:00:57,000 Then we will also be discussing about agents. 17 00:00:57,000 --> 00:00:59,000 We will be developing multiple projects. 18 00:00:59,000 --> 00:01:03,000 Okay, but in this video we will try to focus on LLM powered chatbots. 19 00:01:03,000 --> 00:01:03,000 Okay. 20 00:01:03,000 --> 00:01:10,000 So first of all, as usual again over here I am going to use the open source uh uh models from grok 21 00:01:10,000 --> 00:01:11,000 API instead of using OpenAI. 22 00:01:12,000 --> 00:01:12,000 Okay. 23 00:01:12,000 --> 00:01:15,000 So first of all what I will say I will go ahead and say import OS. 24 00:01:15,000 --> 00:01:23,000 Uh, along with this, I will go ahead and write from uh dot env import load underscore dot env. 25 00:01:23,000 --> 00:01:26,000 And then we have this load underscore dot env. 26 00:01:26,000 --> 00:01:30,000 Here we are just uh loading all the environment variables. 27 00:01:30,000 --> 00:01:33,000 All the environment variables okay. 28 00:01:33,000 --> 00:01:37,000 And then after this, uh, what do we exactly do. 29 00:01:37,000 --> 00:01:41,000 We basically get our grok underscore API underscore key. 30 00:01:41,000 --> 00:01:46,000 And you know, that we have in our environment variable which we had updated last time. 31 00:01:46,000 --> 00:01:48,000 So this is my grok API key. 32 00:01:48,000 --> 00:01:50,000 So I will just go ahead and call it right. 33 00:01:50,000 --> 00:01:55,000 And in order to call it I will just go ahead and write OS dot get env. 34 00:01:55,000 --> 00:01:58,000 And here I'm just going to call this right. 35 00:01:59,000 --> 00:01:59,000 Perfect. 36 00:01:59,000 --> 00:02:03,000 So if I go ahead and print my grok API key this is the output. 37 00:02:03,000 --> 00:02:05,000 Uh, please do not use this. 38 00:02:05,000 --> 00:02:11,000 Anyhow, uh, after I upload all these videos, uh, in the course, uh, there's all keys will be deleted. 39 00:02:11,000 --> 00:02:14,000 Okay, so now I have my grok API key. 40 00:02:14,000 --> 00:02:17,000 Now the next thing is that we will go ahead and see how to call our LLM model. 41 00:02:17,000 --> 00:02:19,000 So that already we have seen over here. 42 00:02:19,000 --> 00:02:22,000 So here we will be importing chat grok. 43 00:02:22,000 --> 00:02:23,000 Right. 44 00:02:23,000 --> 00:02:26,000 So here you can see lang chain underscore grok import chat grok. 45 00:02:26,000 --> 00:02:30,000 And here we are going to use chat grok with model gamma two nine bit. 46 00:02:30,000 --> 00:02:33,000 This is the most newest model that we are going to use over here. 47 00:02:33,000 --> 00:02:35,000 Let me just go ahead and execute this. 48 00:02:35,000 --> 00:02:37,000 And here I have actually given the grok API key. 49 00:02:37,000 --> 00:02:38,000 Okay. 50 00:02:38,000 --> 00:02:41,000 So this is how you specifically load a model okay. 51 00:02:42,000 --> 00:02:46,000 Um, now after loading a model, what I am actually going to do is that we are going to, first of all, 52 00:02:46,000 --> 00:02:47,000 start with basic things. 53 00:02:47,000 --> 00:02:47,000 Okay. 54 00:02:47,000 --> 00:02:54,000 So let me just go ahead and write from long chain underscore code I am going to go ahead and import 55 00:02:54,000 --> 00:02:55,000 this messages. 56 00:02:55,000 --> 00:03:00,000 And from this messages I'll import the first type of message which is called as human message. 57 00:03:00,000 --> 00:03:00,000 Okay. 58 00:03:00,000 --> 00:03:06,000 Whenever I have a human message that basically whenever I'm using this particular class, uh, it indicates 59 00:03:06,000 --> 00:03:08,000 that a user is giving a message itself. 60 00:03:08,000 --> 00:03:08,000 Right. 61 00:03:08,000 --> 00:03:12,000 So let me just go ahead and write model dot invoke over here. 62 00:03:12,000 --> 00:03:16,000 And uh, with respect to this, let me go ahead and create my human message, uh, class. 63 00:03:16,000 --> 00:03:17,000 Let me initialize it. 64 00:03:17,000 --> 00:03:21,000 And here we are going to talk about content is equal to. 65 00:03:21,000 --> 00:03:35,000 And here I'm going to say hey hi my name is Chris and I am, uh, chief AI engineer. 66 00:03:35,000 --> 00:03:35,000 Okay. 67 00:03:35,000 --> 00:03:39,000 So I'm just giving this kind of message over here and let me go ahead and invoke this. 68 00:03:39,000 --> 00:03:44,000 So here you'll be able to see as soon as I give the message over here, this we are invoking with respect 69 00:03:44,000 --> 00:03:49,000 to this particular model, that basically means this information is going to my, uh, model that is 70 00:03:49,000 --> 00:03:50,000 gamma2 over here. 71 00:03:50,000 --> 00:03:51,000 And I'm getting the response. 72 00:03:51,000 --> 00:03:52,000 Hello, Chris. 73 00:03:52,000 --> 00:03:54,000 It's nice to meet you as a chief engineer. 74 00:03:54,000 --> 00:03:57,000 What kind of projects are you working on these days? 75 00:03:57,000 --> 00:04:02,000 I'm always eager to learn about the exciting work being done in the field of AI, and these are my all 76 00:04:02,000 --> 00:04:07,000 the metadata info Information like how much is the input tokens, what is the output tokens, and what 77 00:04:07,000 --> 00:04:08,000 is the total number of tokens. 78 00:04:08,000 --> 00:04:14,000 So here, uh, just a simple conversation by using a human message, uh, class, I've actually done 79 00:04:14,000 --> 00:04:14,000 it right. 80 00:04:14,000 --> 00:04:17,000 Still, I have not initialized anything from the system message itself. 81 00:04:17,000 --> 00:04:23,000 I have not told, uh, like, what kind of behavior the LM model should probably go ahead with right 82 00:04:23,000 --> 00:04:24,000 now. 83 00:04:24,000 --> 00:04:24,000 Let me do one thing. 84 00:04:24,000 --> 00:04:25,000 Let me. 85 00:04:25,000 --> 00:04:26,000 And you know this, right? 86 00:04:26,000 --> 00:04:30,000 Whenever we get this AI message, that basically means we are getting the response from our model. 87 00:04:30,000 --> 00:04:32,000 Now let me go ahead and write from Long Chain. 88 00:04:33,000 --> 00:04:33,000 Oops. 89 00:04:34,000 --> 00:04:38,000 Uh, over here I'll go ahead and write from Long Chain. 90 00:04:38,000 --> 00:04:38,000 Okay. 91 00:04:38,000 --> 00:04:44,000 Now, I think I was searching it over here from long chain underscore core dot messages. 92 00:04:44,000 --> 00:04:48,000 I'm also going to import I message okay. 93 00:04:48,000 --> 00:04:51,000 Now again here I'll go ahead and write model dot invoke. 94 00:04:51,000 --> 00:04:56,000 Uh, and let me put some combination of AI and human message. 95 00:04:56,000 --> 00:05:02,000 Now understand, one thing is that, uh, we can give a list of messages like how we basically developed 96 00:05:02,000 --> 00:05:03,000 a simple LM application. 97 00:05:03,000 --> 00:05:03,000 Right? 98 00:05:03,000 --> 00:05:05,000 So let's say first message is my human message. 99 00:05:05,000 --> 00:05:11,000 And here I will go ahead and write content with respect to this entire information. 100 00:05:11,000 --> 00:05:11,000 Hi. 101 00:05:11,000 --> 00:05:14,000 My name is so and so and I'm a chief engineer. 102 00:05:14,000 --> 00:05:15,000 I'll put it over here itself. 103 00:05:15,000 --> 00:05:16,000 Okay? 104 00:05:16,000 --> 00:05:21,000 Uh, and I will close this quickly, and then I will go ahead and take my next message. 105 00:05:21,000 --> 00:05:25,000 The next message will be basically, uh, AI message. 106 00:05:25,000 --> 00:05:29,000 And here in the AI message, I'm just going to go ahead and write this particular content and let me 107 00:05:29,000 --> 00:05:32,000 just go ahead and copy this entire thing, whatever content is over here. 108 00:05:32,000 --> 00:05:33,000 Right. 109 00:05:33,000 --> 00:05:35,000 Uh, in the field of AI. 110 00:05:35,000 --> 00:05:35,000 Right. 111 00:05:35,000 --> 00:05:39,000 So I'm just going to go ahead and write this particular content over here. 112 00:05:39,000 --> 00:05:39,000 Right. 113 00:05:39,000 --> 00:05:43,000 So this is the I, uh, message that I've actually got a response. 114 00:05:43,000 --> 00:05:47,000 So now here I'm actually going to create a sequence of messages from this particular text. 115 00:05:47,000 --> 00:05:49,000 I got the I'm just hard coding. 116 00:05:49,000 --> 00:05:49,000 Okay. 117 00:05:49,000 --> 00:05:52,000 I got this particular message from this particular model. 118 00:05:52,000 --> 00:05:56,000 And now I will go ahead and just trace again a human message. 119 00:05:56,000 --> 00:06:09,000 And now I will go and ask, hey, hey, Hey, what's my name and what do I do? 120 00:06:09,000 --> 00:06:12,000 Okay, so this is the question that I'm going to ask over here. 121 00:06:13,000 --> 00:06:14,000 Now, this is amazing. 122 00:06:14,000 --> 00:06:15,000 Okay, let's see the output. 123 00:06:15,000 --> 00:06:21,000 What I'm actually going to do because I'm, I'm first of all feeding this particular information with 124 00:06:21,000 --> 00:06:22,000 respect to the human message. 125 00:06:22,000 --> 00:06:25,000 Hey, my name is Krish and I am a chief AI engineer. 126 00:06:25,000 --> 00:06:28,000 And from this AI message I got this entire value. 127 00:06:28,000 --> 00:06:30,000 So I am probably putting it up. 128 00:06:30,000 --> 00:06:34,000 And I'm asking it whether it is able to remember my no. 129 00:06:34,000 --> 00:06:39,000 It is whether it is able to remember the information that I have actually given since I'm passing it 130 00:06:39,000 --> 00:06:42,000 in the form of list of messages here. 131 00:06:42,000 --> 00:06:42,000 Right. 132 00:06:43,000 --> 00:06:45,000 I think it should be able to remember it. 133 00:06:45,000 --> 00:06:45,000 Okay. 134 00:06:45,000 --> 00:06:46,000 So now once I go ahead and execute it. 135 00:06:46,000 --> 00:06:51,000 So here you can say you said your name is Krish and you are a chief AI engineer, right? 136 00:06:51,000 --> 00:06:54,000 Is there anything else you would like me to remember about you? 137 00:06:54,000 --> 00:06:57,000 Okay, so all this information is basically coming. 138 00:06:57,000 --> 00:07:02,000 That basically means, uh, one important thing is that whatever conversation I'm giving inside this 139 00:07:02,000 --> 00:07:05,000 list of messages with the human message, I message human message. 140 00:07:05,000 --> 00:07:08,000 It is also able to remember the previous context. 141 00:07:08,000 --> 00:07:08,000 Okay. 142 00:07:08,000 --> 00:07:14,000 Now that is where we are going to discuss about message history okay. 143 00:07:14,000 --> 00:07:19,000 And this message history will probably help you to understand the all the context that it will help 144 00:07:19,000 --> 00:07:22,000 the LM to remember all the context that we are probably discussing about. 145 00:07:22,000 --> 00:07:29,000 This is just an idea that whether the I, the LM model is able to remember things or not. 146 00:07:29,000 --> 00:07:34,000 But at the end of the day, whenever we develop real world application, there will be different, different 147 00:07:34,000 --> 00:07:35,000 sessions that will be happening. 148 00:07:35,000 --> 00:07:36,000 Right? 149 00:07:36,000 --> 00:07:37,000 Let's say I'm using ChatGPT. 150 00:07:37,000 --> 00:07:43,000 Someone is using ChatGPT, some other people are using some other LLM models and they are chatting. 151 00:07:43,000 --> 00:07:43,000 Right. 152 00:07:43,000 --> 00:07:49,000 How that particular uh, models are able to remember their context, that is with respect to session. 153 00:07:49,000 --> 00:07:55,000 So in order to understand that how these sessions are managed, we will be discussing about a very important 154 00:07:55,000 --> 00:07:57,000 property which is called as message history. 155 00:07:57,000 --> 00:07:57,000 Okay. 156 00:07:57,000 --> 00:08:01,000 Now let's discuss about this message history. 157 00:08:01,000 --> 00:08:06,000 And this is a very important component altogether in the, uh, long chain itself. 158 00:08:06,000 --> 00:08:07,000 Right. 159 00:08:07,000 --> 00:08:10,000 Which will actually help you to work with the LM model. 160 00:08:10,000 --> 00:08:10,000 Okay. 161 00:08:11,000 --> 00:08:17,000 Now, what we really need to use or understand how message history works. 162 00:08:17,000 --> 00:08:21,000 So here I'm going to put a message saying that we can use message history class to wrap our model and 163 00:08:21,000 --> 00:08:22,000 make it stateful. 164 00:08:22,000 --> 00:08:27,000 We have to make it stateful so that it remembers all the context with respect to any kind of person 165 00:08:27,000 --> 00:08:30,000 who are actually interacting with that model. 166 00:08:30,000 --> 00:08:34,000 This will keep track of our inputs and output of the model and store them in some data store. 167 00:08:34,000 --> 00:08:38,000 Further interaction will then load these messages and pass them into the chain as part of the input. 168 00:08:38,000 --> 00:08:42,000 Okay, so let's see how this can be actually done. 169 00:08:42,000 --> 00:08:49,000 So first of all for this we require another library which is called as long chain underscore community. 170 00:08:49,000 --> 00:08:51,000 So please make sure that you install this. 171 00:08:51,000 --> 00:08:55,000 And once you install this I have already done this particular installation so that it does not take 172 00:08:55,000 --> 00:08:56,000 much time. 173 00:08:56,000 --> 00:08:58,000 You have to go ahead and update in the requirement dot txt okay. 174 00:08:58,000 --> 00:09:03,000 So first of all go ahead and update this in the requirement dot txt and make sure that whatever installation 175 00:09:03,000 --> 00:09:06,000 that you have done, you are doing it in your, uh, virtual environment. 176 00:09:06,000 --> 00:09:07,000 Okay. 177 00:09:08,000 --> 00:09:08,000 Perfect. 178 00:09:08,000 --> 00:09:15,000 So from here, uh, what I will do is that I, uh, Langton community is basically used because, uh, 179 00:09:15,000 --> 00:09:20,000 uh, any integration related to measure history will be available in this Langton underscore community. 180 00:09:20,000 --> 00:09:20,000 Okay. 181 00:09:20,000 --> 00:09:27,000 So, uh, let me go ahead and import some very important libraries, let's say from long chain underscore 182 00:09:27,000 --> 00:09:28,000 community. 183 00:09:28,000 --> 00:09:31,000 Uh, I will be using something called as chat message histories. 184 00:09:31,000 --> 00:09:31,000 Okay. 185 00:09:31,000 --> 00:09:38,000 Inside this chat message history we will be importing chat chat message history okay I will talk about 186 00:09:38,000 --> 00:09:38,000 this. 187 00:09:38,000 --> 00:09:40,000 What exactly is this chat message history? 188 00:09:40,000 --> 00:09:43,000 Just for now, let me just import some more libraries. 189 00:09:43,000 --> 00:09:51,000 So I will go ahead and write lecture underscore code dot chat history okay I'm going to import bass 190 00:09:51,000 --> 00:09:53,000 chat history okay. 191 00:09:53,000 --> 00:09:54,000 Bass chat message history. 192 00:09:55,000 --> 00:10:00,000 And then finally I will go ahead and import from luncheon underscore core dot runnables okay. 193 00:10:00,000 --> 00:10:06,000 Since this is for the chain and for the rain chain, also, we basically go ahead and uh, import this 194 00:10:06,000 --> 00:10:10,000 particular history and I will go ahead and import runnable with message history. 195 00:10:11,000 --> 00:10:14,000 Now this is something really important okay. 196 00:10:14,000 --> 00:10:19,000 Here you'll be able to see that we can import the relevant class over here. 197 00:10:19,000 --> 00:10:23,000 Set up our chain uh which wraps the model and add in the message history. 198 00:10:23,000 --> 00:10:25,000 That is what we really need to do, right? 199 00:10:25,000 --> 00:10:32,000 Every message that we are probably putting up, it needs to be, uh, added in the message history itself, 200 00:10:32,000 --> 00:10:32,000 how it is done. 201 00:10:32,000 --> 00:10:34,000 I'll just talk about it sometime. 202 00:10:34,000 --> 00:10:35,000 Okay. 203 00:10:35,000 --> 00:10:41,000 But the thing that you really need to understand is that whenever different, different users are chatting 204 00:10:41,000 --> 00:10:47,000 with the LM model, how we are going to make sure that one session is completely different from the 205 00:10:47,000 --> 00:10:48,000 other session. 206 00:10:48,000 --> 00:10:53,000 So for that reason, what we will do is that we will create one amazing function, okay. 207 00:10:53,000 --> 00:10:59,000 And this particular function will be definition get underscore session underscore history. 208 00:10:59,000 --> 00:11:01,000 Now what this function is going to do. 209 00:11:01,000 --> 00:11:03,000 Let me talk about it okay. 210 00:11:03,000 --> 00:11:11,000 Here this function will be creating a session ID okay this session ID and this session id will be a 211 00:11:11,000 --> 00:11:12,000 string okay. 212 00:11:13,000 --> 00:11:16,000 Right I've written a session ID and this will be a string. 213 00:11:16,000 --> 00:11:21,000 And this will get inherited to your base I mean the return type. 214 00:11:21,000 --> 00:11:21,000 Sorry. 215 00:11:21,000 --> 00:11:22,000 Not inherited. 216 00:11:22,000 --> 00:11:26,000 The return type of this particular function will be base chat message history. 217 00:11:27,000 --> 00:11:27,000 So whatever. 218 00:11:27,000 --> 00:11:33,000 Chat history is basically getting created right that is imported from this particular library. 219 00:11:33,000 --> 00:11:38,000 So what I'm actually going to do is that this get session history right with respect to a session ID 220 00:11:38,000 --> 00:11:39,000 will be created. 221 00:11:39,000 --> 00:11:39,000 Right. 222 00:11:39,000 --> 00:11:44,000 And this session ID will be used to distinguish one chat session with the other. 223 00:11:44,000 --> 00:11:45,000 Okay. 224 00:11:45,000 --> 00:11:48,000 So here, uh, the return type is based chat history. 225 00:11:48,000 --> 00:11:53,000 I will say hey if session ID not in store. 226 00:11:53,000 --> 00:11:58,000 So let me just go ahead and create one variable called as store which will be a dictionary over here. 227 00:11:58,000 --> 00:12:01,000 So here I will just go ahead and write store okay. 228 00:12:01,000 --> 00:12:01,000 Okay. 229 00:12:02,000 --> 00:12:03,000 Let me talk more about it. 230 00:12:03,000 --> 00:12:05,000 And the explanation will be very good. 231 00:12:05,000 --> 00:12:08,000 And I will show you the practical example. 232 00:12:08,000 --> 00:12:19,000 I will go ahead and write store of session underscore ID is equal to chat message history okay. 233 00:12:20,000 --> 00:12:22,000 So now see this everything. 234 00:12:22,000 --> 00:12:24,000 Please try to understand it right. 235 00:12:24,000 --> 00:12:28,000 First of all, for this particular message history. 236 00:12:28,000 --> 00:12:34,000 So you can see that we have a class which is called as chat message history. 237 00:12:34,000 --> 00:12:34,000 Okay. 238 00:12:34,000 --> 00:12:41,000 Now this chat message history, whenever we are creating a session ID right, we are storing the session 239 00:12:41,000 --> 00:12:43,000 ID in this particular dictionary. 240 00:12:43,000 --> 00:12:43,000 Right. 241 00:12:43,000 --> 00:12:49,000 With respect to this particular value, whatever session ID I'm actually creating, I am I'm initializing 242 00:12:49,000 --> 00:12:51,000 a chat message history. 243 00:12:51,000 --> 00:12:51,000 Right. 244 00:12:51,000 --> 00:12:54,000 So that basically means this is an object of a chat message history. 245 00:12:54,000 --> 00:13:00,000 And whatever chat is specifically happening automatically, it should be going inside this session ID 246 00:13:00,000 --> 00:13:00,000 okay. 247 00:13:00,000 --> 00:13:08,000 Next thing is that after this I will just go ahead and return store of session underscore ID. 248 00:13:08,000 --> 00:13:10,000 Now see you will. 249 00:13:10,000 --> 00:13:15,000 You know I know there is some confusion with respect to the definition, but let's understand this definition. 250 00:13:15,000 --> 00:13:17,000 So what is the definition over here? 251 00:13:17,000 --> 00:13:20,000 This is a function which is called as get session history. 252 00:13:20,000 --> 00:13:26,000 Whenever I give a session ID, it should be able to check whether the session ID is present in this 253 00:13:26,000 --> 00:13:27,000 particular dictionary or not. 254 00:13:27,000 --> 00:13:33,000 If it is present, we are going to get the entire chat message history from that and we will return 255 00:13:33,000 --> 00:13:34,000 it. 256 00:13:34,000 --> 00:13:34,000 Okay. 257 00:13:34,000 --> 00:13:36,000 And that is how we are going to distinguish it. 258 00:13:37,000 --> 00:13:43,000 Two things this get session history will give you a response type of base chat history. 259 00:13:43,000 --> 00:13:46,000 Uh, it is an abstract class for storing the chat message history. 260 00:13:46,000 --> 00:13:52,000 If I talk about chat message history it is an in-memory implementation of chat message history stores. 261 00:13:52,000 --> 00:13:54,000 Messages are in an memory list. 262 00:13:54,000 --> 00:13:56,000 So it is very much simple. 263 00:13:56,000 --> 00:13:58,000 Whenever I give a session ID it should. 264 00:13:58,000 --> 00:14:02,000 First of all go ahead and check in this particular dictionary whether it is available or not. 265 00:14:02,000 --> 00:14:07,000 If it is available, it will go ahead and pick up the entire chat message that we have discussed with 266 00:14:07,000 --> 00:14:11,000 respect to, or whatever questions we have asked with respect to the LM model for that session ID, 267 00:14:11,000 --> 00:14:15,000 and it is just going to return it now, see how it is going to work. 268 00:14:15,000 --> 00:14:15,000 Okay. 269 00:14:15,000 --> 00:14:16,000 So I'll execute this. 270 00:14:17,000 --> 00:14:19,000 Uh, I will talk about this runnable with message history. 271 00:14:19,000 --> 00:14:21,000 Uh, soon it will be used as we go ahead. 272 00:14:21,000 --> 00:14:22,000 Okay. 273 00:14:22,000 --> 00:14:27,000 Now the best thing over here will be that we will be seeing how we will be using this. 274 00:14:27,000 --> 00:14:28,000 Okay? 275 00:14:28,000 --> 00:14:32,000 So first of all, we will go ahead and create our config okay. 276 00:14:33,000 --> 00:14:35,000 So I'll write hey this is my config over here. 277 00:14:35,000 --> 00:14:40,000 And this config will be nothing but it will called as configurable. 278 00:14:40,000 --> 00:14:44,000 Okay I will just give this key value pairs configurable. 279 00:14:44,000 --> 00:14:46,000 And here I'm just going to give my session ID. 280 00:14:46,000 --> 00:14:49,000 So let's say my session ID is something like this. 281 00:14:49,000 --> 00:14:51,000 Right now I'm hard coding it. 282 00:14:51,000 --> 00:14:59,000 Let's say my session ID is something like um chat one okay I'm just going to write it as chat one. 283 00:14:59,000 --> 00:15:01,000 Now let's go ahead and execute this okay. 284 00:15:02,000 --> 00:15:03,000 Let's go ahead and execute. 285 00:15:03,000 --> 00:15:05,000 So this basically becomes my config okay. 286 00:15:05,000 --> 00:15:07,000 My session ID is chat one. 287 00:15:07,000 --> 00:15:11,000 Now let me use this particular session ID and let me chat with the LM model. 288 00:15:11,000 --> 00:15:17,000 So here you'll be able to see that I will write response is equal to width. 289 00:15:17,000 --> 00:15:22,000 Uh, so before this let me do one more thing okay. 290 00:15:22,000 --> 00:15:26,000 Uh, see, I have used the Get session history right now. 291 00:15:26,000 --> 00:15:29,000 In order to execute this, I really need to use this runnable. 292 00:15:29,000 --> 00:15:30,000 Right. 293 00:15:30,000 --> 00:15:35,000 Unless and until we don't have this runnable, uh, it is not going to probably execute this entire 294 00:15:35,000 --> 00:15:35,000 chain. 295 00:15:35,000 --> 00:15:39,000 So here you say see runnable that manage chat message history for other runnables. 296 00:15:39,000 --> 00:15:39,000 Okay. 297 00:15:39,000 --> 00:15:44,000 The chat message history is a sequence of messages that represent a conversation. 298 00:15:44,000 --> 00:15:49,000 So here what I'm actually going to do I'll go ahead and write width underscore message underscore history. 299 00:15:49,000 --> 00:15:53,000 And here I will go ahead and write a runnable with message history. 300 00:15:53,000 --> 00:15:58,000 And here the first parameter will get is model whatever model we are going to use. 301 00:15:58,000 --> 00:16:01,000 And the second is get session history Okay. 302 00:16:01,000 --> 00:16:05,000 So this two parameters will be specifically required, uh, for this right. 303 00:16:05,000 --> 00:16:08,000 And this width session with message history. 304 00:16:08,000 --> 00:16:12,000 We will try to invoke this as soon as we invoke this. 305 00:16:12,000 --> 00:16:14,000 This is going to take this particular model. 306 00:16:14,000 --> 00:16:16,000 This is going to take the entire chat history. 307 00:16:16,000 --> 00:16:21,000 What we are getting based on the session ID again how to pass the session ID I'll let you know. 308 00:16:21,000 --> 00:16:21,000 Okay. 309 00:16:21,000 --> 00:16:25,000 And then it will be able to give you the entire information. 310 00:16:25,000 --> 00:16:31,000 So here, let me just go ahead and write double s, uh, because I don't want to write triple s. 311 00:16:31,000 --> 00:16:31,000 Okay. 312 00:16:31,000 --> 00:16:34,000 So here so many S's are here okay. 313 00:16:34,000 --> 00:16:35,000 Perfect. 314 00:16:36,000 --> 00:16:36,000 Right. 315 00:16:36,000 --> 00:16:38,000 So again let me repeat it. 316 00:16:38,000 --> 00:16:39,000 What we have exactly done. 317 00:16:39,000 --> 00:16:41,000 First of all we have created a dictionary over here. 318 00:16:41,000 --> 00:16:48,000 Then we have created a function which will be able to retrieve whatever uh, chat history is there for 319 00:16:48,000 --> 00:16:49,000 a specific session ID. 320 00:16:49,000 --> 00:16:56,000 Then we are converting, or we are using this runnable with message history so that we can, um, you 321 00:16:56,000 --> 00:16:58,000 know, interact with our LM model based on the chat history. 322 00:16:58,000 --> 00:17:01,000 So for this we give two parameters that is model and get session history. 323 00:17:01,000 --> 00:17:05,000 Now we have created a configuration called as session ID as chat one. 324 00:17:05,000 --> 00:17:10,000 And now what we will do is that we will call this with message history. 325 00:17:10,000 --> 00:17:16,000 And here I'm just going to go ahead and invoke invoke over here. 326 00:17:16,000 --> 00:17:18,000 First of all I will give a human message. 327 00:17:18,000 --> 00:17:21,000 Let's say this is my human message over here. 328 00:17:21,000 --> 00:17:22,000 Now see see the magic okay. 329 00:17:22,000 --> 00:17:23,000 This will be amazing. 330 00:17:23,000 --> 00:17:24,000 I will go ahead and write. 331 00:17:24,000 --> 00:17:27,000 This is my content is equal to. 332 00:17:27,000 --> 00:17:29,000 And let me say hi. 333 00:17:29,000 --> 00:17:32,000 Uh, or I'll give the same message over here. 334 00:17:32,000 --> 00:17:32,000 Right. 335 00:17:32,000 --> 00:17:36,000 So hi, my name is Krish and I am an AI engineer. 336 00:17:36,000 --> 00:17:37,000 Okay. 337 00:17:37,000 --> 00:17:41,000 Now see, I am trying to interact with based on the session ID. 338 00:17:41,000 --> 00:17:43,000 Okay, now see this all information is over here. 339 00:17:44,000 --> 00:17:48,000 But along with this we can go ahead and give our configuration. 340 00:17:48,000 --> 00:17:52,000 The configuration will be given inside this invoke. 341 00:17:52,000 --> 00:17:53,000 And here I will go ahead and write. 342 00:17:53,000 --> 00:17:56,000 This is my config which will be equal to config. 343 00:17:57,000 --> 00:18:03,000 Now when we are giving the config that basically means for this session id we are interacting okay. 344 00:18:03,000 --> 00:18:06,000 So based on the session id it will be able to remember all this context. 345 00:18:06,000 --> 00:18:09,000 So if I go ahead and execute this okay. 346 00:18:09,000 --> 00:18:12,000 So with message history is not defined because I did not execute this. 347 00:18:12,000 --> 00:18:13,000 Let me execute it. 348 00:18:13,000 --> 00:18:15,000 Now let's go ahead and do this. 349 00:18:15,000 --> 00:18:18,000 So here you can see I message hi crush. 350 00:18:18,000 --> 00:18:19,000 It's great to meet you. 351 00:18:19,000 --> 00:18:22,000 Being a chief engineer is an exciting role. 352 00:18:22,000 --> 00:18:28,000 All this information is basically coming Okay, now if you want to get this particular content, I will 353 00:18:28,000 --> 00:18:30,000 also save this in my response variable. 354 00:18:30,000 --> 00:18:30,000 Okay. 355 00:18:30,000 --> 00:18:34,000 And let me just go ahead and write response dot content okay. 356 00:18:34,000 --> 00:18:35,000 I can also go ahead and write like this. 357 00:18:35,000 --> 00:18:38,000 So this is what is the message that I will be getting. 358 00:18:38,000 --> 00:18:41,000 Now let me do one thing okay. 359 00:18:41,000 --> 00:18:47,000 Let me take another response or let me go ahead and again use this with message history. 360 00:18:47,000 --> 00:18:51,000 And this time I'm sending a message of human message like what's my name? 361 00:18:51,000 --> 00:18:53,000 And I'll write config is equal to config. 362 00:18:54,000 --> 00:18:59,000 Now in this case if I see the output here you'll it will say that hey your name is Chris okay. 363 00:18:59,000 --> 00:19:02,000 White is able to remember it because we are giving the same config. 364 00:19:02,000 --> 00:19:08,000 So let's say that now if I change the config, will it be able to remember this particular context? 365 00:19:08,000 --> 00:19:14,000 Okay, now here I'm just going to change the config When I say change the config I'm changing the session 366 00:19:14,000 --> 00:19:15,000 ID right. 367 00:19:15,000 --> 00:19:18,000 So I'm changing the session id okay. 368 00:19:18,000 --> 00:19:21,000 Now if I give the session ID some some different session ID. 369 00:19:21,000 --> 00:19:25,000 So for giving the session id I will again use this config over here. 370 00:19:25,000 --> 00:19:27,000 The first session id was chat one. 371 00:19:27,000 --> 00:19:29,000 Now this will be my config one. 372 00:19:29,000 --> 00:19:32,000 Okay let me just go ahead and give it. 373 00:19:32,000 --> 00:19:35,000 So this will basically be my config one okay. 374 00:19:35,000 --> 00:19:40,000 Now with respect to this particular config I will again go ahead and use my response and I'll say, 375 00:19:40,000 --> 00:19:47,000 hey, uh, with this particular message history right over here, I will go ahead and I dot invoke and 376 00:19:47,000 --> 00:19:50,000 here I will go ahead and give my message. 377 00:19:50,000 --> 00:19:52,000 The first message is nothing but my human message. 378 00:19:52,000 --> 00:19:59,000 The human message will be nothing but content is equal to I'll say, what's my name? 379 00:19:59,000 --> 00:19:59,000 Okay. 380 00:20:00,000 --> 00:20:02,000 Now see, before also I asked what's my name? 381 00:20:02,000 --> 00:20:05,000 But I give the previous session ID with respect to this particular configuration. 382 00:20:05,000 --> 00:20:07,000 But here I'm going to give my new configuration. 383 00:20:07,000 --> 00:20:14,000 So I'll go ahead and write config is equal to I will go ahead and write config one okay. 384 00:20:14,000 --> 00:20:17,000 Now do you think it will be able to remember or not. 385 00:20:17,000 --> 00:20:23,000 Just try to guess, take a guess and make sure that you write down in the comment section the Q&A section 386 00:20:23,000 --> 00:20:24,000 that we have. 387 00:20:24,000 --> 00:20:28,000 Okay, so here you will be able to see that if I go ahead and execute this response dot text. 388 00:20:28,000 --> 00:20:31,000 Uh sorry, it should be response dot content. 389 00:20:33,000 --> 00:20:36,000 The reason it is able to remember because I have given the same session ID. 390 00:20:36,000 --> 00:20:39,000 Now let me change this particular session ID I'll say chat two. 391 00:20:39,000 --> 00:20:40,000 Okay. 392 00:20:40,000 --> 00:20:41,000 Now if I go ahead and execute it. 393 00:20:41,000 --> 00:20:46,000 Hey, as an AI, I have no memory of past conversation and don't know your name. 394 00:20:46,000 --> 00:20:48,000 Now what exactly I've actually done over here. 395 00:20:48,000 --> 00:20:55,000 I've just changed the session ID now, whatever conversation I have right now, uh, it will entirely 396 00:20:55,000 --> 00:21:01,000 be saved in the session ID chat two okay, so that in the later stages, whenever we use this with message 397 00:21:01,000 --> 00:21:06,000 history, it will just give the session ID of chat two and it will retrieve the information. 398 00:21:06,000 --> 00:21:07,000 Now let me do one thing. 399 00:21:07,000 --> 00:21:10,000 Let me quickly go ahead and copy this entire thing. 400 00:21:10,000 --> 00:21:10,000 Okay. 401 00:21:11,000 --> 00:21:15,000 Let me go ahead and write and say, uh, I'll say, hey, my name is John. 402 00:21:15,000 --> 00:21:15,000 Okay. 403 00:21:15,000 --> 00:21:19,000 Hey, my name is John. 404 00:21:19,000 --> 00:21:20,000 Okay. 405 00:21:21,000 --> 00:21:23,000 And now I will go ahead and execute with this config okay. 406 00:21:23,000 --> 00:21:26,000 So if I go ahead and say, hey John, it's nice to meet you. 407 00:21:26,000 --> 00:21:33,000 Now if I go ahead and execute the same code with this particular config okay. 408 00:21:33,000 --> 00:21:36,000 Now you'll be able to see that it will understand. 409 00:21:36,000 --> 00:21:38,000 Your name is John, I remember. 410 00:21:38,000 --> 00:21:40,000 Is there anything I, uh, else I can help you with? 411 00:21:40,000 --> 00:21:47,000 Okay, so with the help of this session ID, you'll be able to see that how easily we are able to switch 412 00:21:47,000 --> 00:21:50,000 with respect to the context of conversation that we are having. 413 00:21:50,000 --> 00:21:51,000 Okay. 414 00:21:51,000 --> 00:21:53,000 So I hope you got an idea. 415 00:21:53,000 --> 00:21:57,000 I hope you got an idea with respect to this particular message history, why we are specifically using 416 00:21:57,000 --> 00:22:05,000 it, but just a easy way to understand if just highlight it over that particular class, let's say over 417 00:22:05,000 --> 00:22:07,000 here, base chat message history is there. 418 00:22:07,000 --> 00:22:11,000 This is nothing, but this is an abstract base class for storing chat message history and chat. 419 00:22:11,000 --> 00:22:13,000 Message history is basically present over here. 420 00:22:13,000 --> 00:22:16,000 This is an in memory implementation of chat message history. 421 00:22:16,000 --> 00:22:20,000 As soon as I give my session ID whenever I call this. 422 00:22:20,000 --> 00:22:23,000 So what is with underscore message underscore history. 423 00:22:23,000 --> 00:22:24,000 It is a runnable. 424 00:22:24,000 --> 00:22:26,000 It is a chain between model and get session history. 425 00:22:26,000 --> 00:22:32,000 Whenever I am going to interact with the model, it is first of all going to, uh, you know, probably 426 00:22:32,000 --> 00:22:34,000 go ahead and give your get session history. 427 00:22:34,000 --> 00:22:39,000 That basically means based on the session ID, it is going to pick up your entire chat session. 428 00:22:39,000 --> 00:22:40,000 Okay. 429 00:22:40,000 --> 00:22:42,000 And based on that it is going to remember the context. 430 00:22:42,000 --> 00:22:49,000 So this was just an idea with respect to implementing uh, you know, or how you can basically work 431 00:22:49,000 --> 00:22:52,000 with chat history, uh, whenever you're developing a chat bot. 432 00:22:52,000 --> 00:23:00,000 Now we will try to implement all this chat history, uh, in our, uh, prompt template. 433 00:23:00,000 --> 00:23:02,000 And that is what we are going to see in the next video. 434 00:23:02,000 --> 00:23:04,000 So here I hope you are able to understand this. 435 00:23:04,000 --> 00:23:06,000 This was it from my side. 436 00:23:06,000 --> 00:23:09,000 I'll see you all in the next video where I will be discussing about Prompt template. 437 00:23:09,000 --> 00:23:10,000 So thank you. 438 00:23:10,000 --> 00:23:11,000 I'll see you all in the next video.