Loading...
In this video we'll put together everything we've learned in the previous videos to create an end-to-end example of a customer service assistant. We'll go through the following steps. First, we'll check the input to see if it flags the moderation API. Second, if it doesn't, we'll extract the list of products. Third, if the products are found, we'll try to look them up. And four, we'll answer the user question with the model. And finally, we'll put the answer through the moderation API. And if it is not flagged, we'll return it to the user. So we have our setup. We have this additional import. This is a Python package we'll use for a chatbot UI. And so I'm going to paste in a function "process_user_message". And I'm actually going to run an example first and then we'll talk through the function. So we have our user input that we've been using. Tell me about the smartx pro phone and the camera. Also tell me about TVs. So let's run this. First, let's run this. So as you can see, we're going through the steps to answer the user question. So the first step, the moderation step, the second step is extracting the list of products. The third step is looking up the product information. And so now with this product information, the model is trying to answer the question as we've seen in the previous videos. And then finally, it puts the response through the moderation API again to make sure it's safe to show to the user. And so this is the response that we're now familiar with. And so let's talk through a little bit of what's happening. So we have our helper function "process_user_message". It takes in the user input, which is the current message, and an array of all of the messages so far and this is when we build the chatbot UI. So the first step, checking to see if the input flags the moderation API. We covered this in a previous video. If the input is flagged, then we tell the user that we can't process the request. If it is not flagged, we try to extract the list of products as we did in the previous video. Then we try to look up the products. And in this case, if no products are found, this will just be an empty string. Then we answer the user question, so we give the conversation history and the new messages with the relevant product information. So we get the response, and then we run this response through the moderation API. If it's flagged, we tell the user that we can't provide this information. Maybe we'll say something like, let me connect you, and you could take some subsequent step. And so, let's put this all together with a nice UI, and try to have a conversation. So here we have a function that will just accumulate the messages as we interact with the assistant. And feel free to pause the video to understand in more detail how this works. And of course feel free to do the same for the really long "process_user_message" function that we just saw as well. So now we're going to paste in our code to display the chatbot UI. So if we run this, now let's try and have a conversation with the customer service assistant. So let's ask, "what TVs do you have". And under the hood, the assistant is going through all of the steps in the "process_user_message" function. So it's listed a variety of different TVs. So now let's ask, "which is the cheapest". Again, it's going through all the same steps, but this time it's passing the conversation history as well into the prompts. Okay, so it's telling us that this speaker is the cheapest TV related product we have. Interesting. Let's see what the most expensive is. So the most expensive TV is the CineView 8K TV. So let's ask for more information about it, "tell me more about it". And so we've received some more information about this television. So in this example, we've combined the techniques we've learned throughout the course to create a comprehensive system with a chain of steps that evaluates user inputs, processes them, and then checks the output. By monitoring the quality of the system across a larger number of inputs, you can alter the steps and improve the overall performance of your system. Maybe we might find that our prompts could be better for some of the steps, maybe some of the steps aren't even necessary. Maybe we'll find a better retrieval method, and so on. We'll discuss this more in the next video.