What we're building

0:00 In this video, we're gonna learn how to build a B2B agent below platform that allows us to create sales agents, customer support agents, lead qualifier agents, business conversational agents, and many more. As you can see, this is the chat page on the admin side, where I can see the interaction between the AI and human. Now, I'm going to deploy an agent. As you can see at the right, we have a deploy agent into an error react app.

0:30 To deploy an agent into an error react app, our platform provides an SDK. This allows us to embed an agent into an error react app. Now, I'm going to try to interact with the agents. Let's say I am the customer, the right here. Now I'm gonna pass a message to the agents. Hello, if I pass a hello, I want to contact your manager. This is the message that I'm passing to the agents. As I pass first, hello here, the agent's going to react immediately.

1:04 As you can see, the agent reacts. And at the left here, I can see the conversation between the AI and customer. I want to contact your manager. I want to contact your manager. And then I pass here, my name is Ben. I can pass this. And then as I pass this, at the left, I can see the conversation between the AI and human as a manager. And then the, the agent is going to react again. And as it is a memory, it's going to save everything within the database.

1:38 It is a long-term memory and short-term memory. It cannot forget anything, guys, okay? It is a strong memory. As you can see, at the left, the agent reacts. Okay, as you can see, at the left, the agent reacts. Now, let me take over the chat as a human. And as you can see, at the right, we have human take over the chat. We use WebSockets. I'm going to pass here, Hi Ben. As you can see, at the left, human.

2:11 We have a human message. How can I help you? How can I help you? I'm going to pass this to the agents. And then at the right here, as a customer, I can see, Hi Ben, I want to, I want to, I forgot my password. My password. And then as the admin, I can see the customer message, I forgot my password. And then I'm going to pass here, ask the AI. Okay, and then I'm going to quit here, the conversation.

2:51 And then the AI is going to take over the chat again. As you can see, the AI is going to take over the chat again. And here, as the customer, I can continue chatting with the agents. Okay, if I click here again, I'm going to take over the chat as human. Every input that we pass to the agents, it is traced. As you can see, we use Langsmith in order to trace every input that we're going to pass to the to the agents.

3:21 Every input, we can see where it was called by the agents, which model has been invoked by the agent, okay? And the response. We got literally everything here within Langsmith. We can see if we got latency. As you can see, this is Langsmith. It allows us to debug and trace our agents. I can see every inputs that that I have passed to the agents in here. Now, let me present you other stuff. As you can see, our app is running into Docker container.

3:54 And we got here nine services. We got the task service, which handles AI logic. We got notification service for sending email. We got the API gateway. We have the OL service. We got Redis, RabbitMQ for message broker. PG vector is a vector database. We are going to use a third-party vector database, but instead we're going to use PG vector. This is the service, and we got here MangoDB, okay? And as you can see, this is our app.

4:27 It's running into Docker container. I'm running actually nine servers. All of them are running. In order to allow these services to interact together, we use gRPC. Remote procedure call. It is a protocol buffer used in microservice architecture. Now, let's go back in here. And I'm presenting you other page. You can see this is the the pages, the agent page where I can head in and create an agent. I can pass the name, category, goal, personal, and company context.

5:07 We have here the lead page where we can see all leads tracked by the agents. We can see here sessions. We have here the chat interface, where we can see the conversation between the AI and customer. And we have here test agent page. When we finish create an agent, we need to test it. And here, I can move here and then I can start testing the agents that I have created. We have the knowledge base page, where I can upload my FAQ questions of my business.

5:27 I can upload my popular information. And then if a customer asks a given question about a particular product, the agent can go ahead and retrieve it in effect database. Okay, that's how it works. We have a lot of things to do in here. And we have here the login page. And as you can see, this is the login page. I can sign up, then pass you a random email and then pass you a password. Then we're going to click here, sign up.

5:00 They're going to send us a OTP code. As you can see, they send us a OTP code. Now I'm going to head on to mail trap in order to get the OTP code. And then we're going to capture the OTP code provided by mail trap. Now we're going to copy the OTP code and then we're going to pass it in here in order to validate our email. Then I'm going to click here, validate. And then we're going to verify our email.

5:00 Okay. That's what we're going to learn, guys. Okay? That's what we're going to learn throughout this video. We're going to use microservice architecture. All right, guys. Now we're going to get a quick overview of the system that we're going to build. As you can see, we're going to run our app within Docker. We're going to use React as the front end. And at the back end, we're going to use Node.js.