Build a Multi-Agent AI Wedding Planner with LangChain | Full Project Step-by-Step

Look at this diagram, this is what we will be building today. But not only that, we will not just be building an amazing AI multi-agent system that is basically our wedding planner, but also we will be understanding how agents and sub-agents work together in linkchain to create an amazing project. By the end of this project not only you will be having an amazing project to put on your portfolio that is basically gonna make you stand out on other engineers, but also you will be understanding how to create different other projects relying on this blueprint. You're not convinced yet? Let me show you what we will be building today. This is the application that we will be building. This is a multi-agent wedding planner. We can see a lot of information here. All this information basically are we are gathering data points that our groom and our bride would like to have in their wedding, including their names, how many guests, the budget, the currency, and a lot of other details that we will see in this project. All we have to do then is generate our wedding plan. And once we hit the button, we will see multiple agents coming together, working together, searching the web, doing all the research, and we will see this in detail step by step to accomplish one goal and one goal only, the goal that we are giving it, and this goal is to create the perfect wedding. Let me show you what will happen. I've skipped a bit just for you to understand because it took 140 seconds, 39 seconds to do this work. Why? Because this is a complete research. We are doing a complete research, all the agents coming together. We, for example, set our example what we would like to do, and here we have a complete step by step on everything that our bride and groom would like to have. The budget, the scenarios, the timeline, etc. This is, guys, a perfect, perfect project for you to have on your portfolio, and if you are new to creating a multi-agent applications, this is the right place to start. I hope you are excited to build this project because now we are going to see the full roadmap to go from zero to hero on this project. In this project, not only we are going to be copy pasting code now, but we will also be going into the details. We will be understanding how to study, what is length chain, project setup, prompt preparation, tool preparation, agent preparation. We will be making sure that we will see everything in the CLI, and then at the end we also will be adding UI. And once everything is done, congratulations, the project is done, you can add it now to your portfolio and stand out between millions of other people. If you like what you're seeing here, welcome to AI Career Lab, the place where we build the future AI engineers that companies are willing to pay top dollars for. So make sure to subscribe and not miss any of my videos. One more thing, if you like what you are seeing here and you would like to see even more complex projects, please leave a like. It's the only way for me to know that these topics are interesting for you guys. With that being said, let's start to create this amazing project. To study effectively, you need to understand and go and work with three main important points. First, I would like to tell you and advise you not to use AI when you are studying. Guys, you have to think for yourself. You have to go and it has to be hard for you to find information. You have to try your best to be able to do something. If you do not find information directly, please, it's okay, work a little bit harder, but you will be able to remember that later on. It's the only way. Do not rely on AI and coding tools to be able to generate your code. It will not help you and you will not be able to advance and become better AI engineers. Don't get me wrong, AI tools are really good and they will come in handy later on once you really are capable of doing it yourself. Second thing I would like to say, studying is like a marathon. It's not like a, it's not a sprint. So take your time and study bit by bit and don't try to rush your studying process. Go slowly but steady, which means it's better for you to do one hour a day than doing five hours one day and then stop. Okay, one tool I would like to suggest is this tool. This tool basically it's a simple, I use it in my work, it's a simple timer that basically you work for 25 minutes and after 25 minutes it will give you a five minute short break and it will repeat the cycle and after three or four times it will give you a long break of 15 minutes. This is a really good tool because it allows you to focus on what you are doing because if you, if there's something, if your phone rang you can tell yourself it's okay, I will be catching up with this notification in 25 minutes. It's okay, it's not the end of the world. So you can keep up, you can keep your focus and you can continue working. Great, one more thing I would like to add and it's the last thing before we start to understand a bit more about the project. I would like to also tell you when we are working together, look at what I'm doing. Understand first what I'm doing and then stop the video, do it yourself and then do this over and over and over. Why am I saying this? Do not just copy and paste without thinking. You need to be thinking, you need to be understanding and you need to be implementing things that we are doing together on your own. So you can do this while we are doing it step by step, you see what I do and then you try to do it for yourself. If you fail, re-watch and then re-apply. It is important, just don't just copy paste, copy paste without really understanding. So understand first and then try to do it yourself and then obviously sometimes you will have some mistakes but this is the right thing, this is the good thing, this is the right approach to do it. Do not just copy paste for no reason and make sure by the end of this project that you are 100% capable of recreating this project from scratch on your own. Great, let's start with understanding what is Lankchain. If you really want to go into the details of Lankchain, I would suggest for you to watch this video right here. I have put together a complete ultimate crash course on Lankchain that will allow you to understand what we are doing today, going into the details of Lankchain, going into what is agents, what is models, what is creating, what is initializing, what is prompting, all these kind of things are in that Lankchain crash course. I would highly recommend for you to study it before moving on this project but if you already have a bit of the initialization of Lankchain, you have a little bit of knowledge then it's perfect, you don't really have to go through that crash course but I would still recommend it if you think that your knowledge is a bit shaky and once you're done please come back here because we have an amazing project, you must finish this one, you will really really like it. Now that we finished, let's start with setting up our project. We will start by opening our VS Code. Once we have done once we opened our VS Code, we are going to go and actually create our folder. I've created a folder that is called multi-agent ACL. Once we created our folder, this is basically the folder where we are going to put a lot of, basically all of our projects. We are going to go back to our VS Code and then a lot of you guys asked me if I can make the VS Code bigger, I hope this helps and then we are going to go to the folder and then we are going to select our multi-agent ACL. ACL stands for AI Career Lab. We are going to close our AI tool because we don't need it and then we are going to open our terminal. The first thing we need to do in our terminal is actually to make sure that we have Python. Perfect, we do have Python. Next, we will be checking if we have the pip which is our packet manager and also we will be checking if we have uv which is the packet manager that I prefer and we do have it. Once we have it and if you guys don't have it, please just run this command and you will be able to use the uv as well. Just make sure that you run this command after that to see if it's good. Now that it's good, we are going to initialize our project. Perfect, I just initialized our project. We have our gitignore Python version, we have our main project.tom and the readme file. Okay, very good. I will be also adding our .env and we will also be are going to be adding this is where we are going to put our keys, our special keys. We are going to put them here and then let's put .env here so we make sure that it will never be added to our github. We need to also create our virtual environment just like that and then we created our virtual environment. Perfect, and now we need to basically activate our virtual environment and just like that we can see that we are actually now in our activation virtual environment which is perfect. Now I will be adding some of the few libraries that we will be using. Of course there are more that we are going to add on the way but that's completely fine. It's not a big deal, it happened. So let's start with linkchain, openai and yeah so .env we need to add. Okay, perfect. The packet manager is making sure that we have the right things. Now we are going to run uvsync. Very good and our main is here. It's okay for now, we can just run it to make sure that our main is running well and hello for multi-agent acr. Okay, very good. This is a good start. Let's go to our to see openai. Okay, now let's add the keys. We need the keys so let's do that and we are going to go to our .env file and inside our .env file we are going to add two important keys. The first key is going to be openai api key and the second thing we are putting taveli. I will explain what is that in a second. If you watched my previous videos you would understand but if you didn't I will explain it so don't worry. Openai key is basically the key that our agents are going to rely on in terms of the model that we are going to be using and taveli openai key is going to be the api that allows our agents to use multiple tools such as the search functionality to the internet. So we know that we have our model which is our smart model like openai and we have our agent which is also going to be relying on our model but also if the agent would like to do some search it needs a way to do that and it's going to be using taveli and the reason why I'm choosing taveli you will see in a second why I'm choosing that. We are going to go and we are going to open our openai dashboard and we are going to create a new key. I'm going to call it multi-agent I'm going to call it multi-agent ACL and then I'm going to create our secret key. I'm going to copy this key and make sure to never share this key with anyone because this is very important for the sake of this video I'm putting it but by the end well by the end of this video I'm going to delete it as well so there's no need for you there's no need for you guys to try to have it because you're not going to have it so this is one we did openai key and the second thing we're going to have sorry API key we are going here as you can see I already use it for other things and we are going to call it also multi-agent ACL and if you guys don't have if you guys don't have any of a dashboard or the ACL make sure to create accounts it's very very straightforward just create accounts and then create the API keys and make sure that you do have money included for the openai for tavli is a 100% free for the first credit and you will not use more than that for the things that you need so don't worry about that for for example for the openai dashboard I do have I think three dollars or five dollars on my API key just for the sake of actually doing my personal project okay this is perfect and now we are going to add the tavli key okay now we have the two keys here we do have our environment okay very good we do have our sync we can see our lock we can see our project it's very good okay perfect and now that we finished this part let me just remove this one I'm gonna go back to my I'm gonna go back to here and with that being done I'm going to make sure that we finished this part in a very good way very nice now let's move to our prompt preparation to move to our prompt preparation we will come to our project and we will create prompt.py this is basically where we are going to so we can use them later on in the future okay I hope you guys are keeping up with everything I'm making sure okay I hope you can see very well and also make sure to like and subscribe once again guys now let's start by we are going to add two prompts here the first prompt is going to be wedding planner agent prompt and the second one is going to be our user prompt and the the reason why I'm starting with the prompts guys is because it's the easiest the easiest part and it's good to start I like to start things from easiest to hardest because when you do that you accomplish a lot more than when you do the opposite and you're going to be demotivated so now I'm going to give it actually I'm gonna give my agent what I want it to be so I'm gonna tell it you are an expert in wedding planning I have prepared the I've actually prepared the I have actually prepared this prompt so I'm going to copy paste it here and then explain it to you and then I also have done the same thing for here so for the sake of not losing a lot of time I'm going to read it for you so what are what are we doing we are saying you are an expert in wedding planning your task is to assist the user in planning their wedding by providing a recommendation and suggestion based on their preference and requirements these are this is this these are the requirements and then we have put requirements between between two brackets and not brackets sorry I forgot the name of these and the reason why we did that because later on we are going to actually be injecting the requirement inside of our prompt to be able to get what we want because don't forget every couple is going to have different set of requirements and then we also are telling our agent that hey you have actually two subagents that can help you with your task subagent one and subagent two they are experts in searching the web for information providing relevant results so we are also telling our agent that it has two subagents so like two employees under the agent that will help you in search and they are an expert in search for information okay and we will see how these subagents will come to live in a bit and at the same time we have our user prompt that is basically just telling our telling our agent please can you help me with my wedding these are the requirements I provided okay perfect and just like that we are really starting to get into the project we have our agent prompt we have our user prompt very good now we are going to start by actually having some very interesting things so let's start by the heart of our project we have moved now to tool preparation to be able to make sure that our tool is going to work perfectly and that our agent later on is going to work also perfectly we need to actually start by making sure that our model which which we just put the api for which is our open ai model is working in a very good condition to do that we are actually going to create models.py so this is the model basically this is where we are going to actually give I'm going to call it models that's just in case I add another model later on and this is basically where we are going to give all our other files the models so we only need to change the model in one place instead of changing it multiple times and this is a really good way of working because we don't have to repeat yourself it's a very important thing in coding to try not to repeat yourself to be able to start and make sure our model is working and that our openai dashboard and account is working let's start by creating our model so to do that we will start by getting from linkchain and we are going to do chat models and then we are going to initializing the chat model and if you feel that these this information are a bit weird for you please make sure to check our my linkchain crash course that is basically going to details and what is this exactly and then we are going to get from that and we are going to get our load once we did our load we are going to run our load to be able to get our key that we just did and then i'm gonna call it openai model and then i'm gonna run the init chat model and i'm gonna make sure that i would like to pick gpt5 nano which is an extremely cheap model to work with once we do that let's print and then i'm going to actually do openai model dot invoke and then what is the color of the sky just to make sure that our model is working and everything's fine and then i'm gonna call models and then we'll see if it's gonna be fine it's taking a little bit of time i hope it's going in the right direction there's only one way to know it's by getting the response and we did have a response which is amazing however the the content that we got back is a bit there's a lot of information so i'm just going to put that content here to show you guys uh the actual response that we got and basically it's just for me i'm just actually selecting the response instead of having everything and as you can see here it depends on the time and the condition but usually during the day it's sky blue orange pink gray and white etc very good now we know that actually our model is working we are going to save this and we are moving next to our tool now that we have our model working we can actually start by creating our tool and this is basically the tool that our agents are going to use to be able to do some research if they want to okay we are going by select tools dot buy and inside our file we are going to import from lengthchain tools i'm gonna import tool and then from taveli i'm gonna import our taveli clients okay it looks like we did not add taveli so we are going to do that now and then we are going to sync our project perfect we are going to clear what we have now our vscode recognize it my client and then i'm gonna import my load very good and then at the end i'm gonna import logging because it's gonna be so nice to see what's happening in the cli once we can actually know oh okay this basically this tool is being used etc i'm going to configure my logging and then i'm gonna say logging dot basic configure level logging info this is just telling me okay give me info and then the format is going to be as follow it's a bit of it's a bit it's a bit annoying this part but it's okay this part is best just basically we are configuring how we would like our logging to look like so don't worry a lot if you don't really understand this part this is generally just a preference that i use and then one more thing message s and then level okay this is just configuring and now let's do the actual ai engineering work loading and then family client is going to be family client and then now we are going to put our tool annotation to basically tell that this is a tool we are going to define our function and once again guys if you feel like we are going too fast please do check our link chain my link chain video it goes way smoother on this and way way slower and i would like to keep this project this video actually for the people who already understand and would like to create an amazing project so now we created our function and we put the tool annotation to tell basically this is a tool that we can later on pass in our link chain agents don't forget basically we are going to use the annotation to explain what does this do the web for a given topic and return the results and one more thing here i would like to say okay i remember i used to do this just give me a second um i don't think so but okay no problem we will continue we'll continue with that and then here we are going to logging you remember we got the logging for this exact reason where we are saying searching the web for topic and then okay basically whenever an our agent is using our tool we will be logging it to understand what is our agent actually searching and this is where it's going to get really really beautiful because we can see what are sub-agents working and what is the main agent dedicating to the sub-agents and what are they searching so we are searching for this topic and then we return the search basically it's a functionality from tabuli that will basically run an api call to search the web for a specific topic and it will return the information about this specific topic very good now that we have our tool inset we created our tool let's check the tools that we do have our tools are in check and let's move to our agent preparations in our agent we are going to create two different sub-agents and these agents are basically going to be also used as tools why because we are going to give our main agent two sub-agents and we're going to tell it okay use these sub-agents as tools to do whatever you feel is good to achieve the goal that we are going to give you which is creating our wedding plan okay so agents import create agent we are also also from linkchain.tools import tool from linkchain.messages import human message from models import openai model this is basically the one that we created from tools import search web which is our functionality that we created also and at the end we're gonna import logging and we are going to basically do the same thing so we are going to take configuration of this one we are going to put it here and then we are going to create our sub-agents now sub-agents we are going to log.info and we are saying initializing agent one and then here we are going to say sub-agent one is going to be create agent the model that this agent will rely on is exactly the model that we created model tools we are going to allow it to have the search web tool and then name we are going to call it sub-agent one so this is basically the name that this is basically the name that our main agents will be able to call this one we are going to copy this and then we are going to change some new things perfect now we have two sub-agents and now let's call it tool and this tool basically is going to be delegating tasks for our agent okay so i will create and then i will explain it to you i'm gonna be like delegate sub-agent one and then we are going to give it a query and then it's going to actually give me a string back and then it's going to be like search the web for a given topic by delegating the task to sub-agent one to do so and then as usual i would like to be logging this information to be able to track what is actually happening behind the scenes going to put an f string and then we are going to be delegating query to sub-agent one and this is the query and then response is going to be a sub-agent one dot invoke we are going to give it messages and inside of messages we are going to have an array of messages and we are going to put a human message and inside the human message we are going to say that the content is the query at the end we are going to return the response messages minus one so only defer the last message i would like to have and then i would like to have the content of that last message so what are we doing guys we are creating an agent this agent is based on the model that we created and it also has a tool that is searching the web then we are creating this tool this tool is basically to delegate subject to sub-agent one so whenever our main agent would like to give a task to another agent it all it has to do it simply use this tool give the query that what the main agent would like to search this query then will be given to our sub-agent by invoking our agent one and telling it that this is what we would like to know this is what we would like to search the sub-agent then is going to search using the search web and give us the respond and this is the content that we will be receiving and then returning to the main agent to do whatever he pleases with okay we are going to do the exact same thing but for sub-agent two and the reason why we are creating two tools there's basically kind of do the same thing is because in in a wedding or in an any different project we might actually be wanting to give an agent just one thing to do at the time and this will allow our main agent to know that which topic is coming from which which agent without really having to have a lot of things to do itself so all it has to do is to think about the task and if it has to do any other task we would like to give the sub-agents to do it so it can just focus on the task that we have now that we have this we uh let me see what do we have on our here sub-agent and the main great we have our sub-agents preparation so it's good to move to our main agent now very good for our main agent i'm gonna use the main actually i'm gonna go to main and then here where the magic all the magic is going to so this is where we are going to create our main agent give our main agent the prompts and also i'm gonna give it the servants which are like sub-agents to do whatever it sees fit okay so first we're gonna get the prompt so obviously we have wedding planner and user second from linkchain.agents i'm gonna import create agent third i'm gonna get some models open ai models and this is the beauty guys of not having to repeat yourself because whenever at any point i would like to have multiple models i can simply just come here create the models that i want if right now i would like all my agents to be just based on gpt5 i can just remove this and it will work perfectly i don't need to do hundreds of changes okay this is the good thing about not repeating yourself this is the right way to code and now from agents i'm gonna import delegate to sub-agent one and delegate to sub-agent two these are the two tools that we created that are based on our sub-agent and then i'm gonna import logging and then from linkchain messages i'm gonna import human messages and human messages is basically telling the agent that this message is coming from a human for it to do the job not from an ai or not from a system okay again i'm going to repeat the basic logging configuration and then here i'm gonna be asking user for their requirements for their wedding and then i'm gonna be logging the information and this information i'm gonna be like telling basically in my cli that we are asking user for their requirements and then user requirements don't forget that in our prompts we created a part that we can inject the requirements we are going to be asking please enter your requirements and references for your wedding and then at the end i'm just gonna put slash n so we can go to the to the line another thing i'm going to be like i'm going to updating the system prompt so the system prompt is going to be updated with updated with a new requirement that we have and then it's going to be from our wedding planner format and then inside it we're gonna be requirements the prompt take the requirement and then basically just like that okay very good and next i'm gonna create our main agent and then i'm gonna be logging the information initializing main wedding planner agent and then main wedding planner agent is going to be create agent the model is going to be openai model tools are going to be delegate one delegate two name is going to be main wedding planner agent and then i would like my system prompt to be the updated system prompt very good you see guys how we're creating a really good very big project in very short amount of time because we understand how to manipulate link chain and if you feel like we are going very fast please double check the link chain course it would really help you to understand everything we are doing okay and then at the end invoke main agent so basically we are calling the main agent we are logging the info again and then we are saying that we are invoking main wedding planner agent main wedding planner asian dot invoke and then we are going to give it uh we are going to give it the messages the messages and we are going to give it a list of messages which is a human content and then the human message and then the content is going to be basically the user prompt for main okay so we are basically creating our agent on the model giving it the sub agents to use how it please giving the system prompt where i'm telling him okay i would like you to create this beautiful wedding plan and then we are running the agent by telling it hey user prompt is like please help me with my wedding and then it understand that it's a wedding planner okay very good now just as a last thing we are going to print the response of the main agent and then print and planner and then here main wedding planner and we're gonna get messages and we're gonna get the last message and only the content of the okay so let's see if this whole thing is going to work and then we will explain what happened okay so we're gonna run python main and then let's see what we will have so first of all we see that we are initializing some agents some agents too we are asking for the requirements and then here we are saying please enter requirement of preference okay i don't have a preference but i do have a very small i mean that's not small but a small budget of 10k us dollars i mean it's a definitely not small but okay so let's remove small a budget of 10k us dollars okay so we are initializing our wedding planner agents we are invoking our wedding planner agents see how beautiful it is it's we are really understanding what's going on behind the scenes so let's see what's actually going to happen are is our main agents going to use our sub agents let's see and we will know in a bit so now okay we do have an issue and the issue is that so our issue is with the messages okay very good let's solve that issue so we are they are saying that main wedding planner messages content state graph object not subscribable okay very good oh did we do a mistake did i give the right messages i did give the right messages i did give human prompt object is not subscribable okay so make sure that it is an agent that we created oh okay makes sense makes sense makes sense it's not it's actually the response of main agent that we need okay so what happened here guys is that yes of course the response is not in our agent it's actually in the response of our agent okay so i'm going to say like this and then we are going to remove this and then okay now it should be fine let's run it again it's a bit uh yes let's give it the same preference so we did get a nice response however we did not get we did not see that the sub agents have are being used so let's see what's going on in here okay so basically after i've seen after i rerun the same i didn't change anything but i did change my preference and i realized that our preference attack it was so small that it didn't really need to delegate any of the tasks however when i said my preference is places on the beach we can see that delegating query to sub agents best and you can see how the agents are working in the background and this is fascinating guys it's fascinating how these agents are working together look how sub agents is also searching for wedding venues and here's wedding destination and at the end we can see that it got all the things all the things that from sub agents one and two to come up with the experience that we have so it's such an amazing project guys i really hope that you like it because so far everything that we have here is what you are going to be to be doing as a ai engineer okay let's go here and actually finish this one and since we actually added the logins as well so i hope you like it now for the ui it is not really important for you to finish the ui part because the ui is simply just for you to see it but if you are interested i would highly recommend for you to come to my github and please come and check my project and you will see here this project that i'm just explaining to you go to our app and simply copy the whole thing because it has a lot of streamlit applications and a lot of css it's not really important for you to showcase that but it is for sure and nice to have if you are interested in me showing that please write it in the comments below i have created multiple projects so far and i didn't receive any comments saying that someone is interested in actually learning the the ui part so if it's still the case perfect i will simply be doing this often because i think not a lot of people are interested in the ui so all you have to do is bring the ui copy and make sure to star my project guys so come here and make sure you star it okay so i know you've been here go to our agent we are going to create app.py and then we are going to our project and then here we are going to say uv run streamlit app and before we are going to add streamlit you see guys how we are facing the mistakes we are learning in without really relying on ai because this is how you learn okay it's very important we are going to sync our application and then we are going to try again and here you go this is the project that i told you about you can simply generate and it will generate the wedding okay and you will have exactly what we have in the cli okay guys let's just put two more here and then finally the project is done you have created this project right here we are taking the streamlit the command line apps the wedding requirements the prompts the wedding agents the wedding agent is delegating to sub agents the sub agents are using the search web tools the search web tools are going to the api taverly we are sharing the openai model between different sub agents between our main agents we are going we are using the gpt5 nano the final wedding we are printing it in the cli we are also printing it in our ui i hope you guys really really enjoyed this project a lot i if you reach this level then really congratulations i'm really proud of you i'm really proud of the work you are doing and this is just the beginning guys i promise you this is just the beginning you can see here all the api going on in the background guys this is one of the projects that is really makes you love working with agents it makes you love working it makes you really love being an ai engineer i have a lot to teach you i have a lot to inform you so if you really like it i would really appreciate the support because if i can see more support i can probably put more effort into this because it takes hours and hours just to plan these projects to make them sure to make sure they are people are interested in these projects before even going into the details with that being said i really hope you enjoy it please subscribe please like and i'll see you on the next one and if you have any project ideas please leave them in the comments below have a good one and i'll see you in the next project happy learning