Inspiration
We noticed that whenever we were looking to purchase something online, we always look for the cheaper option. However, it can be quite an annoying process as normally we need to look at different applications like MakeMyTrip for travel and hotels or Amazon and other retailers for the cheapest price. However, the internet is so vast, it is not totally possible to always find the cheapest price possible to purchase the good.
Also, depending on our location at times in different regions within America, there is different regional markups, that the consumer has to pay. Thus, we decided to build a new product that helps consumers to make good economic decision, and get the cheapest price possible. It also lets consumers know on what are the extra costs that the consumer needs to pay.
What it does
It is a website that is an autonomous, AI-driven shopping and price verification platform for different types of products that consumers purchase. The website, directly navigates the consumers to the checkout link, thus ensuring that consumers can directly access the site with the cheapest price for their product. It is stateless, which means that each time in the chat, there is a new message sent to Grok Bot.
How we built it
We also used the Cursor AI Agent to help us write our front end, visualize and write parts of our backend and lastly at the end run exhaustive tests on various differen functions for the backend, to check how they are working and also checking if there is proper communication between the database, n8n, Grok Bot using Curl requests. Cursor also helped us with the docker and FastAPI set up to ensure that n8n, and the front and backend is set up in our Vultur Server.
When a user submits a request on the web app, an n8n and Google Gemini orchestration pipeline translates their constraints (such as delivery ZIP code, dates, or quantities) into a custom extraction schema and self-audits the plan for accuracy before dispatching it through a FastAPI and Tiger Data job queue. First, the program checks through Tiger Data SQL database first for each day to check and quickly give pricing data for that day as it is accurate. The database also keeps updating new pricing fo the same product when asked by the consumer, to make the processes faster and more accurate. Grok Bot then claims the job and launches three isolated parallel Helper Bot browser sessions configured with distinct regional storefront settings, walking through actual cart and checkout previews to capture real shipping fees, local taxes, and mandatory surcharges before ranking and displaying the verified cheapest option.
Challenges we ran into
There were some problems when setting up n8n, as we had to migrate from the local n8n running on a computer to the Vultur server instance we made up. This meant that a lot of API connections had to be set up again, and there was no real opportunity to test and ensure that the n8n concept was problely working. Then there was also issues with connecting everything to Grok Bot due to the production URL's resulting in FastAPI needed to be set up.
This then took longer time with the code having to be written and tested to ensure it did not break anything we already coded and ensured everything communicates well with each other. Lastly, we need the perfect timing to ensure that the whole thinking process works quickly using parallel agents and optimizing the prompt for Grok Bot.
Accomplishments that we're proud of
We are proud of getting to learn, use and implement technologies like FastAPI, docker and n8n to ensure that we can properly set up and keep the system as reliable as possible. Then, we got to try Grok Bot and implement it in a sense, than just a conventional chat sense. We got to work and link Grok Bot to our n8n, in order to then utilize the power of Grok bot computing to meet the purpose of our application. We are also proud to get a proper website that is functional and properly implemented, with our main aim.
What we learned
We learned to use new technologies and implemets some DevOp's standards such as Docker, FastAPI ensuring that our code base could be easily hosted and could communicate with each other as we were using a combination of code and orchestration combined. Then we learnt how to use Grok Bot effectively in our particular case . Lastly, we had to learn more on how to understand the SQL code and also the proper n8n orchestratoon using webhooks.
What's next for TrueCart
Next for TrueCart, we are going to be adding more of chat history and authentication to make it user by user. Then after that we want to add more changes in the logic so that we can get different discount codes to find the most cheapest price. Then we want to make this for more currencies and open it up to work for different countries. Lastly, we want to add a secure payment service either using Stripe or somehting else which ensures that we are not really having to store any personal financial info and it becomes more convinient for the customer to work on our website.
