VALORANT Team Builder

Inspiration

Our passion for VALORANT, having played since day one, and a love for generative AI and data preprocessing drove us to create this project. We wanted to combine our interests in gaming and AI to build something impactful for the Esports community.

What it does

Our project is an AI-powered VALORANT team builder and data analyzer that helps team managers scout and recruit players more effectively. It enables users to scrape and analyze player stats, compare players across leagues, and build optimal teams based on data-driven insights.

How we built it

We used Streamlit for the frontend, LangChain for LLM integration, and Selenium and BeautifulSoup for extensive data scraping over several days. The platform is hosted on an EC2 instance and leverages AWS services such as ChromaDB for vector storage and S3 for managing large datasets.

Challenges we ran into

One of the biggest challenges was setting up agents and their prompt flows. We encountered constant invoke errors in the code that we couldn’t resolve, which consumed much of our time. Additionally, scraping data presented issues such as CAPTCHAs and incomplete player profiles, forcing us to implement Selenium to automate Google searches and scrape player data manually.

Accomplishments that we're proud of

We’re proud of the comprehensive dataset we built and the powerful prompt system for the AI agent. Although we couldn’t finish the frontend UI in time, the backend and data infrastructure worked seamlessly.

What we learned

We gained experience in data scraping, using AWS services like EC2 and S3, creating agent prompt flows, managing vector storage with ChromaDB, and handling data preprocessing at scale.

What's next for Valorant TeamBuilder

We plan to enhance the UI and further expand the dataset by including players not available on VLR.gg or Liquipedia, which were our main data sources. To address this, we’ll publish the platform online and invite the community to contribute by submitting VLR.gg or Tracker.gg links for esports players. Additionally, we aim to build a page for comparing players and teams, allowing users to talk to our LLM and gain deeper insights from these comparisons.